Research ArticleCoal and Gangue Underground Pneumatic Separation EffectEvaluation Influenced by Different Airflow Directions

Kehong Zheng12 Changlong Du1 Jianping Li1 and Bingjing Qiu1

1College of Mechanical and Electrical Engineering China University of Mining and Technology Xuzhou 221116 China2Department of Metallurgical Engineering College of Mines and Earth Sciences University of Utah Salt Lake CityUT 84112-0114 USA

Correspondence should be addressed to Kehong Zheng zkhcumtgmailcom

Received 23 June 2015 Accepted 9 November 2015

Academic Editor Charles C Sorrell

Copyright copy 2016 Kehong Zheng et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Coal and gangue underground pneumatic separation is of key importance for green mining Two kinds of arrangement schemesfor high-pressure value used in pneumatic separation system are proposed in this study Pneumatic separation effects are examinedunder different arrangement of high-pressure value Here theoretical pneumatic separation distance formulas of mineral particlesaffected by different airflow directions are derived and validated by a series of numerical simulations and orthogonal experimentsIn the following analysis the effects of gangue diameter (119889) conveyor velocity (V

0) and the height difference between conveyor belt

and air nozzle (ℎ119901) are mainly considered The numerical simulation and experimental results indicate that pneumatic separation

effects under the condition of 119906 and V0being in the opposite directionwill be better than that of 119906 and V

0being in the same direction

The pneumatic separation distance Δ119878 shows a decreasing trend with the increasing of the three factors The study also shows thatgangue diameter has the most significant influence on separation distance followed by conveyor velocity V

0and height difference

ℎ119901

1 Introduction

Coal has been regarded as the most widely used fuel in manycountries because of the lack of oil and gas resources which isexpected to be prolonged over decades Initial undergroundseparation for coal and gangue is an indispensable processof mine production and has great significance for coalpreparation Gangue from underground separation can beused for filling mined-out areas improving the quality ofraw coal decreasing the cost of preparation saving transportcapacity and reducing the amount of waste polluting theenvironment above the ground [1 2] Currently severaltechnologies of underground separation coal and gangue areproposed which include an automated hydraulic separationtechnology and a rotating impact method of coal gangueseparation underground [3ndash6] However underground sepa-ration technologies mentioned above have the disadvantagesof high cost being time-consuming and low automation

With the development of digital image processing tech-nology scholars mainly probed into a general machine vision

approach for online identification of coal and gangue [7ndash10]The influences of camera location particle overlap imageblur degree and conveyor velocity were investigated duringthe coal and gangue particles online transportation [11ndash14]From the above discussion machine vision process has theadvantages of real-time processing lower cost and higherintelligence

Lots of researches have conducted the online identifica-tion of coal and gangue which have proved the feasibility ofimage recognition of coal and gangue Nevertheless pneu-matic separation of coal and gangue has not been discussedadequately and has not been applied in mineral separation

Currently pneumatic separation has been widely usedin agriculture field and rare metal recycling field A cottonseed online separation system based on machine vision wasdesigned to realize automatic separation of red-brown andblack cotton seed by compressed airflow in the air [15] Guoet al proposed a series of methods to realize the recyclingof printed circuit boards (PCBs) the whole process that

Hindawi Publishing CorporationAdvances in Materials Science and EngineeringVolume 2016 Article ID 6465983 13 pageshttpdxdoiorg10115520166465983

2 Advances in Materials Science and Engineering

involves crushing and electrostatic and pneumatic separationhas formed a closed cycle that can return material andprovide salable product [16] Xu et al described an effectivemechanical process including impact crushing from printedcircuit boards and investigated the pneumatic separation formetal recovery scraps [17] Kumar et al report a simple andeco-friendly physical pneumatic separation process for therecycling of metallic values from PCBs [18 19] The effect oforifices on spherical particles was clarified by Hayashi andOki through numerical simulations of air-solid multiphaseflow in a vertical single-column pneumatic separator torealize recycle important raremetals (such as tantalum) in thePCBs ofwaste electronic equipment [20]Havlik et al focusedon studying mechanical-physical pretreatment of PCBs fromused consumer equipment followed by extraction of copperand tin from residue fractions by leaching in hydrochloricacid solutions [21]

Moreover researchers also have done lots of researchon pneumatic separation of coal particles (particle size iemainly less than 5mm) in fluidized conveyor Liu et alinvestigated the pneumatic classification behavior in a labo-ratory CCMC reactor with such a configuration by removingthe coal fraction below a given size (eg 30mm) from0 to 200mm [22 23] Yang et al attempted to carry outa systematic process analysis of fine coal preparation in avibrated gas-fluidized bed (VGFB) [24]

The research group has proved the feasibility of pneu-matic separation technology for coal and gangue yet pneu-matic separation effects influenced by different arrangementsof high-pressure value have not been presented In this papertheoretical models were established for the study of pneu-matic separation process affected by different high-pressureairflow direction Then a series of numerical simulationsand orthogonal experiments were conducted to evaluate theseparation effect under different airflow direction Finally thetheoretical models were corrected based on experiment data

2 Principle

Underground coal and gangue pneumatic separation systemwhich is shown in Figure 1 contains three parts prelimi-nary crush separation system machine version system andpneumatic separation system Preliminary crush separationsystem mainly contains roll-type crusher spiral size screenand several conveyor belts which aims to limit the granularityof coal and gangue sent into machine vision system withina range of 50mmsim100mm Machine vision system is ahighly automated technology using digital image processingtechnology to identify the size and location of gangue basedon identification algorithms then the identified informationwill be sent to pneumatic separation system through imagesensor Pneumatic separation system is controlled by com-puter to realize coal and gangue pneumatic separation

Two kinds of arrangement schemes for high-pressurevalue used in pneumatic separation system are presented inFigure 1 One is that particlesrsquo motion direction is the sameas that of high-pressure airflow the other is that particlesrsquomotion direction is opposite to that of high-pressure airflowAs shown in Figure 1 the high-pressure values 12(a) are

1

68 7

4

5

32

91011

13

14

15

16 17

18 19

2012(a)

12(b)

Figure 1 Separation system of coal and gangue (1) High-pressuregasholder (2) air compressor (3) separation equipment (4) vibrat-ing sieve (5) sieve plate (6) coal conveyor belt (7) conveyerbelt for coal and gangue recognition (8) coal (9) gangue (1011) materials conveyor belt (12) (a b) high-pressure value array(13) control equipment (14) computer (15) image acquisitionequipment (16) CCD camera (17) camera box (18) conveyor belt(19) coal conveyor belt (20) LED light array

arranged under the conveyor belt (7) in the first arrangementscheme mixed mineral materials are thrown from conveyorbelt (7) coal materials unaffected by high-pressure airflowwill fall onto conveyor belt (10) and ganguematerials affectedby high-pressure airflow will fall onto conveyor belt (11) Inthe second arrangement scheme high-pressure values 12(b)are arranged in front of conveyor belt (7) mixed mineralmaterials are thrown from conveyor belt (7) coal materialsunaffected by high-pressure airflow will fall on conveyor belt(11) and gangue materials affected by high-pressure airflowwill fall on conveyor belt (10)

3 Model Development

31 Theoretical Model The movement of mineral materialsaffected by high-pressure airflow can be divided into threestages (1) particlersquos movement before falling into airflowdomain (2) movement of particle in airflow domain (3)movement of particle after leaving the airflow domainWhenthe movement direction of mineral materials is the sameas that of high-pressure airflow coal particle trajectoryunaffected by high-pressure airflow is shown in Figure 2(a)of curve 1 and gangue particle trajectory affected by high-pressure airflow is shown in Figure 2(a) of curve 2 When themovement direction of mineral materials is opposite to thatof high-pressure airflow coal particle trajectory unaffected byhigh-pressure airflow is shown in Figure 2(b) of curve 1 andgangue particle trajectory affected by high-pressure airflow isshown in Figure 2(b) of curve 2

As shown in Figure 2 Δ119878 represents the pneumaticseparation distance of coal and gangue ℎ

119901represents the

height difference between gangue mass center on conveyorbelt and upper boundary of the airflow domain 119905

119901represents

themotion time before gangue came into the airflow domainV11990112

is the velocity in direction 119910 of gangue before falling

Advances in Materials Science and Engineering 3

Gangue particle trajectory 2

Airflow domain

Coal particle trajectory 1

hp tp

hj tj1

fSj1Sitf1

hf

S0

Sf1

ΔS

x

(a) and 119906 are in the same direction

Gangue particle trajectory 2

Airflow domain

Coal particle trajectory 1

hj tj2

Sj2

Si tf2

hfS0

Sf2

ΔS

hptp

(b) and 119906 are in the opposite direction

Figure 2 Pneumatic separation model of coal and gangue

into airflow domain and ℎ119895represents the height of airflow

domain 119878119894represents the displacement in 119909 direction of

gangue before entering airflow domain 11990511989512

represents themotion time of gangue in airflow domain and 119878

119895represents

the motion displacement in 119909 direction of gangue in airflowdomain V

11989112represents the motion velocity in 119910 direction

of gangue in airflow domain (11990511989512) represents the motion

velocity in 119909 direction of gangue in airflow domain 11990511989112

represents the motion time of gangue after leaving airflowdomain and 119878

11989112represents the motion displacement in

119909 direction of gangue after leaving airflow domain In thefollowing equations subscript 1 is used to represent whenairflow velocity and conveyor belt velocity are in the samedirection and subscript 2 is used to represent when airflowvelocity and conveyor belt velocity are in the oppositedirection

Coal and ganguematerials will be recognized bymachinevision system and the recognized image information willbe sent to pneumatic separation system through imagesensor Coal particles will fall from the conveyor belt freelywithout the effect of high-pressure airflow When the motiondirection of mineral materials is the same as that of high-pressure airflow gangue material will change its trajectoryand will be thrown significantly farther than that of coalparticle When the motion direction of mineral materials isopposite to that of high-pressure airflow ganguematerial willchange its trajectory to an opposite direction compared withthat of coal particleThemotions of coal and gangue particlesare analyzed respectively in the following three sections

When coal materials are recognized by machine visionsystem coal particles will fall from the conveyor belt freelyunaffected by high-pressure airflow 119905

0and 1198780can be obtained

through (1) and (2) Gangue motion law at this stage is thesame as coal materials 119878

119894and V119901can be also shown by (3) and

(4) where 1199050represents the motion time of coal particle and

1198780represents the displacement of coal particle in 119909 direction

1199050= radic

2 (ℎ119901+ ℎ119895+ ℎ119891)

119892 (1)

1198780= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892 (2)

119878119894= V0radic2ℎ119901

119892 (3)

V119901= 119892radic

2ℎ119901

119892 (4)

In order to calculate the acting force of high pressureon gangue particles multiple rectangular polyline is used toapproximate gangue particlersquos physical shape if the dynamicpressure head of the airflow domain is known in speciallocationThe acting force of high-pressure airflow on ganguecan be presented by

119865 = ∬119860

119875119889119889119860 =

119898

sum

119894=1

119875119889119894119860119894 (5)

where 119865 represents the acting force of the high airflowon gangue 119875

119889119894represents the high pressure in ring 119894 119860

119894

represents the area of ring 119894 and 119898 represents the totalnumber of rings

As shown in Figure 3 gangue is affected by airflowwith different directions In order to analyze the movementof gangue particle in high-pressure airflow domain airresistance and horizontal momentumrsquos increment of airflowdomain are ignored The theoretical formula of ganguedisplacement in direction 119910 can be expressed as 119904 = int11990511989512

0(V119901+

11989211990511989512)119889119905 = ℎ minus 119889 then 119905

11989512can be obtained as shown in

11990511989512

=

radicV2119901+ 2119892 (ℎ minus 119889) minus V

119901

119892 (6)

As shown in (6) gangue affected by different airflowdirection has the same motion time in airflow domainGangue will be affected by two forces after falling into high-pressure airflow domain The two forces applied on gangue

4 Advances in Materials Science and Engineering

y

xo

d2

Airflow direction

u

hj tj1

(a) and 119906 are in the same direction

y

xo

d2

Airflow direction

u

hj tj2

(b) and 119906 are in the opposite direction

Figure 3 Schematic diagram of ganguersquos motion in airflow domain

particle are gravity and high dynamic pressure Airflowdynamic pressure will convert to static pressure on conditionthat the velocity of airflow 119906 is larger than ganguersquos horizontalvelocity V

0and airflow must keep dynamic pressure 119901

119889=

12058822 The pressure difference can be expressed as Δ119901

119889=

120588(1199062minus 2)2 According to (5) the formula of airflow force

can be expressed by

119865 = ∬119901119889119860 asymp sum

119894

119901119894119860119894=1

2120588sum

119894

(1199062

119894minus 2)119860119894 (7)

When the curvature of gangue surface is not too large itcan be approximately seen as a plane so forall119906

119894= 119906 (7) can be

changed to

119865 =1

2120588 (1199062minus 2)119860 (8)

According to Newtonrsquos second law (8) can be changed to

119898 =1

2120588119860 (119906

2minus 2)

and 119906 are in same direction

119898 = minus1

2120588119860 (119906

2+ 2)

and 119906 are in opposite direction

(9)

As shown in (9) the expressions are corresponding todifferent airflow directions The general solution of displace-ment in 119909 direction of gangue moving in the airflow can beobtained by taking Laplace transform to (9) and the result isshown in

119909 (1199051198951)

=1198601205881198622+ 119898 ln (119890(21198621119906minus1198601199051198951120588119906)119898 + 1) minus 119898119862

1+ 1198601199051198951120588119906

119860120588

and 119906 are in same direction

119909 (1199051198952)

=1198601205881198624+ 2119898 ln (cos ((119860119905

1198952119906120588 minus 2119898119906119862

3) 2119898))

119860120588

and 119906 are in opposite direction

(10)

The velocity of gangue in 119909 direction in airflow domaincan be obtained by taking derivation of (10) The generalsolution is shown in

1199091015840(1199051198951) =

1199061198901198601199051198951119906120588119898

11989021198621119906119898 + 1198901198601199051198951119906120588119898

and 119906 are in same direction

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898119906119862

3

2119898)

and 119906 are in opposite direction

(11)

In order to calculate the constant coefficients 1198621 1198622 1198623

and 1198624in (10) the initial constraint conditions are as follows

the initial position of the gangue is 119909(0) = 0 and the initialvelocity of gangue is 1199091015840(0) = V

0

When and 119906 are in the same direction the generalsolution of 119909(119905

119895) and 1199091015840(119905

119895) can be obtained by (10) and (11)

1198621and 119862

2are shown in

1198621=119898 ln (119906V

0minus 1)

2119906

1198622=1198982 ln (119906V

0minus 1) minus 2119898119906 ln (119906V

0)

2119860119906120588

(12)

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown in

119909 (1199051198951)

=119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

1199091015840(1199051198951) =

1199061198901198601199051198951119906120588119898

119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898

(13)

When and 119906 are in the opposite direction the generalsolution of 119909(119905

119895) and 1199091015840(119905

119895) can be obtained by (10) and (11)

1198623and 119862

4are shown by

1198623=

arccos(119906radicV20+ 1199062)

119906

1198624=

minus2119898 ln(119906radicV20+ 1199062)

119860120588

(14)

Advances in Materials Science and Engineering 5

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown by

(15) and (16) respectively

119909 (1199051198952) =

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

(15)

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898) (16)

When gangue particles left the boundary of airflowdomain gangue particles would do flat parabolicmotionTheformula of velocity V

119891in direction 119910 is shown in

V11989112

= int

11990511989512

0

(V119901+ 119892119905) 119889119905 (17)

11990511989112

can be solved through (18) and the result is shown by(19)

ℎ119891= int

11990511989112

0

(V119891+ 119892119905) 119889119905

= int

11990511989112

0

119892119905 119889119905 + int

11990511989112

0

int

119905119895

0

(V119901+ 119892119905) 119889

2119905

(18)

11990511989112

=

minus1198921199052

11989512minus 211990511989512

V119901+ radic8119892ℎ

119891+ (119892119905211989512

+ 211990511989512

V119901)2

2119892

(19)

The displacement 11987811989112

of gangue in direction 119909 afterleaving the airflow domain can be expressed by

11987811989112

= 1199091015840(11990511989512) 11990511989112

(20)

To sum up when and 119906 are in the same direction theseparation distance Δ119878 can be calculated by

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780= V0radic2ℎ119901

119892

+119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

(21)

When and119906 are in the opposite direction the separationdistance Δ119878 can be obtained by

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892

(22)

The separation distance Δ119878 is shown by (21) and (22)which reflects the basic motion law of gangue when influ-enced by airflow field under different airflow directionsFor the limitation of assumption there exist big differ-ences between theoretical model and practical model Inorder to simplify the calculation and correct the differencebetween theoretical value and practical value parameters 119896

