AN EXPERIMENTAL STUDYON THE COMBUSTION CHARACTERISTICS OF SAWDUST

IN A CYCLONE COMBUSTOR

B. FungtammasanDept. of Mechanical Engineering

King Mongkut's Institute of Technology North BangkokBangsue, Bangkok

THAILAND

P. JittrepitProvincial Water Works

Bangkhen, BangkokTHAILAND

J. Torero & P. JoulainLaboratoire de Chimie - Physique de la Combustion

Vniversite de Poi tiers - UA 872 au CNRSENSMA BP 109 - Site du Futuroscope

86960 Futuroscope CedexFRANCE

ABSTRACT

This paper discusses experimental results on the combustion characteristics of sawdustin a cyclone combustor in which the fuel (sawdust) and air are injected into the combustionchamber through two tangential inlets. Data obtained for flow visualization under isothermalflow conditions, flame stability, combustion temperature mapping and combustion-productcomposition analyses are presented. It is observed that sawdust particles only follow the airflow when the particles are very fine. Thus the residence time of sawdust particles is independentof air flow rate but decreases with the sawdust flow rate. Flame stability tests indicate thatstable flames can be obtained