119899

(nonlinear correction factor) and 119896119903(linear correction factor)

are introduced 119896119899reflects the convergence rate of the fitting

function and 119896119903reflects the convergent gain and is used to

adjust the fitting effect of formula based on experimentalvalue Theoretical value will get close to experiment value byadjusting the value of 119896

119899and 119896119903 The modified formulas with

the correction factors are shown in

6 Advances in Materials Science and Engineering

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus1198601198961198991199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601198961198991199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601198961198991199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 119896119903

(23)

Δ119878 = 1198780+ 119878119895+ 119878119891minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((119860119896119899119905119895119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(119860119905119895120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

119905119895(119892119905119895+ 2V119901)minus V0radic2ℎ119901

119892minus 119896119903

(24)

32 Air-Solid Multiphase Pneumatic Separation SimulationIn order to evaluate the pneumatic separation effect undertwo kinds of arrangement schemes for high-pressure valuea ldquofixed coarse-gridrdquo fluid scheme is applied in PFC3Dfor pneumatic separation simulations In the fluid scheme550 (22 times 5 times 5 119909 119910 and 119911 directions) fluid cells arecreated in a rectangular space (119909 = [0 7000mm] 119910 =[minus400mm 400mm] and 119911 = [300mm 400mm]) whichcovers the rectangular space A pneumatic boundary shouldbe set for the fluid grid Driving forces from the fluid flow areapplied to the particles as body forces These forces are alsoadded to the fluid equations and cause change inmomentumas reflected by the change in the pressure gradient in the flowdirection

As shown in Figures 4(a) and 4(b) two available modelsare established to reduce the computation time without lossin accuracy The gangue hopper containing 300 balls is builtat the top right of conveying belt 1 it is aimed at reducing thecomputation timeThe front view of the simulationmodels atan initial stage under different airflow directions is shown inFigures 4(a) and 4(b) respectively

In Figure 4(a) the velocity V0of conveyor belt 1 and

airflow velocity 119906 are in opposite directions When ganguefalls into the airflow domain gangue particle will change itstrajectory to an opposite direction and fall onto conveyor belt2 The conveyor belt 2 has the same movement direction asairflow velocity finally gangue particles will be transportedto collecting box by conveyor belt 2 In Figure 4(b) conveyorbelt V0and airflow velocity 119906 are in the same directionWhen

gangue falls into the airflow domain gangue particle willthrow much farther and fall on conveyor belt 2 The motionof conveyor belt 2 has the same motion direction as conveyorbelt 1 finally the gangue particles are transported to collectingbox by conveyor belt 2

In the fluid scheme a pneumatic boundary is set for twoavailable models During the simulations an approximation

is made by specifying the velocity boundary at the rightend of the model and a pressure boundary as 00 Pa at leftend with 119909 = 0mm The slip boundary in which thefluid velocity parallel to the wall surface is nonzero at thewall surface is specified at the surrounding four walls Insimulation as shown by Figure 4(b) an approximation ismade by specifying the velocity boundary at the left end ofmodel and a pressure boundary as 00 Pa at the right endwith 119909 = 7000mm The initial setup of the slip boundaryin Figure 4(b) is the same as that shown in Figure 4(a) andmaterial properties are shown in Table 1

As shown in Figure 4(a) airflow is injectedwith a velocityof 300ms from the negative direction of 119909 at the startingpoint while in Figure 4(b) air is injected from the positivedirection of 119909 When the pneumatic boundaries are appliedat the initial stage the two side walls applied to confine theassembly are removed simultaneously Figure 5 shows thefront view of simulation result from the initial stage to 1 sec

It can be obtained from Figure 5 that gangue particleshave different motion trajectories under different airflowvelocity As shown in Figures 5(a) and 5(b) gangue particleswill do horizontal projectile motion before entering airflowdomain as can be seen at 119905 = 02 s and 119905 = 04 s As shown inFigure 5(a) velocities of conveyor belt 1 V

0and airflow 119906 are in

the opposite direction and gangue particles will change theirmotion trajectories to an opposite direction compared withthat of coal without the effect of airflow which can be seen at119905 = 06 s 119905 = 08 s and 119905 = 10 s As shown in Figure 5(b)velocities of the conveyor belt 1 V

0and airflow 119906 are in the

same direction and gangue particles will be blown muchfarther than that of coal without being affected by airflow ascan be seen at 119905 = 06 s 119905 = 08 s and 119905 = 10 s

As can be seen in Figures 5(a) and 5(b) gangue particleswill change their trajectories in areaA and areaCAreas B andD as shown in Figures 5(a) and 5(b) illustrated that gangueparticles with smaller diameters can be blown much farther

Advances in Materials Science and Engineering 7

Gangue hopperConveying belt 1

Conveying belt 2Collecting box

Fluid boundary

hp

0

u

(a) V0 and 119906 in the opposite direction

Gangue hopper

Conveying belt 1

Collecting box

Fluid boundary

Conveying belt 2

hp

0

u

(b) V0 and 119906 in the same direction

Figure 4 Air-solid multiphase pneumatic separation

BA

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(a) V0 and 119906 are in the opposite direction

DC

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(b) V0 and 119906 are in the same direction

Figure 5 Front view (initial stage to 1 sec apparent velocity plusmn260ms)

Table 1 Materials properties

Parameter Value UnitsBall

Diameter 50sim100 mmNumber 300 mdashDensity 2700 kgm3

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 07 mdash

AirDensity 1205 kgm3

Viscosity 18 times 10minus6 PasdotsWall

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 03 mdash

In order to study the pneumatic separation distanceinfluenced by different airflow directions airflow velocitiesplusmn300ms are chosen for the research of separation effectConveyor velocity V

0and height difference ℎ

119901between con-

veyor belt and air nozzle are kept constant The relationshipsbetween separation distances and particle diameters underdifferent airflow velocity directions are shown in Figure 6

45 50 55 60 65 70 75 80 85 90 95 100 10514

16

18

20

22

24

26

28

30

32

ΔS

(m)

d (mm)u = minus300ms and 0 = 1msu = 300ms and 0 = 1ms

Figure 6The relationship between particle diameter and separationdistance under different direction airflow velocity (119906 = plusmn300msV0= +1ms)

As can be seen in Figure 6 separation distance decreaseswith the increase of particle diameter It also can be con-cluded that the separation effect under airflow velocityminus300ms is better than that under airflow velocity 300ms

8 Advances in Materials Science and Engineering

Table 2 Levels of factors

Level Factor119860 gangue diameter 119889 (mm) 119861 height difference ℎ

119901(m) 119862 conveyor belt velocity V

0(ms)

1 50 040 0502 80 080 1003 100 110 200

Table 3 Experimental results of separation effect

Experiment number 119860 119861 119862Separation distance Δ119878 (m)

119906 and V0are in the opposite direction 119906 and V

0are in the same direction

1 50 04 05 305 2212 50 08 10 22 1143 50 11 20 205 0784 80 04 10 215 1085 80 08 20 148 0656 80 11 05 125 0427 100 04 20 185 0988 100 08 05 155 0379 100 11 10 140 028

The separation effect also can be analyzed from (9) dynamicpressure differenceΔ119875

119889can be expressed as 120588(1199062minus2)2when

119906 and V0are in the same direction while when 119906 and V

0are in

the opposite direction the dynamic pressure difference Δ119875119889

can be expressed as 120588(1199062 + 2)2 Thus it can be obtainedfrom the above analysis that separation effect under airflowvelocity minus300ms is significantly better than that of underairflow velocity 300ms

33 Orthogonal Experiment of Pneumatic Separation Digitalimage processing technology has been used to identify thetarget of various patterns of coal and gangue in undergroundpneumatic separation system Before mineral materials aresent to machine vision system coal and gangue have beencrushed to 100mm by impact crusher The size of materialsis ranging from 50mm to 100mm These materials are sentby the conveyor belt for coal and gangue digital imageinformation identification (as shown in Figure 1)

In this study separation distance Δ119878 (m) is selectedas the primary index to evaluate the pneumatic separationeffect Pneumatic separation influenced by different airflowdirection can be achieved by changing the arrangement ofhigh-pressure value The pneumatic separation test systemmainly consists of conveyor belt queuing system machinevision system control system and high-pressure air injectionsystem When coal and gangue materials fell down from theconveyor belt gangue will be identified by machine visionsystem and the information is transported to control systemthrough image sensor to drive the electromagnetic valueopenThus coal gangue pneumatic separation is realizedTheair compressor used in pneumatic separation testing systemis LG-6510 its working pressure is 10Mpa and certifiedcapacity is 65m3min

From the analysis shown in Section 3 the conveyorvelocity V

0 height difference ℎ

119901 and gangue diameter 119889

are selected as the three factors Factors and levels arelisted in Table 2 According to the identified level of factorsorthogonal table L

9(34) is applied in the test Orthogonal

experiment arrangement and results are shown in Table 3

4 Result and Discussion

41 Variance Analysis and Range Analysis Based on Orthogo-nal Test As can be seen from Table 2 each factor at differentlevels is approximate linearity so the method of regressionanalysis is to obtain the linear function relationship betweenthem appropriately [25] Thus the regression equationsof coal gangue separation distance under different airflowdirection are derived Equation (25) represents the regressionequationwhen 119906 and V

0are in the opposite direction and (26)

represents the regression equation when 119906 and V0are in the

same direction where 119910 is the separation distance of coal andgangue 119909

1represents the diameter of gangue 119909

2is the height

difference and 1199093is the velocity of conveyor belt

119910 = 467 minus 00191199091minus 1773119909

2+ 0092119909

3 (25)

119910 = 360 minus 0021199091minus 1599119909

2minus 014119909

3 (26)

Variance analysis is carried out on the regression equationto make significance test and the results are shown inTables 4 and 5 In order to determine the optimal pneumaticseparation solutions comparative analyses of the rangesbetween various levels of each factor are carried out Therange analysis is shown in Table 6

In Table 6 subscripts 1 and 2 used in influence factors(119860 119861 119862) represent the fact that 119906 and V

0are in the opposite

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

2 Advances in Materials Science and Engineering

involves crushing and electrostatic and pneumatic separationhas formed a closed cycle that can return material andprovide salable product [16] Xu et al described an effectivemechanical process including impact crushing from printedcircuit boards and investigated the pneumatic separation formetal recovery scraps [17] Kumar et al report a simple andeco-friendly physical pneumatic separation process for therecycling of metallic values from PCBs [18 19] The effect oforifices on spherical particles was clarified by Hayashi andOki through numerical simulations of air-solid multiphaseflow in a vertical single-column pneumatic separator torealize recycle important raremetals (such as tantalum) in thePCBs ofwaste electronic equipment [20]Havlik et al focusedon studying mechanical-physical pretreatment of PCBs fromused consumer equipment followed by extraction of copperand tin from residue fractions by leaching in hydrochloricacid solutions [21]

Moreover researchers also have done lots of researchon pneumatic separation of coal particles (particle size iemainly less than 5mm) in fluidized conveyor Liu et alinvestigated the pneumatic classification behavior in a labo-ratory CCMC reactor with such a configuration by removingthe coal fraction below a given size (eg 30mm) from0 to 200mm [22 23] Yang et al attempted to carry outa systematic process analysis of fine coal preparation in avibrated gas-fluidized bed (VGFB) [24]

The research group has proved the feasibility of pneu-matic separation technology for coal and gangue yet pneu-matic separation effects influenced by different arrangementsof high-pressure value have not been presented In this papertheoretical models were established for the study of pneu-matic separation process affected by different high-pressureairflow direction Then a series of numerical simulationsand orthogonal experiments were conducted to evaluate theseparation effect under different airflow direction Finally thetheoretical models were corrected based on experiment data

2 Principle

Underground coal and gangue pneumatic separation systemwhich is shown in Figure 1 contains three parts prelimi-nary crush separation system machine version system andpneumatic separation system Preliminary crush separationsystem mainly contains roll-type crusher spiral size screenand several conveyor belts which aims to limit the granularityof coal and gangue sent into machine vision system withina range of 50mmsim100mm Machine vision system is ahighly automated technology using digital image processingtechnology to identify the size and location of gangue basedon identification algorithms then the identified informationwill be sent to pneumatic separation system through imagesensor Pneumatic separation system is controlled by com-puter to realize coal and gangue pneumatic separation

Two kinds of arrangement schemes for high-pressurevalue used in pneumatic separation system are presented inFigure 1 One is that particlesrsquo motion direction is the sameas that of high-pressure airflow the other is that particlesrsquomotion direction is opposite to that of high-pressure airflowAs shown in Figure 1 the high-pressure values 12(a) are

1

68 7

4

5

32

91011

13

14

15

16 17

18 19

2012(a)

12(b)

Figure 1 Separation system of coal and gangue (1) High-pressuregasholder (2) air compressor (3) separation equipment (4) vibrat-ing sieve (5) sieve plate (6) coal conveyor belt (7) conveyerbelt for coal and gangue recognition (8) coal (9) gangue (1011) materials conveyor belt (12) (a b) high-pressure value array(13) control equipment (14) computer (15) image acquisitionequipment (16) CCD camera (17) camera box (18) conveyor belt(19) coal conveyor belt (20) LED light array

arranged under the conveyor belt (7) in the first arrangementscheme mixed mineral materials are thrown from conveyorbelt (7) coal materials unaffected by high-pressure airflowwill fall onto conveyor belt (10) and ganguematerials affectedby high-pressure airflow will fall onto conveyor belt (11) Inthe second arrangement scheme high-pressure values 12(b)are arranged in front of conveyor belt (7) mixed mineralmaterials are thrown from conveyor belt (7) coal materialsunaffected by high-pressure airflow will fall on conveyor belt(11) and gangue materials affected by high-pressure airflowwill fall on conveyor belt (10)

3 Model Development

31 Theoretical Model The movement of mineral materialsaffected by high-pressure airflow can be divided into threestages (1) particlersquos movement before falling into airflowdomain (2) movement of particle in airflow domain (3)movement of particle after leaving the airflow domainWhenthe movement direction of mineral materials is the sameas that of high-pressure airflow coal particle trajectoryunaffected by high-pressure airflow is shown in Figure 2(a)of curve 1 and gangue particle trajectory affected by high-pressure airflow is shown in Figure 2(a) of curve 2 When themovement direction of mineral materials is opposite to thatof high-pressure airflow coal particle trajectory unaffected byhigh-pressure airflow is shown in Figure 2(b) of curve 1 andgangue particle trajectory affected by high-pressure airflow isshown in Figure 2(b) of curve 2

As shown in Figure 2 Δ119878 represents the pneumaticseparation distance of coal and gangue ℎ

119901represents the

height difference between gangue mass center on conveyorbelt and upper boundary of the airflow domain 119905

119901represents

themotion time before gangue came into the airflow domainV11990112

is the velocity in direction 119910 of gangue before falling

Advances in Materials Science and Engineering 3

Gangue particle trajectory 2

Airflow domain

Coal particle trajectory 1

hp tp

hj tj1

fSj1Sitf1

hf

S0

Sf1

ΔS

x

(a) and 119906 are in the same direction

Gangue particle trajectory 2

Airflow domain

Coal particle trajectory 1

hj tj2

Sj2

Si tf2

hfS0

Sf2

ΔS

hptp

(b) and 119906 are in the opposite direction

Figure 2 Pneumatic separation model of coal and gangue

into airflow domain and ℎ119895represents the height of airflow

domain 119878119894represents the displacement in 119909 direction of

gangue before entering airflow domain 11990511989512

represents themotion time of gangue in airflow domain and 119878

119895represents

the motion displacement in 119909 direction of gangue in airflowdomain V

11989112represents the motion velocity in 119910 direction

of gangue in airflow domain (11990511989512) represents the motion

velocity in 119909 direction of gangue in airflow domain 11990511989112

represents the motion time of gangue after leaving airflowdomain and 119878

11989112represents the motion displacement in

119909 direction of gangue after leaving airflow domain In thefollowing equations subscript 1 is used to represent whenairflow velocity and conveyor belt velocity are in the samedirection and subscript 2 is used to represent when airflowvelocity and conveyor belt velocity are in the oppositedirection

Coal and ganguematerials will be recognized bymachinevision system and the recognized image information willbe sent to pneumatic separation system through imagesensor Coal particles will fall from the conveyor belt freelywithout the effect of high-pressure airflow When the motiondirection of mineral materials is the same as that of high-pressure airflow gangue material will change its trajectoryand will be thrown significantly farther than that of coalparticle When the motion direction of mineral materials isopposite to that of high-pressure airflow ganguematerial willchange its trajectory to an opposite direction compared withthat of coal particleThemotions of coal and gangue particlesare analyzed respectively in the following three sections

When coal materials are recognized by machine visionsystem coal particles will fall from the conveyor belt freelyunaffected by high-pressure airflow 119905

0and 1198780can be obtained

through (1) and (2) Gangue motion law at this stage is thesame as coal materials 119878

119894and V119901can be also shown by (3) and

(4) where 1199050represents the motion time of coal particle and

1198780represents the displacement of coal particle in 119909 direction

1199050= radic

2 (ℎ119901+ ℎ119895+ ℎ119891)

119892 (1)

1198780= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892 (2)

119878119894= V0radic2ℎ119901

119892 (3)

V119901= 119892radic

2ℎ119901

119892 (4)

In order to calculate the acting force of high pressureon gangue particles multiple rectangular polyline is used toapproximate gangue particlersquos physical shape if the dynamicpressure head of the airflow domain is known in speciallocationThe acting force of high-pressure airflow on ganguecan be presented by

119865 = ∬119860

119875119889119889119860 =

119898

sum

119894=1

119875119889119894119860119894 (5)

where 119865 represents the acting force of the high airflowon gangue 119875

119889119894represents the high pressure in ring 119894 119860

119894

represents the area of ring 119894 and 119898 represents the totalnumber of rings

As shown in Figure 3 gangue is affected by airflowwith different directions In order to analyze the movementof gangue particle in high-pressure airflow domain airresistance and horizontal momentumrsquos increment of airflowdomain are ignored The theoretical formula of ganguedisplacement in direction 119910 can be expressed as 119904 = int11990511989512

0(V119901+

11989211990511989512)119889119905 = ℎ minus 119889 then 119905

11989512can be obtained as shown in

11990511989512

=

radicV2119901+ 2119892 (ℎ minus 119889) minus V

119901

119892 (6)

As shown in (6) gangue affected by different airflowdirection has the same motion time in airflow domainGangue will be affected by two forces after falling into high-pressure airflow domain The two forces applied on gangue

4 Advances in Materials Science and Engineering

y

xo

d2

Airflow direction

u

hj tj1

(a) and 119906 are in the same direction

y

xo

d2

Airflow direction

u

hj tj2

(b) and 119906 are in the opposite direction

Figure 3 Schematic diagram of ganguersquos motion in airflow domain

particle are gravity and high dynamic pressure Airflowdynamic pressure will convert to static pressure on conditionthat the velocity of airflow 119906 is larger than ganguersquos horizontalvelocity V

0and airflow must keep dynamic pressure 119901

119889=

12058822 The pressure difference can be expressed as Δ119901

119889=

120588(1199062minus 2)2 According to (5) the formula of airflow force

can be expressed by

119865 = ∬119901119889119860 asymp sum

119894

119901119894119860119894=1

2120588sum

119894

(1199062

119894minus 2)119860119894 (7)

When the curvature of gangue surface is not too large itcan be approximately seen as a plane so forall119906

119894= 119906 (7) can be

changed to

119865 =1

2120588 (1199062minus 2)119860 (8)

According to Newtonrsquos second law (8) can be changed to

119898 =1

2120588119860 (119906

2minus 2)

and 119906 are in same direction

119898 = minus1

2120588119860 (119906

2+ 2)

and 119906 are in opposite direction

(9)

As shown in (9) the expressions are corresponding todifferent airflow directions The general solution of displace-ment in 119909 direction of gangue moving in the airflow can beobtained by taking Laplace transform to (9) and the result isshown in

119909 (1199051198951)

=1198601205881198622+ 119898 ln (119890(21198621119906minus1198601199051198951120588119906)119898 + 1) minus 119898119862

1+ 1198601199051198951120588119906

119860120588

and 119906 are in same direction

119909 (1199051198952)

=1198601205881198624+ 2119898 ln (cos ((119860119905

1198952119906120588 minus 2119898119906119862

3) 2119898))

119860120588

and 119906 are in opposite direction

(10)

The velocity of gangue in 119909 direction in airflow domaincan be obtained by taking derivation of (10) The generalsolution is shown in

1199091015840(1199051198951) =

1199061198901198601199051198951119906120588119898

11989021198621119906119898 + 1198901198601199051198951119906120588119898

and 119906 are in same direction

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898119906119862

3

2119898)

and 119906 are in opposite direction

(11)

In order to calculate the constant coefficients 1198621 1198622 1198623

and 1198624in (10) the initial constraint conditions are as follows

the initial position of the gangue is 119909(0) = 0 and the initialvelocity of gangue is 1199091015840(0) = V

0

When and 119906 are in the same direction the generalsolution of 119909(119905

119895) and 1199091015840(119905

119895) can be obtained by (10) and (11)

1198621and 119862

2are shown in

1198621=119898 ln (119906V

0minus 1)

2119906

1198622=1198982 ln (119906V

0minus 1) minus 2119898119906 ln (119906V

0)

2119860119906120588

(12)

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown in

119909 (1199051198951)

=119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

1199091015840(1199051198951) =

1199061198901198601199051198951119906120588119898

119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898

(13)

When and 119906 are in the opposite direction the generalsolution of 119909(119905

119895) and 1199091015840(119905

119895) can be obtained by (10) and (11)

1198623and 119862

4are shown by

1198623=

arccos(119906radicV20+ 1199062)

119906

1198624=

minus2119898 ln(119906radicV20+ 1199062)

119860120588

(14)

Advances in Materials Science and Engineering 5

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown by

(15) and (16) respectively

119909 (1199051198952) =

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

(15)

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898) (16)

When gangue particles left the boundary of airflowdomain gangue particles would do flat parabolicmotionTheformula of velocity V

119891in direction 119910 is shown in

V11989112

= int

11990511989512

0

(V119901+ 119892119905) 119889119905 (17)

11990511989112

can be solved through (18) and the result is shown by(19)

ℎ119891= int

11990511989112

0

(V119891+ 119892119905) 119889119905

= int

11990511989112

0

119892119905 119889119905 + int

11990511989112

0

int

119905119895

0

(V119901+ 119892119905) 119889

2119905

(18)

11990511989112

=

minus1198921199052

11989512minus 211990511989512

V119901+ radic8119892ℎ

119891+ (119892119905211989512

+ 211990511989512

V119901)2

2119892

(19)

The displacement 11987811989112

of gangue in direction 119909 afterleaving the airflow domain can be expressed by

11987811989112

= 1199091015840(11990511989512) 11990511989112

(20)

To sum up when and 119906 are in the same direction theseparation distance Δ119878 can be calculated by

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780= V0radic2ℎ119901

119892

+119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

(21)

When and119906 are in the opposite direction the separationdistance Δ119878 can be obtained by

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892

(22)

The separation distance Δ119878 is shown by (21) and (22)which reflects the basic motion law of gangue when influ-enced by airflow field under different airflow directionsFor the limitation of assumption there exist big differ-ences between theoretical model and practical model Inorder to simplify the calculation and correct the differencebetween theoretical value and practical value parameters 119896

119899

(nonlinear correction factor) and 119896119903(linear correction factor)

are introduced 119896119899reflects the convergence rate of the fitting

function and 119896119903reflects the convergent gain and is used to

adjust the fitting effect of formula based on experimentalvalue Theoretical value will get close to experiment value byadjusting the value of 119896

119899and 119896119903 The modified formulas with

the correction factors are shown in

6 Advances in Materials Science and Engineering

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus1198601198961198991199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601198961198991199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601198961198991199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 119896119903

(23)

Δ119878 = 1198780+ 119878119895+ 119878119891minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((119860119896119899119905119895119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(119860119905119895120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

119905119895(119892119905119895+ 2V119901)minus V0radic2ℎ119901

119892minus 119896119903

(24)

32 Air-Solid Multiphase Pneumatic Separation SimulationIn order to evaluate the pneumatic separation effect undertwo kinds of arrangement schemes for high-pressure valuea ldquofixed coarse-gridrdquo fluid scheme is applied in PFC3Dfor pneumatic separation simulations In the fluid scheme550 (22 times 5 times 5 119909 119910 and 119911 directions) fluid cells arecreated in a rectangular space (119909 = [0 7000mm] 119910 =[minus400mm 400mm] and 119911 = [300mm 400mm]) whichcovers the rectangular space A pneumatic boundary shouldbe set for the fluid grid Driving forces from the fluid flow areapplied to the particles as body forces These forces are alsoadded to the fluid equations and cause change inmomentumas reflected by the change in the pressure gradient in the flowdirection

As shown in Figures 4(a) and 4(b) two available modelsare established to reduce the computation time without lossin accuracy The gangue hopper containing 300 balls is builtat the top right of conveying belt 1 it is aimed at reducing thecomputation timeThe front view of the simulationmodels atan initial stage under different airflow directions is shown inFigures 4(a) and 4(b) respectively

In Figure 4(a) the velocity V0of conveyor belt 1 and

airflow velocity 119906 are in opposite directions When ganguefalls into the airflow domain gangue particle will change itstrajectory to an opposite direction and fall onto conveyor belt2 The conveyor belt 2 has the same movement direction asairflow velocity finally gangue particles will be transportedto collecting box by conveyor belt 2 In Figure 4(b) conveyorbelt V0and airflow velocity 119906 are in the same directionWhen

gangue falls into the airflow domain gangue particle willthrow much farther and fall on conveyor belt 2 The motionof conveyor belt 2 has the same motion direction as conveyorbelt 1 finally the gangue particles are transported to collectingbox by conveyor belt 2

In the fluid scheme a pneumatic boundary is set for twoavailable models During the simulations an approximation

is made by specifying the velocity boundary at the rightend of the model and a pressure boundary as 00 Pa at leftend with 119909 = 0mm The slip boundary in which thefluid velocity parallel to the wall surface is nonzero at thewall surface is specified at the surrounding four walls Insimulation as shown by Figure 4(b) an approximation ismade by specifying the velocity boundary at the left end ofmodel and a pressure boundary as 00 Pa at the right endwith 119909 = 7000mm The initial setup of the slip boundaryin Figure 4(b) is the same as that shown in Figure 4(a) andmaterial properties are shown in Table 1

As shown in Figure 4(a) airflow is injectedwith a velocityof 300ms from the negative direction of 119909 at the startingpoint while in Figure 4(b) air is injected from the positivedirection of 119909 When the pneumatic boundaries are appliedat the initial stage the two side walls applied to confine theassembly are removed simultaneously Figure 5 shows thefront view of simulation result from the initial stage to 1 sec

It can be obtained from Figure 5 that gangue particleshave different motion trajectories under different airflowvelocity As shown in Figures 5(a) and 5(b) gangue particleswill do horizontal projectile motion before entering airflowdomain as can be seen at 119905 = 02 s and 119905 = 04 s As shown inFigure 5(a) velocities of conveyor belt 1 V

0and airflow 119906 are in

the opposite direction and gangue particles will change theirmotion trajectories to an opposite direction compared withthat of coal without the effect of airflow which can be seen at119905 = 06 s 119905 = 08 s and 119905 = 10 s As shown in Figure 5(b)velocities of the conveyor belt 1 V

0and airflow 119906 are in the

same direction and gangue particles will be blown muchfarther than that of coal without being affected by airflow ascan be seen at 119905 = 06 s 119905 = 08 s and 119905 = 10 s

As can be seen in Figures 5(a) and 5(b) gangue particleswill change their trajectories in areaA and areaCAreas B andD as shown in Figures 5(a) and 5(b) illustrated that gangueparticles with smaller diameters can be blown much farther

Advances in Materials Science and Engineering 7

Gangue hopperConveying belt 1

Conveying belt 2Collecting box

Fluid boundary

hp

0

u

(a) V0 and 119906 in the opposite direction

Gangue hopper

Conveying belt 1

Collecting box

Fluid boundary

Conveying belt 2

hp

0

u

(b) V0 and 119906 in the same direction

Figure 4 Air-solid multiphase pneumatic separation

BA

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(a) V0 and 119906 are in the opposite direction

DC

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(b) V0 and 119906 are in the same direction

Figure 5 Front view (initial stage to 1 sec apparent velocity plusmn260ms)

Table 1 Materials properties

Parameter Value UnitsBall

Diameter 50sim100 mmNumber 300 mdashDensity 2700 kgm3

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 07 mdash

AirDensity 1205 kgm3

Viscosity 18 times 10minus6 PasdotsWall

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 03 mdash

In order to study the pneumatic separation distanceinfluenced by different airflow directions airflow velocitiesplusmn300ms are chosen for the research of separation effectConveyor velocity V

0and height difference ℎ

119901between con-

veyor belt and air nozzle are kept constant The relationshipsbetween separation distances and particle diameters underdifferent airflow velocity directions are shown in Figure 6

45 50 55 60 65 70 75 80 85 90 95 100 10514

16

18

20

22

24

26

28

30

32

ΔS

(m)

d (mm)u = minus300ms and 0 = 1msu = 300ms and 0 = 1ms

Figure 6The relationship between particle diameter and separationdistance under different direction airflow velocity (119906 = plusmn300msV0= +1ms)

As can be seen in Figure 6 separation distance decreaseswith the increase of particle diameter It also can be con-cluded that the separation effect under airflow velocityminus300ms is better than that under airflow velocity 300ms

8 Advances in Materials Science and Engineering

Table 2 Levels of factors

Level Factor119860 gangue diameter 119889 (mm) 119861 height difference ℎ

119901(m) 119862 conveyor belt velocity V

0(ms)

1 50 040 0502 80 080 1003 100 110 200

Table 3 Experimental results of separation effect

Experiment number 119860 119861 119862Separation distance Δ119878 (m)

119906 and V0are in the opposite direction 119906 and V

0are in the same direction

1 50 04 05 305 2212 50 08 10 22 1143 50 11 20 205 0784 80 04 10 215 1085 80 08 20 148 0656 80 11 05 125 0427 100 04 20 185 0988 100 08 05 155 0379 100 11 10 140 028

The separation effect also can be analyzed from (9) dynamicpressure differenceΔ119875

119889can be expressed as 120588(1199062minus2)2when

119906 and V0are in the same direction while when 119906 and V

0are in

the opposite direction the dynamic pressure difference Δ119875119889

can be expressed as 120588(1199062 + 2)2 Thus it can be obtainedfrom the above analysis that separation effect under airflowvelocity minus300ms is significantly better than that of underairflow velocity 300ms

33 Orthogonal Experiment of Pneumatic Separation Digitalimage processing technology has been used to identify thetarget of various patterns of coal and gangue in undergroundpneumatic separation system Before mineral materials aresent to machine vision system coal and gangue have beencrushed to 100mm by impact crusher The size of materialsis ranging from 50mm to 100mm These materials are sentby the conveyor belt for coal and gangue digital imageinformation identification (as shown in Figure 1)

In this study separation distance Δ119878 (m) is selectedas the primary index to evaluate the pneumatic separationeffect Pneumatic separation influenced by different airflowdirection can be achieved by changing the arrangement ofhigh-pressure value The pneumatic separation test systemmainly consists of conveyor belt queuing system machinevision system control system and high-pressure air injectionsystem When coal and gangue materials fell down from theconveyor belt gangue will be identified by machine visionsystem and the information is transported to control systemthrough image sensor to drive the electromagnetic valueopenThus coal gangue pneumatic separation is realizedTheair compressor used in pneumatic separation testing systemis LG-6510 its working pressure is 10Mpa and certifiedcapacity is 65m3min

From the analysis shown in Section 3 the conveyorvelocity V

0 height difference ℎ

119901 and gangue diameter 119889

are selected as the three factors Factors and levels arelisted in Table 2 According to the identified level of factorsorthogonal table L

9(34) is applied in the test Orthogonal

experiment arrangement and results are shown in Table 3

4 Result and Discussion

41 Variance Analysis and Range Analysis Based on Orthogo-nal Test As can be seen from Table 2 each factor at differentlevels is approximate linearity so the method of regressionanalysis is to obtain the linear function relationship betweenthem appropriately [25] Thus the regression equationsof coal gangue separation distance under different airflowdirection are derived Equation (25) represents the regressionequationwhen 119906 and V

0are in the opposite direction and (26)

represents the regression equation when 119906 and V0are in the

same direction where 119910 is the separation distance of coal andgangue 119909

1represents the diameter of gangue 119909

2is the height

difference and 1199093is the velocity of conveyor belt

119910 = 467 minus 00191199091minus 1773119909

2+ 0092119909

3 (25)

119910 = 360 minus 0021199091minus 1599119909

2minus 014119909

3 (26)

Variance analysis is carried out on the regression equationto make significance test and the results are shown inTables 4 and 5 In order to determine the optimal pneumaticseparation solutions comparative analyses of the rangesbetween various levels of each factor are carried out Therange analysis is shown in Table 6

In Table 6 subscripts 1 and 2 used in influence factors(119860 119861 119862) represent the fact that 119906 and V

0are in the opposite

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

Advances in Materials Science and Engineering 3

Gangue particle trajectory 2

Airflow domain

Coal particle trajectory 1

hp tp

hj tj1

fSj1Sitf1

hf

S0

Sf1

ΔS

x

(a) and 119906 are in the same direction

Gangue particle trajectory 2

Airflow domain

Coal particle trajectory 1

hj tj2

Sj2

Si tf2

hfS0

Sf2

ΔS

hptp

(b) and 119906 are in the opposite direction

Figure 2 Pneumatic separation model of coal and gangue

into airflow domain and ℎ119895represents the height of airflow

domain 119878119894represents the displacement in 119909 direction of

gangue before entering airflow domain 11990511989512

represents themotion time of gangue in airflow domain and 119878

119895represents

the motion displacement in 119909 direction of gangue in airflowdomain V

11989112represents the motion velocity in 119910 direction

of gangue in airflow domain (11990511989512) represents the motion

velocity in 119909 direction of gangue in airflow domain 11990511989112

represents the motion time of gangue after leaving airflowdomain and 119878

11989112represents the motion displacement in

119909 direction of gangue after leaving airflow domain In thefollowing equations subscript 1 is used to represent whenairflow velocity and conveyor belt velocity are in the samedirection and subscript 2 is used to represent when airflowvelocity and conveyor belt velocity are in the oppositedirection

Coal and ganguematerials will be recognized bymachinevision system and the recognized image information willbe sent to pneumatic separation system through imagesensor Coal particles will fall from the conveyor belt freelywithout the effect of high-pressure airflow When the motiondirection of mineral materials is the same as that of high-pressure airflow gangue material will change its trajectoryand will be thrown significantly farther than that of coalparticle When the motion direction of mineral materials isopposite to that of high-pressure airflow ganguematerial willchange its trajectory to an opposite direction compared withthat of coal particleThemotions of coal and gangue particlesare analyzed respectively in the following three sections

When coal materials are recognized by machine visionsystem coal particles will fall from the conveyor belt freelyunaffected by high-pressure airflow 119905

0and 1198780can be obtained

through (1) and (2) Gangue motion law at this stage is thesame as coal materials 119878

119894and V119901can be also shown by (3) and

(4) where 1199050represents the motion time of coal particle and

1198780represents the displacement of coal particle in 119909 direction

1199050= radic

2 (ℎ119901+ ℎ119895+ ℎ119891)

119892 (1)

1198780= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892 (2)

119878119894= V0radic2ℎ119901

119892 (3)

V119901= 119892radic

2ℎ119901

119892 (4)

In order to calculate the acting force of high pressureon gangue particles multiple rectangular polyline is used toapproximate gangue particlersquos physical shape if the dynamicpressure head of the airflow domain is known in speciallocationThe acting force of high-pressure airflow on ganguecan be presented by

119865 = ∬119860

119875119889119889119860 =

119898

sum

119894=1

119875119889119894119860119894 (5)

where 119865 represents the acting force of the high airflowon gangue 119875

119889119894represents the high pressure in ring 119894 119860

119894

represents the area of ring 119894 and 119898 represents the totalnumber of rings

As shown in Figure 3 gangue is affected by airflowwith different directions In order to analyze the movementof gangue particle in high-pressure airflow domain airresistance and horizontal momentumrsquos increment of airflowdomain are ignored The theoretical formula of ganguedisplacement in direction 119910 can be expressed as 119904 = int11990511989512

0(V119901+

11989211990511989512)119889119905 = ℎ minus 119889 then 119905

11989512can be obtained as shown in

11990511989512

=

radicV2119901+ 2119892 (ℎ minus 119889) minus V

119901

119892 (6)

As shown in (6) gangue affected by different airflowdirection has the same motion time in airflow domainGangue will be affected by two forces after falling into high-pressure airflow domain The two forces applied on gangue

4 Advances in Materials Science and Engineering

y

xo

d2

Airflow direction

u

hj tj1

(a) and 119906 are in the same direction

y

xo

d2

Airflow direction

u

hj tj2

(b) and 119906 are in the opposite direction

Figure 3 Schematic diagram of ganguersquos motion in airflow domain

particle are gravity and high dynamic pressure Airflowdynamic pressure will convert to static pressure on conditionthat the velocity of airflow 119906 is larger than ganguersquos horizontalvelocity V

0and airflow must keep dynamic pressure 119901

119889=

12058822 The pressure difference can be expressed as Δ119901

119889=

120588(1199062minus 2)2 According to (5) the formula of airflow force

can be expressed by

119865 = ∬119901119889119860 asymp sum

119894

119901119894119860119894=1

2120588sum

119894

(1199062

119894minus 2)119860119894 (7)

When the curvature of gangue surface is not too large itcan be approximately seen as a plane so forall119906

119894= 119906 (7) can be

changed to

119865 =1

2120588 (1199062minus 2)119860 (8)

According to Newtonrsquos second law (8) can be changed to

119898 =1

2120588119860 (119906

2minus 2)

and 119906 are in same direction

119898 = minus1

2120588119860 (119906

2+ 2)

and 119906 are in opposite direction

(9)

As shown in (9) the expressions are corresponding todifferent airflow directions The general solution of displace-ment in 119909 direction of gangue moving in the airflow can beobtained by taking Laplace transform to (9) and the result isshown in

119909 (1199051198951)

=1198601205881198622+ 119898 ln (119890(21198621119906minus1198601199051198951120588119906)119898 + 1) minus 119898119862

1+ 1198601199051198951120588119906

119860120588

and 119906 are in same direction

119909 (1199051198952)

=1198601205881198624+ 2119898 ln (cos ((119860119905

1198952119906120588 minus 2119898119906119862

3) 2119898))

119860120588

and 119906 are in opposite direction

(10)

The velocity of gangue in 119909 direction in airflow domaincan be obtained by taking derivation of (10) The generalsolution is shown in

1199091015840(1199051198951) =

1199061198901198601199051198951119906120588119898

11989021198621119906119898 + 1198901198601199051198951119906120588119898

and 119906 are in same direction

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898119906119862

3

2119898)

and 119906 are in opposite direction

(11)

In order to calculate the constant coefficients 1198621 1198622 1198623

and 1198624in (10) the initial constraint conditions are as follows

the initial position of the gangue is 119909(0) = 0 and the initialvelocity of gangue is 1199091015840(0) = V

0

When and 119906 are in the same direction the generalsolution of 119909(119905

119895) and 1199091015840(119905

119895) can be obtained by (10) and (11)

1198621and 119862

2are shown in

1198621=119898 ln (119906V

0minus 1)

2119906

1198622=1198982 ln (119906V

0minus 1) minus 2119898119906 ln (119906V

0)

2119860119906120588

(12)

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown in

119909 (1199051198951)

=119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

1199091015840(1199051198951) =

1199061198901198601199051198951119906120588119898

119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898

(13)

When and 119906 are in the opposite direction the generalsolution of 119909(119905

119895) and 1199091015840(119905

119895) can be obtained by (10) and (11)

1198623and 119862

4are shown by

1198623=

arccos(119906radicV20+ 1199062)

119906

1198624=

minus2119898 ln(119906radicV20+ 1199062)

119860120588

(14)

Advances in Materials Science and Engineering 5

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown by

(15) and (16) respectively

119909 (1199051198952) =

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

(15)

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898) (16)

When gangue particles left the boundary of airflowdomain gangue particles would do flat parabolicmotionTheformula of velocity V

119891in direction 119910 is shown in

V11989112

= int

11990511989512

0

(V119901+ 119892119905) 119889119905 (17)

11990511989112

can be solved through (18) and the result is shown by(19)

ℎ119891= int

11990511989112

0

(V119891+ 119892119905) 119889119905

= int

11990511989112

0

119892119905 119889119905 + int

11990511989112

0

int

119905119895

0

(V119901+ 119892119905) 119889

2119905

(18)

11990511989112

=

minus1198921199052

11989512minus 211990511989512

V119901+ radic8119892ℎ

119891+ (119892119905211989512

+ 211990511989512

V119901)2

2119892

(19)

The displacement 11987811989112

of gangue in direction 119909 afterleaving the airflow domain can be expressed by

11987811989112

= 1199091015840(11990511989512) 11990511989112

(20)

To sum up when and 119906 are in the same direction theseparation distance Δ119878 can be calculated by

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780= V0radic2ℎ119901

119892

+119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

(21)

When and119906 are in the opposite direction the separationdistance Δ119878 can be obtained by

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892

(22)

The separation distance Δ119878 is shown by (21) and (22)which reflects the basic motion law of gangue when influ-enced by airflow field under different airflow directionsFor the limitation of assumption there exist big differ-ences between theoretical model and practical model Inorder to simplify the calculation and correct the differencebetween theoretical value and practical value parameters 119896

119899

(nonlinear correction factor) and 119896119903(linear correction factor)

are introduced 119896119899reflects the convergence rate of the fitting

function and 119896119903reflects the convergent gain and is used to

adjust the fitting effect of formula based on experimentalvalue Theoretical value will get close to experiment value byadjusting the value of 119896

119899and 119896119903 The modified formulas with

the correction factors are shown in

6 Advances in Materials Science and Engineering

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus1198601198961198991199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601198961198991199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601198961198991199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 119896119903

(23)

Δ119878 = 1198780+ 119878119895+ 119878119891minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((119860119896119899119905119895119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(119860119905119895120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

119905119895(119892119905119895+ 2V119901)minus V0radic2ℎ119901

119892minus 119896119903

(24)

32 Air-Solid Multiphase Pneumatic Separation SimulationIn order to evaluate the pneumatic separation effect undertwo kinds of arrangement schemes for high-pressure valuea ldquofixed coarse-gridrdquo fluid scheme is applied in PFC3Dfor pneumatic separation simulations In the fluid scheme550 (22 times 5 times 5 119909 119910 and 119911 directions) fluid cells arecreated in a rectangular space (119909 = [0 7000mm] 119910 =[minus400mm 400mm] and 119911 = [300mm 400mm]) whichcovers the rectangular space A pneumatic boundary shouldbe set for the fluid grid Driving forces from the fluid flow areapplied to the particles as body forces These forces are alsoadded to the fluid equations and cause change inmomentumas reflected by the change in the pressure gradient in the flowdirection

As shown in Figures 4(a) and 4(b) two available modelsare established to reduce the computation time without lossin accuracy The gangue hopper containing 300 balls is builtat the top right of conveying belt 1 it is aimed at reducing thecomputation timeThe front view of the simulationmodels atan initial stage under different airflow directions is shown inFigures 4(a) and 4(b) respectively

In Figure 4(a) the velocity V0of conveyor belt 1 and

airflow velocity 119906 are in opposite directions When ganguefalls into the airflow domain gangue particle will change itstrajectory to an opposite direction and fall onto conveyor belt2 The conveyor belt 2 has the same movement direction asairflow velocity finally gangue particles will be transportedto collecting box by conveyor belt 2 In Figure 4(b) conveyorbelt V0and airflow velocity 119906 are in the same directionWhen

gangue falls into the airflow domain gangue particle willthrow much farther and fall on conveyor belt 2 The motionof conveyor belt 2 has the same motion direction as conveyorbelt 1 finally the gangue particles are transported to collectingbox by conveyor belt 2

In the fluid scheme a pneumatic boundary is set for twoavailable models During the simulations an approximation

is made by specifying the velocity boundary at the rightend of the model and a pressure boundary as 00 Pa at leftend with 119909 = 0mm The slip boundary in which thefluid velocity parallel to the wall surface is nonzero at thewall surface is specified at the surrounding four walls Insimulation as shown by Figure 4(b) an approximation ismade by specifying the velocity boundary at the left end ofmodel and a pressure boundary as 00 Pa at the right endwith 119909 = 7000mm The initial setup of the slip boundaryin Figure 4(b) is the same as that shown in Figure 4(a) andmaterial properties are shown in Table 1

As shown in Figure 4(a) airflow is injectedwith a velocityof 300ms from the negative direction of 119909 at the startingpoint while in Figure 4(b) air is injected from the positivedirection of 119909 When the pneumatic boundaries are appliedat the initial stage the two side walls applied to confine theassembly are removed simultaneously Figure 5 shows thefront view of simulation result from the initial stage to 1 sec

It can be obtained from Figure 5 that gangue particleshave different motion trajectories under different airflowvelocity As shown in Figures 5(a) and 5(b) gangue particleswill do horizontal projectile motion before entering airflowdomain as can be seen at 119905 = 02 s and 119905 = 04 s As shown inFigure 5(a) velocities of conveyor belt 1 V

0and airflow 119906 are in

the opposite direction and gangue particles will change theirmotion trajectories to an opposite direction compared withthat of coal without the effect of airflow which can be seen at119905 = 06 s 119905 = 08 s and 119905 = 10 s As shown in Figure 5(b)velocities of the conveyor belt 1 V

0and airflow 119906 are in the

same direction and gangue particles will be blown muchfarther than that of coal without being affected by airflow ascan be seen at 119905 = 06 s 119905 = 08 s and 119905 = 10 s

As can be seen in Figures 5(a) and 5(b) gangue particleswill change their trajectories in areaA and areaCAreas B andD as shown in Figures 5(a) and 5(b) illustrated that gangueparticles with smaller diameters can be blown much farther

Advances in Materials Science and Engineering 7

Gangue hopperConveying belt 1

Conveying belt 2Collecting box

Fluid boundary

hp

0

u

(a) V0 and 119906 in the opposite direction

Gangue hopper

Conveying belt 1

Collecting box

Fluid boundary

Conveying belt 2

hp

0

u

(b) V0 and 119906 in the same direction

Figure 4 Air-solid multiphase pneumatic separation

BA

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(a) V0 and 119906 are in the opposite direction

DC

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(b) V0 and 119906 are in the same direction

Figure 5 Front view (initial stage to 1 sec apparent velocity plusmn260ms)

Table 1 Materials properties

Parameter Value UnitsBall

Diameter 50sim100 mmNumber 300 mdashDensity 2700 kgm3

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 07 mdash

AirDensity 1205 kgm3

Viscosity 18 times 10minus6 PasdotsWall

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 03 mdash

In order to study the pneumatic separation distanceinfluenced by different airflow directions airflow velocitiesplusmn300ms are chosen for the research of separation effectConveyor velocity V

0and height difference ℎ

119901between con-

veyor belt and air nozzle are kept constant The relationshipsbetween separation distances and particle diameters underdifferent airflow velocity directions are shown in Figure 6

45 50 55 60 65 70 75 80 85 90 95 100 10514

16

18

20

22

24

26

28

30

32

ΔS

(m)

d (mm)u = minus300ms and 0 = 1msu = 300ms and 0 = 1ms

Figure 6The relationship between particle diameter and separationdistance under different direction airflow velocity (119906 = plusmn300msV0= +1ms)

As can be seen in Figure 6 separation distance decreaseswith the increase of particle diameter It also can be con-cluded that the separation effect under airflow velocityminus300ms is better than that under airflow velocity 300ms

8 Advances in Materials Science and Engineering

Table 2 Levels of factors

Level Factor119860 gangue diameter 119889 (mm) 119861 height difference ℎ

119901(m) 119862 conveyor belt velocity V

0(ms)

1 50 040 0502 80 080 1003 100 110 200

Table 3 Experimental results of separation effect

Experiment number 119860 119861 119862Separation distance Δ119878 (m)

119906 and V0are in the opposite direction 119906 and V

0are in the same direction

1 50 04 05 305 2212 50 08 10 22 1143 50 11 20 205 0784 80 04 10 215 1085 80 08 20 148 0656 80 11 05 125 0427 100 04 20 185 0988 100 08 05 155 0379 100 11 10 140 028

The separation effect also can be analyzed from (9) dynamicpressure differenceΔ119875

119889can be expressed as 120588(1199062minus2)2when

119906 and V0are in the same direction while when 119906 and V

0are in

the opposite direction the dynamic pressure difference Δ119875119889

can be expressed as 120588(1199062 + 2)2 Thus it can be obtainedfrom the above analysis that separation effect under airflowvelocity minus300ms is significantly better than that of underairflow velocity 300ms

33 Orthogonal Experiment of Pneumatic Separation Digitalimage processing technology has been used to identify thetarget of various patterns of coal and gangue in undergroundpneumatic separation system Before mineral materials aresent to machine vision system coal and gangue have beencrushed to 100mm by impact crusher The size of materialsis ranging from 50mm to 100mm These materials are sentby the conveyor belt for coal and gangue digital imageinformation identification (as shown in Figure 1)

In this study separation distance Δ119878 (m) is selectedas the primary index to evaluate the pneumatic separationeffect Pneumatic separation influenced by different airflowdirection can be achieved by changing the arrangement ofhigh-pressure value The pneumatic separation test systemmainly consists of conveyor belt queuing system machinevision system control system and high-pressure air injectionsystem When coal and gangue materials fell down from theconveyor belt gangue will be identified by machine visionsystem and the information is transported to control systemthrough image sensor to drive the electromagnetic valueopenThus coal gangue pneumatic separation is realizedTheair compressor used in pneumatic separation testing systemis LG-6510 its working pressure is 10Mpa and certifiedcapacity is 65m3min

From the analysis shown in Section 3 the conveyorvelocity V

0 height difference ℎ

119901 and gangue diameter 119889

are selected as the three factors Factors and levels arelisted in Table 2 According to the identified level of factorsorthogonal table L

9(34) is applied in the test Orthogonal

experiment arrangement and results are shown in Table 3

4 Result and Discussion

41 Variance Analysis and Range Analysis Based on Orthogo-nal Test As can be seen from Table 2 each factor at differentlevels is approximate linearity so the method of regressionanalysis is to obtain the linear function relationship betweenthem appropriately [25] Thus the regression equationsof coal gangue separation distance under different airflowdirection are derived Equation (25) represents the regressionequationwhen 119906 and V

0are in the opposite direction and (26)

represents the regression equation when 119906 and V0are in the

same direction where 119910 is the separation distance of coal andgangue 119909

1represents the diameter of gangue 119909

2is the height

difference and 1199093is the velocity of conveyor belt

119910 = 467 minus 00191199091minus 1773119909

2+ 0092119909

3 (25)

119910 = 360 minus 0021199091minus 1599119909

2minus 014119909

3 (26)

Variance analysis is carried out on the regression equationto make significance test and the results are shown inTables 4 and 5 In order to determine the optimal pneumaticseparation solutions comparative analyses of the rangesbetween various levels of each factor are carried out Therange analysis is shown in Table 6

In Table 6 subscripts 1 and 2 used in influence factors(119860 119861 119862) represent the fact that 119906 and V

0are in the opposite

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

4 Advances in Materials Science and Engineering

y

xo

d2

Airflow direction

u

hj tj1

(a) and 119906 are in the same direction

y

xo

d2

Airflow direction

u

hj tj2

(b) and 119906 are in the opposite direction

Figure 3 Schematic diagram of ganguersquos motion in airflow domain

particle are gravity and high dynamic pressure Airflowdynamic pressure will convert to static pressure on conditionthat the velocity of airflow 119906 is larger than ganguersquos horizontalvelocity V

0and airflow must keep dynamic pressure 119901

119889=

12058822 The pressure difference can be expressed as Δ119901

119889=

120588(1199062minus 2)2 According to (5) the formula of airflow force

can be expressed by

119865 = ∬119901119889119860 asymp sum

119894

119901119894119860119894=1

2120588sum

119894

(1199062

119894minus 2)119860119894 (7)

When the curvature of gangue surface is not too large itcan be approximately seen as a plane so forall119906

119894= 119906 (7) can be

changed to

119865 =1

2120588 (1199062minus 2)119860 (8)

According to Newtonrsquos second law (8) can be changed to

119898 =1

2120588119860 (119906

2minus 2)

and 119906 are in same direction

119898 = minus1

2120588119860 (119906

2+ 2)

and 119906 are in opposite direction

(9)

As shown in (9) the expressions are corresponding todifferent airflow directions The general solution of displace-ment in 119909 direction of gangue moving in the airflow can beobtained by taking Laplace transform to (9) and the result isshown in

119909 (1199051198951)

=1198601205881198622+ 119898 ln (119890(21198621119906minus1198601199051198951120588119906)119898 + 1) minus 119898119862

1+ 1198601199051198951120588119906

119860120588

and 119906 are in same direction

119909 (1199051198952)

=1198601205881198624+ 2119898 ln (cos ((119860119905

1198952119906120588 minus 2119898119906119862

3) 2119898))

119860120588

and 119906 are in opposite direction

(10)

The velocity of gangue in 119909 direction in airflow domaincan be obtained by taking derivation of (10) The generalsolution is shown in

1199091015840(1199051198951) =

1199061198901198601199051198951119906120588119898

11989021198621119906119898 + 1198901198601199051198951119906120588119898

and 119906 are in same direction

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898119906119862

3

2119898)

and 119906 are in opposite direction

(11)

In order to calculate the constant coefficients 1198621 1198622 1198623

and 1198624in (10) the initial constraint conditions are as follows

the initial position of the gangue is 119909(0) = 0 and the initialvelocity of gangue is 1199091015840(0) = V

0

When and 119906 are in the same direction the generalsolution of 119909(119905

119895) and 1199091015840(119905

119895) can be obtained by (10) and (11)

1198621and 119862

2are shown in

1198621=119898 ln (119906V

0minus 1)

2119906

1198622=1198982 ln (119906V

0minus 1) minus 2119898119906 ln (119906V

0)

2119860119906120588

(12)

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown in

119909 (1199051198951)

=119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

1199091015840(1199051198951) =

1199061198901198601199051198951119906120588119898

119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898

(13)

When and 119906 are in the opposite direction the generalsolution of 119909(119905

119895) and 1199091015840(119905

119895) can be obtained by (10) and (11)

1198623and 119862

4are shown by

1198623=

arccos(119906radicV20+ 1199062)

119906

1198624=

minus2119898 ln(119906radicV20+ 1199062)

119860120588

(14)

Advances in Materials Science and Engineering 5

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown by

(15) and (16) respectively

119909 (1199051198952) =

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

(15)

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898) (16)

When gangue particles left the boundary of airflowdomain gangue particles would do flat parabolicmotionTheformula of velocity V

119891in direction 119910 is shown in

V11989112

= int

11990511989512

0

(V119901+ 119892119905) 119889119905 (17)

11990511989112

can be solved through (18) and the result is shown by(19)

ℎ119891= int

11990511989112

0

(V119891+ 119892119905) 119889119905

= int

11990511989112

0

119892119905 119889119905 + int

11990511989112

0

int

119905119895

0

(V119901+ 119892119905) 119889

2119905

(18)

11990511989112

=

minus1198921199052

11989512minus 211990511989512

V119901+ radic8119892ℎ

119891+ (119892119905211989512

+ 211990511989512

V119901)2

2119892

(19)

The displacement 11987811989112

of gangue in direction 119909 afterleaving the airflow domain can be expressed by

11987811989112

= 1199091015840(11990511989512) 11990511989112

(20)

To sum up when and 119906 are in the same direction theseparation distance Δ119878 can be calculated by

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780= V0radic2ℎ119901

119892

+119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

(21)

When and119906 are in the opposite direction the separationdistance Δ119878 can be obtained by

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892

(22)

The separation distance Δ119878 is shown by (21) and (22)which reflects the basic motion law of gangue when influ-enced by airflow field under different airflow directionsFor the limitation of assumption there exist big differ-ences between theoretical model and practical model Inorder to simplify the calculation and correct the differencebetween theoretical value and practical value parameters 119896

119899

(nonlinear correction factor) and 119896119903(linear correction factor)

are introduced 119896119899reflects the convergence rate of the fitting

function and 119896119903reflects the convergent gain and is used to

adjust the fitting effect of formula based on experimentalvalue Theoretical value will get close to experiment value byadjusting the value of 119896

119899and 119896119903 The modified formulas with

the correction factors are shown in

6 Advances in Materials Science and Engineering

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus1198601198961198991199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601198961198991199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601198961198991199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 119896119903

(23)

Δ119878 = 1198780+ 119878119895+ 119878119891minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((119860119896119899119905119895119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(119860119905119895120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

119905119895(119892119905119895+ 2V119901)minus V0radic2ℎ119901

119892minus 119896119903

(24)

32 Air-Solid Multiphase Pneumatic Separation SimulationIn order to evaluate the pneumatic separation effect undertwo kinds of arrangement schemes for high-pressure valuea ldquofixed coarse-gridrdquo fluid scheme is applied in PFC3Dfor pneumatic separation simulations In the fluid scheme550 (22 times 5 times 5 119909 119910 and 119911 directions) fluid cells arecreated in a rectangular space (119909 = [0 7000mm] 119910 =[minus400mm 400mm] and 119911 = [300mm 400mm]) whichcovers the rectangular space A pneumatic boundary shouldbe set for the fluid grid Driving forces from the fluid flow areapplied to the particles as body forces These forces are alsoadded to the fluid equations and cause change inmomentumas reflected by the change in the pressure gradient in the flowdirection

As shown in Figures 4(a) and 4(b) two available modelsare established to reduce the computation time without lossin accuracy The gangue hopper containing 300 balls is builtat the top right of conveying belt 1 it is aimed at reducing thecomputation timeThe front view of the simulationmodels atan initial stage under different airflow directions is shown inFigures 4(a) and 4(b) respectively

In Figure 4(a) the velocity V0of conveyor belt 1 and

airflow velocity 119906 are in opposite directions When ganguefalls into the airflow domain gangue particle will change itstrajectory to an opposite direction and fall onto conveyor belt2 The conveyor belt 2 has the same movement direction asairflow velocity finally gangue particles will be transportedto collecting box by conveyor belt 2 In Figure 4(b) conveyorbelt V0and airflow velocity 119906 are in the same directionWhen

gangue falls into the airflow domain gangue particle willthrow much farther and fall on conveyor belt 2 The motionof conveyor belt 2 has the same motion direction as conveyorbelt 1 finally the gangue particles are transported to collectingbox by conveyor belt 2

In the fluid scheme a pneumatic boundary is set for twoavailable models During the simulations an approximation

is made by specifying the velocity boundary at the rightend of the model and a pressure boundary as 00 Pa at leftend with 119909 = 0mm The slip boundary in which thefluid velocity parallel to the wall surface is nonzero at thewall surface is specified at the surrounding four walls Insimulation as shown by Figure 4(b) an approximation ismade by specifying the velocity boundary at the left end ofmodel and a pressure boundary as 00 Pa at the right endwith 119909 = 7000mm The initial setup of the slip boundaryin Figure 4(b) is the same as that shown in Figure 4(a) andmaterial properties are shown in Table 1

As shown in Figure 4(a) airflow is injectedwith a velocityof 300ms from the negative direction of 119909 at the startingpoint while in Figure 4(b) air is injected from the positivedirection of 119909 When the pneumatic boundaries are appliedat the initial stage the two side walls applied to confine theassembly are removed simultaneously Figure 5 shows thefront view of simulation result from the initial stage to 1 sec

It can be obtained from Figure 5 that gangue particleshave different motion trajectories under different airflowvelocity As shown in Figures 5(a) and 5(b) gangue particleswill do horizontal projectile motion before entering airflowdomain as can be seen at 119905 = 02 s and 119905 = 04 s As shown inFigure 5(a) velocities of conveyor belt 1 V

0and airflow 119906 are in

the opposite direction and gangue particles will change theirmotion trajectories to an opposite direction compared withthat of coal without the effect of airflow which can be seen at119905 = 06 s 119905 = 08 s and 119905 = 10 s As shown in Figure 5(b)velocities of the conveyor belt 1 V

0and airflow 119906 are in the

same direction and gangue particles will be blown muchfarther than that of coal without being affected by airflow ascan be seen at 119905 = 06 s 119905 = 08 s and 119905 = 10 s

As can be seen in Figures 5(a) and 5(b) gangue particleswill change their trajectories in areaA and areaCAreas B andD as shown in Figures 5(a) and 5(b) illustrated that gangueparticles with smaller diameters can be blown much farther

Advances in Materials Science and Engineering 7

Gangue hopperConveying belt 1

Conveying belt 2Collecting box

Fluid boundary

hp

0

u

(a) V0 and 119906 in the opposite direction

Gangue hopper

Conveying belt 1

Collecting box

Fluid boundary

Conveying belt 2

hp

0

u

(b) V0 and 119906 in the same direction

Figure 4 Air-solid multiphase pneumatic separation

BA

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(a) V0 and 119906 are in the opposite direction

DC

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(b) V0 and 119906 are in the same direction

Figure 5 Front view (initial stage to 1 sec apparent velocity plusmn260ms)

Table 1 Materials properties

Parameter Value UnitsBall

Diameter 50sim100 mmNumber 300 mdashDensity 2700 kgm3

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 07 mdash

AirDensity 1205 kgm3

Viscosity 18 times 10minus6 PasdotsWall

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 03 mdash

In order to study the pneumatic separation distanceinfluenced by different airflow directions airflow velocitiesplusmn300ms are chosen for the research of separation effectConveyor velocity V

0and height difference ℎ

119901between con-

veyor belt and air nozzle are kept constant The relationshipsbetween separation distances and particle diameters underdifferent airflow velocity directions are shown in Figure 6

45 50 55 60 65 70 75 80 85 90 95 100 10514

16

18

20

22

24

26

28

30

32

ΔS

(m)

d (mm)u = minus300ms and 0 = 1msu = 300ms and 0 = 1ms

Figure 6The relationship between particle diameter and separationdistance under different direction airflow velocity (119906 = plusmn300msV0= +1ms)

As can be seen in Figure 6 separation distance decreaseswith the increase of particle diameter It also can be con-cluded that the separation effect under airflow velocityminus300ms is better than that under airflow velocity 300ms

8 Advances in Materials Science and Engineering

Table 2 Levels of factors

Level Factor119860 gangue diameter 119889 (mm) 119861 height difference ℎ

119901(m) 119862 conveyor belt velocity V

0(ms)

1 50 040 0502 80 080 1003 100 110 200

Table 3 Experimental results of separation effect

Experiment number 119860 119861 119862Separation distance Δ119878 (m)

119906 and V0are in the opposite direction 119906 and V

0are in the same direction

1 50 04 05 305 2212 50 08 10 22 1143 50 11 20 205 0784 80 04 10 215 1085 80 08 20 148 0656 80 11 05 125 0427 100 04 20 185 0988 100 08 05 155 0379 100 11 10 140 028

The separation effect also can be analyzed from (9) dynamicpressure differenceΔ119875

119889can be expressed as 120588(1199062minus2)2when

119906 and V0are in the same direction while when 119906 and V

0are in

the opposite direction the dynamic pressure difference Δ119875119889

can be expressed as 120588(1199062 + 2)2 Thus it can be obtainedfrom the above analysis that separation effect under airflowvelocity minus300ms is significantly better than that of underairflow velocity 300ms

33 Orthogonal Experiment of Pneumatic Separation Digitalimage processing technology has been used to identify thetarget of various patterns of coal and gangue in undergroundpneumatic separation system Before mineral materials aresent to machine vision system coal and gangue have beencrushed to 100mm by impact crusher The size of materialsis ranging from 50mm to 100mm These materials are sentby the conveyor belt for coal and gangue digital imageinformation identification (as shown in Figure 1)

In this study separation distance Δ119878 (m) is selectedas the primary index to evaluate the pneumatic separationeffect Pneumatic separation influenced by different airflowdirection can be achieved by changing the arrangement ofhigh-pressure value The pneumatic separation test systemmainly consists of conveyor belt queuing system machinevision system control system and high-pressure air injectionsystem When coal and gangue materials fell down from theconveyor belt gangue will be identified by machine visionsystem and the information is transported to control systemthrough image sensor to drive the electromagnetic valueopenThus coal gangue pneumatic separation is realizedTheair compressor used in pneumatic separation testing systemis LG-6510 its working pressure is 10Mpa and certifiedcapacity is 65m3min

From the analysis shown in Section 3 the conveyorvelocity V

0 height difference ℎ

119901 and gangue diameter 119889

are selected as the three factors Factors and levels arelisted in Table 2 According to the identified level of factorsorthogonal table L

9(34) is applied in the test Orthogonal

experiment arrangement and results are shown in Table 3

4 Result and Discussion

41 Variance Analysis and Range Analysis Based on Orthogo-nal Test As can be seen from Table 2 each factor at differentlevels is approximate linearity so the method of regressionanalysis is to obtain the linear function relationship betweenthem appropriately [25] Thus the regression equationsof coal gangue separation distance under different airflowdirection are derived Equation (25) represents the regressionequationwhen 119906 and V

0are in the opposite direction and (26)

represents the regression equation when 119906 and V0are in the

same direction where 119910 is the separation distance of coal andgangue 119909

1represents the diameter of gangue 119909

2is the height

difference and 1199093is the velocity of conveyor belt

119910 = 467 minus 00191199091minus 1773119909

2+ 0092119909

3 (25)

119910 = 360 minus 0021199091minus 1599119909

2minus 014119909

3 (26)

Variance analysis is carried out on the regression equationto make significance test and the results are shown inTables 4 and 5 In order to determine the optimal pneumaticseparation solutions comparative analyses of the rangesbetween various levels of each factor are carried out Therange analysis is shown in Table 6

In Table 6 subscripts 1 and 2 used in influence factors(119860 119861 119862) represent the fact that 119906 and V

0are in the opposite

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

Advances in Materials Science and Engineering 5

The general solutions of 119909(119905119895) and 1199091015840(119905

119895) are shown by

(15) and (16) respectively

119909 (1199051198952) =

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

(15)

1199091015840(1199051198952) = minus119906 tan(

1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898) (16)

When gangue particles left the boundary of airflowdomain gangue particles would do flat parabolicmotionTheformula of velocity V

119891in direction 119910 is shown in

V11989112

= int

11990511989512

0

(V119901+ 119892119905) 119889119905 (17)

11990511989112

can be solved through (18) and the result is shown by(19)

ℎ119891= int

11990511989112

0

(V119891+ 119892119905) 119889119905

= int

11990511989112

0

119892119905 119889119905 + int

11990511989112

0

int

119905119895

0

(V119901+ 119892119905) 119889

2119905

(18)

11990511989112

=

minus1198921199052

11989512minus 211990511989512

V119901+ radic8119892ℎ

119891+ (119892119905211989512

+ 211990511989512

V119901)2

2119892

(19)

The displacement 11987811989112

of gangue in direction 119909 afterleaving the airflow domain can be expressed by

11987811989112

= 1199091015840(11990511989512) 11990511989112

(20)

To sum up when and 119906 are in the same direction theseparation distance Δ119878 can be calculated by

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780= V0radic2ℎ119901

119892

+119898 ln (119890(119898 ln(119906V0minus1)minus1198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

(21)

When and119906 are in the opposite direction the separationdistance Δ119878 can be obtained by

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((1198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892

(22)

The separation distance Δ119878 is shown by (21) and (22)which reflects the basic motion law of gangue when influ-enced by airflow field under different airflow directionsFor the limitation of assumption there exist big differ-ences between theoretical model and practical model Inorder to simplify the calculation and correct the differencebetween theoretical value and practical value parameters 119896

119899

(nonlinear correction factor) and 119896119903(linear correction factor)

are introduced 119896119899reflects the convergence rate of the fitting

function and 119896119903reflects the convergent gain and is used to

adjust the fitting effect of formula based on experimentalvalue Theoretical value will get close to experiment value byadjusting the value of 119896

119899and 119896119903 The modified formulas with

the correction factors are shown in

6 Advances in Materials Science and Engineering

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus1198601198961198991199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601198961198991199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601198961198991199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 119896119903

(23)

Δ119878 = 1198780+ 119878119895+ 119878119891minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((119860119896119899119905119895119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(119860119905119895120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

119905119895(119892119905119895+ 2V119901)minus V0radic2ℎ119901

119892minus 119896119903

(24)

32 Air-Solid Multiphase Pneumatic Separation SimulationIn order to evaluate the pneumatic separation effect undertwo kinds of arrangement schemes for high-pressure valuea ldquofixed coarse-gridrdquo fluid scheme is applied in PFC3Dfor pneumatic separation simulations In the fluid scheme550 (22 times 5 times 5 119909 119910 and 119911 directions) fluid cells arecreated in a rectangular space (119909 = [0 7000mm] 119910 =[minus400mm 400mm] and 119911 = [300mm 400mm]) whichcovers the rectangular space A pneumatic boundary shouldbe set for the fluid grid Driving forces from the fluid flow areapplied to the particles as body forces These forces are alsoadded to the fluid equations and cause change inmomentumas reflected by the change in the pressure gradient in the flowdirection

As shown in Figures 4(a) and 4(b) two available modelsare established to reduce the computation time without lossin accuracy The gangue hopper containing 300 balls is builtat the top right of conveying belt 1 it is aimed at reducing thecomputation timeThe front view of the simulationmodels atan initial stage under different airflow directions is shown inFigures 4(a) and 4(b) respectively

In Figure 4(a) the velocity V0of conveyor belt 1 and

airflow velocity 119906 are in opposite directions When ganguefalls into the airflow domain gangue particle will change itstrajectory to an opposite direction and fall onto conveyor belt2 The conveyor belt 2 has the same movement direction asairflow velocity finally gangue particles will be transportedto collecting box by conveyor belt 2 In Figure 4(b) conveyorbelt V0and airflow velocity 119906 are in the same directionWhen

gangue falls into the airflow domain gangue particle willthrow much farther and fall on conveyor belt 2 The motionof conveyor belt 2 has the same motion direction as conveyorbelt 1 finally the gangue particles are transported to collectingbox by conveyor belt 2

In the fluid scheme a pneumatic boundary is set for twoavailable models During the simulations an approximation

is made by specifying the velocity boundary at the rightend of the model and a pressure boundary as 00 Pa at leftend with 119909 = 0mm The slip boundary in which thefluid velocity parallel to the wall surface is nonzero at thewall surface is specified at the surrounding four walls Insimulation as shown by Figure 4(b) an approximation ismade by specifying the velocity boundary at the left end ofmodel and a pressure boundary as 00 Pa at the right endwith 119909 = 7000mm The initial setup of the slip boundaryin Figure 4(b) is the same as that shown in Figure 4(a) andmaterial properties are shown in Table 1

As shown in Figure 4(a) airflow is injectedwith a velocityof 300ms from the negative direction of 119909 at the startingpoint while in Figure 4(b) air is injected from the positivedirection of 119909 When the pneumatic boundaries are appliedat the initial stage the two side walls applied to confine theassembly are removed simultaneously Figure 5 shows thefront view of simulation result from the initial stage to 1 sec

It can be obtained from Figure 5 that gangue particleshave different motion trajectories under different airflowvelocity As shown in Figures 5(a) and 5(b) gangue particleswill do horizontal projectile motion before entering airflowdomain as can be seen at 119905 = 02 s and 119905 = 04 s As shown inFigure 5(a) velocities of conveyor belt 1 V

0and airflow 119906 are in

the opposite direction and gangue particles will change theirmotion trajectories to an opposite direction compared withthat of coal without the effect of airflow which can be seen at119905 = 06 s 119905 = 08 s and 119905 = 10 s As shown in Figure 5(b)velocities of the conveyor belt 1 V

0and airflow 119906 are in the

same direction and gangue particles will be blown muchfarther than that of coal without being affected by airflow ascan be seen at 119905 = 06 s 119905 = 08 s and 119905 = 10 s

As can be seen in Figures 5(a) and 5(b) gangue particleswill change their trajectories in areaA and areaCAreas B andD as shown in Figures 5(a) and 5(b) illustrated that gangueparticles with smaller diameters can be blown much farther

Advances in Materials Science and Engineering 7

Gangue hopperConveying belt 1

Conveying belt 2Collecting box

Fluid boundary

hp

0

u

(a) V0 and 119906 in the opposite direction

Gangue hopper

Conveying belt 1

Collecting box

Fluid boundary

Conveying belt 2

hp

0

u

(b) V0 and 119906 in the same direction

Figure 4 Air-solid multiphase pneumatic separation

BA

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(a) V0 and 119906 are in the opposite direction

DC

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(b) V0 and 119906 are in the same direction

Figure 5 Front view (initial stage to 1 sec apparent velocity plusmn260ms)

Table 1 Materials properties

Parameter Value UnitsBall

Diameter 50sim100 mmNumber 300 mdashDensity 2700 kgm3

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 07 mdash

AirDensity 1205 kgm3

Viscosity 18 times 10minus6 PasdotsWall

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 03 mdash

In order to study the pneumatic separation distanceinfluenced by different airflow directions airflow velocitiesplusmn300ms are chosen for the research of separation effectConveyor velocity V

0and height difference ℎ

119901between con-

veyor belt and air nozzle are kept constant The relationshipsbetween separation distances and particle diameters underdifferent airflow velocity directions are shown in Figure 6

45 50 55 60 65 70 75 80 85 90 95 100 10514

16

18

20

22

24

26

28

30

32

ΔS

(m)

d (mm)u = minus300ms and 0 = 1msu = 300ms and 0 = 1ms

Figure 6The relationship between particle diameter and separationdistance under different direction airflow velocity (119906 = plusmn300msV0= +1ms)

As can be seen in Figure 6 separation distance decreaseswith the increase of particle diameter It also can be con-cluded that the separation effect under airflow velocityminus300ms is better than that under airflow velocity 300ms

8 Advances in Materials Science and Engineering

Table 2 Levels of factors

Level Factor119860 gangue diameter 119889 (mm) 119861 height difference ℎ

119901(m) 119862 conveyor belt velocity V

0(ms)

1 50 040 0502 80 080 1003 100 110 200

Table 3 Experimental results of separation effect

Experiment number 119860 119861 119862Separation distance Δ119878 (m)

119906 and V0are in the opposite direction 119906 and V

0are in the same direction

1 50 04 05 305 2212 50 08 10 22 1143 50 11 20 205 0784 80 04 10 215 1085 80 08 20 148 0656 80 11 05 125 0427 100 04 20 185 0988 100 08 05 155 0379 100 11 10 140 028

The separation effect also can be analyzed from (9) dynamicpressure differenceΔ119875

119889can be expressed as 120588(1199062minus2)2when

119906 and V0are in the same direction while when 119906 and V

0are in

the opposite direction the dynamic pressure difference Δ119875119889

can be expressed as 120588(1199062 + 2)2 Thus it can be obtainedfrom the above analysis that separation effect under airflowvelocity minus300ms is significantly better than that of underairflow velocity 300ms

33 Orthogonal Experiment of Pneumatic Separation Digitalimage processing technology has been used to identify thetarget of various patterns of coal and gangue in undergroundpneumatic separation system Before mineral materials aresent to machine vision system coal and gangue have beencrushed to 100mm by impact crusher The size of materialsis ranging from 50mm to 100mm These materials are sentby the conveyor belt for coal and gangue digital imageinformation identification (as shown in Figure 1)

In this study separation distance Δ119878 (m) is selectedas the primary index to evaluate the pneumatic separationeffect Pneumatic separation influenced by different airflowdirection can be achieved by changing the arrangement ofhigh-pressure value The pneumatic separation test systemmainly consists of conveyor belt queuing system machinevision system control system and high-pressure air injectionsystem When coal and gangue materials fell down from theconveyor belt gangue will be identified by machine visionsystem and the information is transported to control systemthrough image sensor to drive the electromagnetic valueopenThus coal gangue pneumatic separation is realizedTheair compressor used in pneumatic separation testing systemis LG-6510 its working pressure is 10Mpa and certifiedcapacity is 65m3min

From the analysis shown in Section 3 the conveyorvelocity V

0 height difference ℎ

119901 and gangue diameter 119889

are selected as the three factors Factors and levels arelisted in Table 2 According to the identified level of factorsorthogonal table L

9(34) is applied in the test Orthogonal

experiment arrangement and results are shown in Table 3

4 Result and Discussion

41 Variance Analysis and Range Analysis Based on Orthogo-nal Test As can be seen from Table 2 each factor at differentlevels is approximate linearity so the method of regressionanalysis is to obtain the linear function relationship betweenthem appropriately [25] Thus the regression equationsof coal gangue separation distance under different airflowdirection are derived Equation (25) represents the regressionequationwhen 119906 and V

0are in the opposite direction and (26)

represents the regression equation when 119906 and V0are in the

same direction where 119910 is the separation distance of coal andgangue 119909

1represents the diameter of gangue 119909

2is the height

difference and 1199093is the velocity of conveyor belt

119910 = 467 minus 00191199091minus 1773119909

2+ 0092119909

3 (25)

119910 = 360 minus 0021199091minus 1599119909

2minus 014119909

3 (26)

Variance analysis is carried out on the regression equationto make significance test and the results are shown inTables 4 and 5 In order to determine the optimal pneumaticseparation solutions comparative analyses of the rangesbetween various levels of each factor are carried out Therange analysis is shown in Table 6

In Table 6 subscripts 1 and 2 used in influence factors(119860 119861 119862) represent the fact that 119906 and V

0are in the opposite

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

6 Advances in Materials Science and Engineering

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus1198601198961198991199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ1198911199061198901198601198961198991199051198951119906120588119898

(119890ln(119906V0minus1) + 1198901198601198961198991199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 119896119903

(23)

Δ119878 = 1198780+ 119878119895+ 119878119891minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892+

2119898 ln(cos((119860119896119899119905119895119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(119860119905119895120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

119905119895(119892119905119895+ 2V119901)minus V0radic2ℎ119901

119892minus 119896119903

(24)

32 Air-Solid Multiphase Pneumatic Separation SimulationIn order to evaluate the pneumatic separation effect undertwo kinds of arrangement schemes for high-pressure valuea ldquofixed coarse-gridrdquo fluid scheme is applied in PFC3Dfor pneumatic separation simulations In the fluid scheme550 (22 times 5 times 5 119909 119910 and 119911 directions) fluid cells arecreated in a rectangular space (119909 = [0 7000mm] 119910 =[minus400mm 400mm] and 119911 = [300mm 400mm]) whichcovers the rectangular space A pneumatic boundary shouldbe set for the fluid grid Driving forces from the fluid flow areapplied to the particles as body forces These forces are alsoadded to the fluid equations and cause change inmomentumas reflected by the change in the pressure gradient in the flowdirection

As shown in Figures 4(a) and 4(b) two available modelsare established to reduce the computation time without lossin accuracy The gangue hopper containing 300 balls is builtat the top right of conveying belt 1 it is aimed at reducing thecomputation timeThe front view of the simulationmodels atan initial stage under different airflow directions is shown inFigures 4(a) and 4(b) respectively

In Figure 4(a) the velocity V0of conveyor belt 1 and

airflow velocity 119906 are in opposite directions When ganguefalls into the airflow domain gangue particle will change itstrajectory to an opposite direction and fall onto conveyor belt2 The conveyor belt 2 has the same movement direction asairflow velocity finally gangue particles will be transportedto collecting box by conveyor belt 2 In Figure 4(b) conveyorbelt V0and airflow velocity 119906 are in the same directionWhen

gangue falls into the airflow domain gangue particle willthrow much farther and fall on conveyor belt 2 The motionof conveyor belt 2 has the same motion direction as conveyorbelt 1 finally the gangue particles are transported to collectingbox by conveyor belt 2

In the fluid scheme a pneumatic boundary is set for twoavailable models During the simulations an approximation

is made by specifying the velocity boundary at the rightend of the model and a pressure boundary as 00 Pa at leftend with 119909 = 0mm The slip boundary in which thefluid velocity parallel to the wall surface is nonzero at thewall surface is specified at the surrounding four walls Insimulation as shown by Figure 4(b) an approximation ismade by specifying the velocity boundary at the left end ofmodel and a pressure boundary as 00 Pa at the right endwith 119909 = 7000mm The initial setup of the slip boundaryin Figure 4(b) is the same as that shown in Figure 4(a) andmaterial properties are shown in Table 1

As shown in Figure 4(a) airflow is injectedwith a velocityof 300ms from the negative direction of 119909 at the startingpoint while in Figure 4(b) air is injected from the positivedirection of 119909 When the pneumatic boundaries are appliedat the initial stage the two side walls applied to confine theassembly are removed simultaneously Figure 5 shows thefront view of simulation result from the initial stage to 1 sec

It can be obtained from Figure 5 that gangue particleshave different motion trajectories under different airflowvelocity As shown in Figures 5(a) and 5(b) gangue particleswill do horizontal projectile motion before entering airflowdomain as can be seen at 119905 = 02 s and 119905 = 04 s As shown inFigure 5(a) velocities of conveyor belt 1 V

0and airflow 119906 are in

the opposite direction and gangue particles will change theirmotion trajectories to an opposite direction compared withthat of coal without the effect of airflow which can be seen at119905 = 06 s 119905 = 08 s and 119905 = 10 s As shown in Figure 5(b)velocities of the conveyor belt 1 V

0and airflow 119906 are in the

same direction and gangue particles will be blown muchfarther than that of coal without being affected by airflow ascan be seen at 119905 = 06 s 119905 = 08 s and 119905 = 10 s

As can be seen in Figures 5(a) and 5(b) gangue particleswill change their trajectories in areaA and areaCAreas B andD as shown in Figures 5(a) and 5(b) illustrated that gangueparticles with smaller diameters can be blown much farther

Advances in Materials Science and Engineering 7

Gangue hopperConveying belt 1

Conveying belt 2Collecting box

Fluid boundary

hp

0

u

(a) V0 and 119906 in the opposite direction

Gangue hopper

Conveying belt 1

Collecting box

Fluid boundary

Conveying belt 2

hp

0

u

(b) V0 and 119906 in the same direction

Figure 4 Air-solid multiphase pneumatic separation

BA

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(a) V0 and 119906 are in the opposite direction

DC

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(b) V0 and 119906 are in the same direction

Figure 5 Front view (initial stage to 1 sec apparent velocity plusmn260ms)

Table 1 Materials properties

Parameter Value UnitsBall

Diameter 50sim100 mmNumber 300 mdashDensity 2700 kgm3

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 07 mdash

AirDensity 1205 kgm3

Viscosity 18 times 10minus6 PasdotsWall

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 03 mdash

In order to study the pneumatic separation distanceinfluenced by different airflow directions airflow velocitiesplusmn300ms are chosen for the research of separation effectConveyor velocity V

0and height difference ℎ

119901between con-

veyor belt and air nozzle are kept constant The relationshipsbetween separation distances and particle diameters underdifferent airflow velocity directions are shown in Figure 6

45 50 55 60 65 70 75 80 85 90 95 100 10514

16

18

20

22

24

26

28

30

32

ΔS

(m)

d (mm)u = minus300ms and 0 = 1msu = 300ms and 0 = 1ms

Figure 6The relationship between particle diameter and separationdistance under different direction airflow velocity (119906 = plusmn300msV0= +1ms)

As can be seen in Figure 6 separation distance decreaseswith the increase of particle diameter It also can be con-cluded that the separation effect under airflow velocityminus300ms is better than that under airflow velocity 300ms

8 Advances in Materials Science and Engineering

Table 2 Levels of factors

Level Factor119860 gangue diameter 119889 (mm) 119861 height difference ℎ

119901(m) 119862 conveyor belt velocity V

0(ms)

1 50 040 0502 80 080 1003 100 110 200

Table 3 Experimental results of separation effect

Experiment number 119860 119861 119862Separation distance Δ119878 (m)

119906 and V0are in the opposite direction 119906 and V

0are in the same direction

1 50 04 05 305 2212 50 08 10 22 1143 50 11 20 205 0784 80 04 10 215 1085 80 08 20 148 0656 80 11 05 125 0427 100 04 20 185 0988 100 08 05 155 0379 100 11 10 140 028

The separation effect also can be analyzed from (9) dynamicpressure differenceΔ119875

119889can be expressed as 120588(1199062minus2)2when

119906 and V0are in the same direction while when 119906 and V

0are in

the opposite direction the dynamic pressure difference Δ119875119889

can be expressed as 120588(1199062 + 2)2 Thus it can be obtainedfrom the above analysis that separation effect under airflowvelocity minus300ms is significantly better than that of underairflow velocity 300ms

33 Orthogonal Experiment of Pneumatic Separation Digitalimage processing technology has been used to identify thetarget of various patterns of coal and gangue in undergroundpneumatic separation system Before mineral materials aresent to machine vision system coal and gangue have beencrushed to 100mm by impact crusher The size of materialsis ranging from 50mm to 100mm These materials are sentby the conveyor belt for coal and gangue digital imageinformation identification (as shown in Figure 1)

In this study separation distance Δ119878 (m) is selectedas the primary index to evaluate the pneumatic separationeffect Pneumatic separation influenced by different airflowdirection can be achieved by changing the arrangement ofhigh-pressure value The pneumatic separation test systemmainly consists of conveyor belt queuing system machinevision system control system and high-pressure air injectionsystem When coal and gangue materials fell down from theconveyor belt gangue will be identified by machine visionsystem and the information is transported to control systemthrough image sensor to drive the electromagnetic valueopenThus coal gangue pneumatic separation is realizedTheair compressor used in pneumatic separation testing systemis LG-6510 its working pressure is 10Mpa and certifiedcapacity is 65m3min

From the analysis shown in Section 3 the conveyorvelocity V

0 height difference ℎ

119901 and gangue diameter 119889

are selected as the three factors Factors and levels arelisted in Table 2 According to the identified level of factorsorthogonal table L

9(34) is applied in the test Orthogonal

experiment arrangement and results are shown in Table 3

4 Result and Discussion

41 Variance Analysis and Range Analysis Based on Orthogo-nal Test As can be seen from Table 2 each factor at differentlevels is approximate linearity so the method of regressionanalysis is to obtain the linear function relationship betweenthem appropriately [25] Thus the regression equationsof coal gangue separation distance under different airflowdirection are derived Equation (25) represents the regressionequationwhen 119906 and V

0are in the opposite direction and (26)

represents the regression equation when 119906 and V0are in the

same direction where 119910 is the separation distance of coal andgangue 119909

1represents the diameter of gangue 119909

2is the height

difference and 1199093is the velocity of conveyor belt

119910 = 467 minus 00191199091minus 1773119909

2+ 0092119909

3 (25)

119910 = 360 minus 0021199091minus 1599119909

2minus 014119909

3 (26)

Variance analysis is carried out on the regression equationto make significance test and the results are shown inTables 4 and 5 In order to determine the optimal pneumaticseparation solutions comparative analyses of the rangesbetween various levels of each factor are carried out Therange analysis is shown in Table 6

In Table 6 subscripts 1 and 2 used in influence factors(119860 119861 119862) represent the fact that 119906 and V

0are in the opposite

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

Advances in Materials Science and Engineering 7

Gangue hopperConveying belt 1

Conveying belt 2Collecting box

Fluid boundary

hp

0

u

(a) V0 and 119906 in the opposite direction

Gangue hopper

Conveying belt 1

Collecting box

Fluid boundary

Conveying belt 2

hp

0

u

(b) V0 and 119906 in the same direction

Figure 4 Air-solid multiphase pneumatic separation

BA

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(a) V0 and 119906 are in the opposite direction

DC

t = 02 s t = 04 s t = 06 s

t = 08 s t = 10 s(b) V0 and 119906 are in the same direction

Figure 5 Front view (initial stage to 1 sec apparent velocity plusmn260ms)

Table 1 Materials properties

Parameter Value UnitsBall

Diameter 50sim100 mmNumber 300 mdashDensity 2700 kgm3

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 07 mdash

AirDensity 1205 kgm3

Viscosity 18 times 10minus6 PasdotsWall

Normal stiffness 1 times 106 NmShear stiffness 1 times 106 NmFriction coefficient 03 mdash

In order to study the pneumatic separation distanceinfluenced by different airflow directions airflow velocitiesplusmn300ms are chosen for the research of separation effectConveyor velocity V

0and height difference ℎ

119901between con-

veyor belt and air nozzle are kept constant The relationshipsbetween separation distances and particle diameters underdifferent airflow velocity directions are shown in Figure 6

45 50 55 60 65 70 75 80 85 90 95 100 10514

16

18

20

22

24

26

28

30

32

ΔS

(m)

d (mm)u = minus300ms and 0 = 1msu = 300ms and 0 = 1ms

Figure 6The relationship between particle diameter and separationdistance under different direction airflow velocity (119906 = plusmn300msV0= +1ms)

As can be seen in Figure 6 separation distance decreaseswith the increase of particle diameter It also can be con-cluded that the separation effect under airflow velocityminus300ms is better than that under airflow velocity 300ms

8 Advances in Materials Science and Engineering

Table 2 Levels of factors

Level Factor119860 gangue diameter 119889 (mm) 119861 height difference ℎ

119901(m) 119862 conveyor belt velocity V

0(ms)

1 50 040 0502 80 080 1003 100 110 200

Table 3 Experimental results of separation effect

Experiment number 119860 119861 119862Separation distance Δ119878 (m)

119906 and V0are in the opposite direction 119906 and V

0are in the same direction

1 50 04 05 305 2212 50 08 10 22 1143 50 11 20 205 0784 80 04 10 215 1085 80 08 20 148 0656 80 11 05 125 0427 100 04 20 185 0988 100 08 05 155 0379 100 11 10 140 028

The separation effect also can be analyzed from (9) dynamicpressure differenceΔ119875

119889can be expressed as 120588(1199062minus2)2when

119906 and V0are in the same direction while when 119906 and V

0are in

the opposite direction the dynamic pressure difference Δ119875119889

can be expressed as 120588(1199062 + 2)2 Thus it can be obtainedfrom the above analysis that separation effect under airflowvelocity minus300ms is significantly better than that of underairflow velocity 300ms

33 Orthogonal Experiment of Pneumatic Separation Digitalimage processing technology has been used to identify thetarget of various patterns of coal and gangue in undergroundpneumatic separation system Before mineral materials aresent to machine vision system coal and gangue have beencrushed to 100mm by impact crusher The size of materialsis ranging from 50mm to 100mm These materials are sentby the conveyor belt for coal and gangue digital imageinformation identification (as shown in Figure 1)

In this study separation distance Δ119878 (m) is selectedas the primary index to evaluate the pneumatic separationeffect Pneumatic separation influenced by different airflowdirection can be achieved by changing the arrangement ofhigh-pressure value The pneumatic separation test systemmainly consists of conveyor belt queuing system machinevision system control system and high-pressure air injectionsystem When coal and gangue materials fell down from theconveyor belt gangue will be identified by machine visionsystem and the information is transported to control systemthrough image sensor to drive the electromagnetic valueopenThus coal gangue pneumatic separation is realizedTheair compressor used in pneumatic separation testing systemis LG-6510 its working pressure is 10Mpa and certifiedcapacity is 65m3min

From the analysis shown in Section 3 the conveyorvelocity V

0 height difference ℎ

119901 and gangue diameter 119889

are selected as the three factors Factors and levels arelisted in Table 2 According to the identified level of factorsorthogonal table L

9(34) is applied in the test Orthogonal

experiment arrangement and results are shown in Table 3

4 Result and Discussion

41 Variance Analysis and Range Analysis Based on Orthogo-nal Test As can be seen from Table 2 each factor at differentlevels is approximate linearity so the method of regressionanalysis is to obtain the linear function relationship betweenthem appropriately [25] Thus the regression equationsof coal gangue separation distance under different airflowdirection are derived Equation (25) represents the regressionequationwhen 119906 and V

0are in the opposite direction and (26)

represents the regression equation when 119906 and V0are in the

same direction where 119910 is the separation distance of coal andgangue 119909

1represents the diameter of gangue 119909

2is the height

difference and 1199093is the velocity of conveyor belt

119910 = 467 minus 00191199091minus 1773119909

2+ 0092119909

3 (25)

119910 = 360 minus 0021199091minus 1599119909

2minus 014119909

3 (26)

Variance analysis is carried out on the regression equationto make significance test and the results are shown inTables 4 and 5 In order to determine the optimal pneumaticseparation solutions comparative analyses of the rangesbetween various levels of each factor are carried out Therange analysis is shown in Table 6

In Table 6 subscripts 1 and 2 used in influence factors(119860 119861 119862) represent the fact that 119906 and V

0are in the opposite

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

8 Advances in Materials Science and Engineering

Table 2 Levels of factors

Level Factor119860 gangue diameter 119889 (mm) 119861 height difference ℎ

119901(m) 119862 conveyor belt velocity V

0(ms)

1 50 040 0502 80 080 1003 100 110 200

Table 3 Experimental results of separation effect

Experiment number 119860 119861 119862Separation distance Δ119878 (m)

119906 and V0are in the opposite direction 119906 and V

0are in the same direction

1 50 04 05 305 2212 50 08 10 22 1143 50 11 20 205 0784 80 04 10 215 1085 80 08 20 148 0656 80 11 05 125 0427 100 04 20 185 0988 100 08 05 155 0379 100 11 10 140 028

The separation effect also can be analyzed from (9) dynamicpressure differenceΔ119875

119889can be expressed as 120588(1199062minus2)2when

119906 and V0are in the same direction while when 119906 and V

0are in

the opposite direction the dynamic pressure difference Δ119875119889

can be expressed as 120588(1199062 + 2)2 Thus it can be obtainedfrom the above analysis that separation effect under airflowvelocity minus300ms is significantly better than that of underairflow velocity 300ms

33 Orthogonal Experiment of Pneumatic Separation Digitalimage processing technology has been used to identify thetarget of various patterns of coal and gangue in undergroundpneumatic separation system Before mineral materials aresent to machine vision system coal and gangue have beencrushed to 100mm by impact crusher The size of materialsis ranging from 50mm to 100mm These materials are sentby the conveyor belt for coal and gangue digital imageinformation identification (as shown in Figure 1)

In this study separation distance Δ119878 (m) is selectedas the primary index to evaluate the pneumatic separationeffect Pneumatic separation influenced by different airflowdirection can be achieved by changing the arrangement ofhigh-pressure value The pneumatic separation test systemmainly consists of conveyor belt queuing system machinevision system control system and high-pressure air injectionsystem When coal and gangue materials fell down from theconveyor belt gangue will be identified by machine visionsystem and the information is transported to control systemthrough image sensor to drive the electromagnetic valueopenThus coal gangue pneumatic separation is realizedTheair compressor used in pneumatic separation testing systemis LG-6510 its working pressure is 10Mpa and certifiedcapacity is 65m3min

From the analysis shown in Section 3 the conveyorvelocity V

0 height difference ℎ

119901 and gangue diameter 119889

are selected as the three factors Factors and levels arelisted in Table 2 According to the identified level of factorsorthogonal table L

9(34) is applied in the test Orthogonal

experiment arrangement and results are shown in Table 3

4 Result and Discussion

41 Variance Analysis and Range Analysis Based on Orthogo-nal Test As can be seen from Table 2 each factor at differentlevels is approximate linearity so the method of regressionanalysis is to obtain the linear function relationship betweenthem appropriately [25] Thus the regression equationsof coal gangue separation distance under different airflowdirection are derived Equation (25) represents the regressionequationwhen 119906 and V

0are in the opposite direction and (26)

represents the regression equation when 119906 and V0are in the

same direction where 119910 is the separation distance of coal andgangue 119909

1represents the diameter of gangue 119909

2is the height

difference and 1199093is the velocity of conveyor belt

119910 = 467 minus 00191199091minus 1773119909

2+ 0092119909

3 (25)

119910 = 360 minus 0021199091minus 1599119909

2minus 014119909

3 (26)

Variance analysis is carried out on the regression equationto make significance test and the results are shown inTables 4 and 5 In order to determine the optimal pneumaticseparation solutions comparative analyses of the rangesbetween various levels of each factor are carried out Therange analysis is shown in Table 6

In Table 6 subscripts 1 and 2 used in influence factors(119860 119861 119862) represent the fact that 119906 and V

0are in the opposite

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

Advances in Materials Science and Engineering 9

Table 4 Variance analysis (119906 and V0are in opposite direction)

Factor Squariance DOF Mean square 119865 value119860 1346 2 0673 Distinctively119861 1013 2 0506 Distinctively119862 0041 2 0020 DistinctivelyRegression 2163 3 0721 11752Error 0307 5 0061Sum 2470 8

Table 5 Variance analysis (119906 and V0are in the same direction)

Factor Squariance DOF Mean square 119865 value119860 1160 2 058 Distinctively119861 1411 2 0705 Distinctively119862 0067 2 0034 DistinctivelyRegression 2509 3 0836 15397Error 0272 5 0054Sum 2781 8

direction and the same direction respectively 119870119895119898

(119898 =

1 2 119899) is the sumof index values corresponding to factorsin column 119895 at level 119898 The value of 119870

119895119898determines the

optimal level and combination of factors in column 119895 119877119895

reflects the ranges of the index with the variation of factorsin column 119895 and the influence of the factor will be moresignificant if the value 119877

119895is greater

According to Tables 4 and 5 all the three factorsmentioned above have significant influence on pneumaticseparation distance As shown in Table 6 the pneumaticseparation distances Δ119878 are all decreasing with the increasedvalue of the three factors The analysis shows that ganguediameters have the most significant influence on separationdistance

42 Analysis of Experiment Results Based on Support Vec-tor Machine (SVM) From the above analysis shown inSection 41 the significant degrees of different factors for sep-aration distance are obtained by variance analysis Besidesthe primary and secondary relations of the influence fac-tors with pneumatic separation distance could be obtainedaccording to the range analysis However the optimal com-bination is a relative definition for the limited levels andhas great one-sided characteristic Most cases occurred inthe experiment the so-called ldquooptimal combination of thefactorsrdquo is a relative optimal not the real optimal

For further analysis of the experiment result SVM [26ndash30] is introduced The detailed functional forms of SVM are

given in the Appendix Optimization settings for factors thathave influences on coal gangue pneumatic separation couldbe divided into the following steps (1) collect the necessarydata using orthogonal experiment (2) set SVM learningmodel parameters and determine SVM kernel function (3)input learning samples and obtain parameters (4) establishthe fittingmodel according to parameters obtained above (5)determine the levels of parameters in a certain range andthen combine these levels to establish a large number of inputvector samples (6) input vector sample into the fitting modeland then obtain the output sample

The relationship of separation distance Δ119878 and variousinfluence factors is obtained through the above analysis Asshown in Figure 7 119909- and 119910-axes represent two of the threeinfluence parameters respectively 14 values equally spacedfrom the range of parameters of orthogonal experimentare taken respectively Thus the comprehensive collectionof the two parameters could form 196 samples and theextreme value is selected in the third parameter in orthogonalexperiment

Figures 7(a) and 7(b) show the relationships of separationdistance Δ119878 and the three factors when 119906 and V

0are in the

opposite direction As can be seen fromFigures 7(a) and 7(b)pneumatic separation distanceΔ119878 decreases with the increaseof conveyor velocity V

0 the height difference ℎ

119901 and gangue

diameter 119889 It also can be concluded that gangue diameter 119889has the greatest influence on separation distance Δ119878

Figures 7(c) and 7(d) have shown the relationships ofseparation distance Δ119878 and factors when 119906 and V

0are in

the same direction As can be seen from Figures 7(c) and7(d) there is the same variation trend as that of Figures 7(a)and 7(b) Through the analysis of the two groups of figurespneumatic separation effect will be better when 119906 and V

0

are in the opposite direction It can be concluded that theSVM intelligent model has important guiding significanceand practical value for coal gangue pneumatic separation

43 Correction of the Theoretical Formula Based on theleast square method (23) and (24) in Section 32 can betransformed into a function of 119896

119899and 119896119903parameters through

variable substitution then set up equations based on theexperiments data The result of 119896

119899and 119896

119903can be calculated

finallyWhen V

0and 119906 are in the same direction nonlinear

correction term can be given as 119896119899= 00205 and linear

correction term 119896119903= 11 Thus the formula can be expressed

as

Δ119878 = 119878119894+ 1198781198951+ 1198781198911minus 1198780

= V0radic2ℎ119901

119892+119898 ln (119890(119898 ln(119906V0minus1)minus002051198601199051198951120588119906)119898 + 1) minus 2119898119906 ln (119906V

0) + 119860119905

1198951120588119906

2119860120588119906

+2ℎ119891119906119890002051198601199051198951119906120588119898

(119890ln(119906V0minus1) + 119890002051198601199051198951119906120588119898) (1198921199051198951+ 2V119901) 1199051198951

minus V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892minus 11

(27)

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

10 Advances in Materials Science and Engineering

Table 6 Range analysis

119860 119861 119862

1198601

1198602

1198611

1198612

1198621

1198622

1198701198951

730 413 705 417 585 3001198701198952

488 215 523 216 575 2491198701198953

48 163 47 148 538 243119877119895

083 083 078 090 016 019

Table 7 Experimental and calculated results of separation effect

Experimentnumber

Experimentvalue Δ119878 (m)

Calculatedvalue Δ1015840119878 (m) Error ()

Experimentvalue Δ119878(m)

Calculatedvalue Δ1015840119878 (m) Error ()

119906 and V0are in the same direction 119906 and V

0are in opposite direction

1 221 233 515 305 316 3482 114 121 578 22 237 7173 078 092 152 205 221 7234 108 123 122 215 236 8895 065 071 845 148 169 12426 042 049 143 125 143 12587 098 114 1403 185 207 10628 037 044 159 155 174 10929 028 033 152 140 156 1026

When V0and 119906 are in the opposite directions nonlinear

correction term 119896119899and linear correction term 119896

119903can be given

as 119896119899= 156 times 105 and 119896

119903= 44 Thus the formula is obtained

as

Δ119878 = 1198780+ 1198781198952+ 1198781198912minus 119878119894

= V0radic2 (ℎ119901+ ℎ119895+ ℎ119891)

119892

+

2119898 ln(cos((156 times 1051198601199051198952119906120588 minus 2119898 arccos(119906radicV2

0+ 1199062)) 2119898)) minus 2119898 ln(119906radicV2

0+ 1199062)

119860120588

+ tan(1198601199051198952120588119906 minus 2119898 arccos(119906radicV2

0+ 1199062)

2119898)

2119906ℎ119891

1199051198952(1198921199051198952+ 2V119901)minus V0radic2ℎ119901

119892minus 44

(28)

In order to verify the effectiveness of theoretical formulanine samples shown in Table 3 are selected as verificationsamples The comparison of experiment results and calcu-lated results is shown in Table 7

As can be seen from Table 7 calculated value and exper-iment value of coal gangue pneumatic separation distancehave high consistent degrees the separation distance errorbetween experiment value and calculated result is less forgangue with smaller diameter and the separation distanceerror increases with the increase of gangue diameter How-ever the separation distance error between experiment value

and calculated result is still less than 16 Thus it canbe concluded that (27) and (28) have important guidingsignificance and practical value for coal gangue pneumaticseparation

5 Conclusions

The established coal and gangue pneumatic separationmodelreflects the basis motion law of gangue affected by airflowand coal without being affected by airflow which providestwo feasible solutions for underground pneumatic separation

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

Advances in Materials Science and Engineering 11

0

1

0505

15

12

15

15

25

35

2

3

4

hp (mm)

ΔS

(m)

0 (ms)

(a) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in opposite direction)

50 60 70 80 90 100

051

15

15

25

35

2

2

3

4

d (mm)

ΔS

(m)

0 (ms)

(b) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in opposite direction)

04 06 08 10 12 141

2

1

2

ΔS

(m)

hp (mm)0 (ms)

15

15

25

05

05

(c) Influence of conveyor belt velocity V0 and height difference ℎ119901 ongangue separation distance (119906 and V0 in the same direction)

50 60 70 80 90 100051

152

05

1

15

2

25

3

ΔS

(m)

d (mm)

0 (ms)

(d) Influence of conveyor belt velocity V0 and gangue diameter 119889 ongangue separation distance (119906 and V0 in the same direction)

Figure 7 Influence of three factors on gangue separation distance

under two kinds of arrangement scheme of high-pressurevalue The analysis in theory is consistent with that obtainedin the experiments which validate the established theoreticalmodel and present the following conclusions

(1) Different high-pressure value arrangement schemeshave great influence on gangue pneumatic separation thetheoretical formulas of coal gangue pneumatic separationdistance affected by different airflow direction are derivedand the expressions of the two formulas are different underdifferent airflow direction

(2) A series of air-solid multiphase flow simulations andorthogonal experiments were conducted to clarify its effectunder different airflow direction Based on the analysispneumatic separation effectwill be better under the conditionof 119906 and V

0being in the opposite direction Pneumatic

separation distance Δ119878 decreases with the increased valuesof the three factors (conveyor velocity V

0 height difference

ℎ119901 and gangue diameter 119889) These analyses also show that

gangue diameters have the most significant influence onseparation distance followed by conveyor velocity V

0and

height difference ℎ119901

(3)The relationship of pneumatic separation distance Δ119878and influence factors was obtained by SVM intelligentmodelthe theoretical formulas of coal gangue pneumatic separationdistance are corrected based on the analysis of orthogonalexperiment data The corrected formula is suitable to serveas the theory basis of coal gangue pneumatic separation

Appendix

Mathematical Quantity for Support VectorMachine Analysis

Original research based on SVMwas originally used in linearfitting problem If function 119891(119909) appears with linear functioncharacteristics it can be expressed as 119910 = 120596x + 119887 Assumethat all the data (x

119894 y119894) (119894 = 1 2 119899) 119909 isin 119877119897 (119877119897 is the real

of 119897 degree) and 119910 isin 119877 Function 119910 can be fitted by linearfunction 119910 = 120596x + 119887 in precision 120576

1003816100381610038161003816y119894 minus 120596x119894 minus 1198871003816100381610038161003816 le 120576 (A1)

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

12 Advances in Materials Science and Engineering

where x119894is the input vector y

119894is a real constant as the output

vector120596 is a normal vector for fitting function 119887 is thresholdvalue and 120576 is the fitting precision

Based on the principle of minimum structural risk theoptimization objective could achieve better generalizationability at the minimum value of 11990822 Considering theexistence of approximation error 120585 (120585 is a real constant) inactual application therefore SVM can be expressed as

min 1

21199082+ 119862

119899

sum

119894=1

(120585119894+ 120585lowast

119894)

st 119910119894minus 119908119909 minus 119887 le 120576 + 120585

119894

119908119909119894+ 119887 minus 119910

119894le 120576 + 120585

lowast

119894

120585119894ge 0 120585

lowast

119894ge 0

(A2)

where 119862 is balance factor which is used to control the degreeof punishment beyond the error sample and 120585

119894and 120585lowast

119894are

relaxation factors 120585119894and 120585lowast119894are of the same nature in general

relaxation factor at the top of fitting curve is recorded as 120585119894

conversely it is recorded as 120585lowast119894

Equation (28) could change into quadratic programmingproblem based on dual theoryThen the Lagrange equation isestablished

119871 (119908 119887 120585119894 120585lowast

119894 120572119894 120572lowast

119894 120578119894 120578lowast

119894)

=1

21199082+ 119862

120572

sum

119894=1

(120585119894+ 120585lowast

119894)

minus

120572

sum

119894=1

120572119894(120576 + 120585

119894minus 119910119894+ 119908119909119894+ 119887)

minus

120572

sum

119894=1

120572lowast

119894(120576 + 120585

lowast

119894+ 119910119894minus 119908119909119894minus 119887)

minus

120572

sum

119894=1

(120578119894120585119894+ 120578lowast

119894120585lowast

119894)

(A3)

where parameters 120572119894and 120572lowast

119894are Lagrange multiplier 120572

119894≫ 0

120572lowast

119894≫ 0 and 120578

119894and 120578lowast119894are temporary variables 120578

119894≫ 0 120578lowast

119894≫

0 120572119894and 120572lowast

119894have the same physical significance with 120578

119894and

120578lowast

119894The optimal solution of (A1) could be derived by calculat-

ing the saddle points of the Lagrange equationThus functionapproximation problem can be obtained

119908 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) 119909119894

119891 (119909) = 119908119909 + 119887 = sum

119909119894isin119878SV

(120572119894minus 120572lowast

119894) (119909119894sdot 119909) + 119887

(A4)

where 119878SV in (A4) is the SVM and the training sample is thesupport vector when (120572

119894minus 120572lowast

119894) is not equal to zero

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Financial support for this work provided by National High-Tech Research andDevelopment Program of China (863 Pro-gram) (no 2012AA062102) Innovation Training Project ofGraduate Student in Jiangsu Province (CXLX13 936) and thePriority Academic Program Development of Jiangsu HigherEducation Institutions (PAPD) is gratefully acknowledged

References

[1] M G Qian J L Xu and X X Miao ldquoTechnique of cleaningmining in coal minerdquo Journal of China University of Mining ampTechnology vol 32 pp 343ndash348 2003

[2] J-X Zhang and X-X Miao ldquoUnderground disposal of waste incoal minerdquo Journal of China University of Mining amp Technologyvol 35 no 2 pp 197ndash200 2006

[3] C S Dong P X Yao and Z H Liu ldquoHydraulic automaticseparation technology of coal and refuse in undergroundminerdquoCoal Science and Technology vol 35 no 3 pp 54ndash56 2007

[4] J Li D Yang and C Du ldquoEvaluation of an undergroundseparation device of coal and ganguerdquo International Journal ofCoal Preparation andUtilization vol 33 no 4 pp 188ndash193 2013

[5] C Luo C Du L Xu and K Zheng ldquoFractal distribution studiesof a rotary crushing mechanismrdquo Recent Patents on MechanicalEngineering vol 7 no 1 pp 44ndash51 2014

[6] J-P Li C-L Du and L-J Xu ldquoImpactive crushing andseparation experiment of coal and ganguerdquo Journal of the ChinaCoal Society vol 36 no 4 pp 687ndash690 2011

[7] S Al-Thyabat and N J Miles ldquoAn improved estimation ofsize distribution from particle profile measurementsrdquo PowderTechnology vol 166 no 3 pp 152ndash160 2006

[8] J Tessier C Duchesne and G Bartolacci ldquoA machine visionapproach to on-line estimation of run-of-mine ore compositionon conveyor beltsrdquo Minerals Engineering vol 20 no 12 pp1129ndash1144 2007

[9] T Andersson M J Thurley and J E Carlson ldquoA machinevision system for estimation of size distributions by weight oflimestone particlesrdquoMinerals Engineering vol 25 no 1 pp 38ndash46 2012

[10] S Al-Thyabat N J Miles and T S Koh ldquoEstimation of the sizedistribution of particles moving on a conveyor beltrdquo MineralsEngineering vol 20 no 1 pp 72ndash83 2007

[11] E Hamzeloo M Massinaei and N Mehrshad ldquoEstimation ofparticle size distribution on an industrial conveyor belt usingimage analysis and neural networksrdquo Powder Technology vol261 pp 185ndash190 2014

[12] Y K Yen C L Lin and J D Miller ldquoParticle overlap and seg-regation problems in on-line coarse particle size measurementrdquoPowder Technology vol 98 no 1 pp 1ndash12 1998

[13] C L Lin Y K Yen and J D Miller ldquoPlant-site evaluations ofthe OPSA system for on-line particle size measurement frommoving belt conveyorsrdquoMinerals Engineering vol 13 no 8 pp897ndash909 2000

[14] C Aldrich G T Jemwa J C van Dyk M J Keyser and J H PVan Heerden ldquoOnline analysis of coal on a conveyor belt by useof machine vision and kernel methodsrdquo International Journalof Coal Preparation and Utilization vol 30 no 6 pp 331ndash3482010

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

Advances in Materials Science and Engineering 13

[15] J X Zhang T Chen Z D Yu andW Li ldquoXinjiang cotton seedcolor separation system based on computer visionrdquo Transac-tions of the Chinese Society of Agricultural Machinery vol 40no 10 pp 161ndash164 2009

[16] C Guo H Wang W Liang J G Fu and X Yi ldquoLiberationcharacteristic and physical separation of printed circuit board(PCB)rdquoWasteManagement vol 31 no 9-10 pp 2161ndash2166 2011

[17] M Xu G M Li J Yin andW Z He ldquoCrushing and pneumaticseparation of printed circuit board scrapsrdquo EnvironmentalScience amp Technology vol 30 pp 72ndash74 2007

[18] V Kumar J-C Lee J Jeong M K Jha B-S Kim andR Singh ldquoNovel physical separation process for eco-friendlyrecycling of rare and valuable metals from end-of-life DVD-PCBsrdquo Separation and Purification Technology vol 111 pp 145ndash154 2013

[19] V Kumar J-C Lee J Jeong M K Jha B-S Kim and RSingh ldquoRecycling of printed circuit boards (PCBs) to generateenriched rare metal concentraterdquo Journal of Industrial andEngineering Chemistry vol 21 pp 805ndash813 2015

[20] N Hayashi and T Oki ldquoEffect of orifice introduction onthe pneumatic separation of spherical particlesrdquo MaterialsTransactions vol 55 no 4 pp 700ndash707 2014

[21] T Havlik D Orac M Berwanger and A Maul ldquoThe effectof mechanical-physical pretreatment on hydrometallurgicalextraction of copper and tin in residue from printed circuitboards from used consumer equipmentrdquoMinerals Engineeringvol 65 pp 163ndash171 2014

[22] Z Liu Y Xie Y Wang J Yu S Gao and G Xu ldquoTandem flu-idized bed elutriatormdashpneumatic classification of coal particlesin a fluidized conveyerrdquo Particuology vol 10 no 5 pp 600ndash6062012

[23] G-H Yang D-C Zheng J-H Zhou Y-M Zhao and Q-RChen ldquoAir classification ofmoist raw coal in a vibrated fluidizedbedrdquoMinerals Engineering vol 15 no 8 pp 623ndash625 2002

[24] X Yang Z Fu J Zhao E Zhou andY Zhao ldquoProcess analysis offine coal preparation using a vibrated gas-fluidized bedrdquoPowderTechnology vol 279 pp 18ndash23 2015

[25] K T Fang C X Ma and J K Li ldquoRecent development oforthogonal factorial designs and their applicationsmdashapplica-tions of regression analysis to orthogonal designsrdquo Applicationof Statistics and Management vol 18 pp 44ndash49 1999

[26] X H Guo and X P Ma ldquoSupport vector machine toolbox inMatlab environmentrdquo Computer Applications and Software vol24 no 12 pp 57ndash59 2007

[27] X Fang Z-J Ding and X-Q Shu ldquoHydrogen yield predictionmodel of hydrogen production from low rank coal basedon support vector machine optimized by genetic algorithmrdquoJournal of the China Coal Society vol 35 no 1 pp 205ndash2092010

[28] Y Q Qiu G H Hu and W L Pan ldquoParallel algorithm ofsupport vector machine based on orthogonal arrayrdquo Journal ofYunnan University vol 28 no 2 pp 93ndash97 2006

[29] J A K Suykens and J Vandewalle ldquoLeast squares supportvector machine classifiersrdquo Neural Processing Letters vol 9 no3 pp 293ndash300 1999

[30] J A K Suyken and J Vandewalle ldquoSparse least squares SupportVector Machine classifiersrdquo in Proceedings of the 8th EuropeanSymposium on Artificial Neural Networks (ESANN rsquo00) pp 37ndash42 Bruges Belgium April 2000

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials

Submit your manuscripts athttpwwwhindawicom

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CorrosionInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Polymer ScienceInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CeramicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CompositesJournal of

NanoparticlesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Biomaterials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

NanoscienceJournal of

TextilesHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Journal of

NanotechnologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

CrystallographyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CoatingsJournal of

Advances in

Materials Science and EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Smart Materials Research

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MetallurgyJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BioMed Research International

MaterialsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Nano

materials

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofNanomaterials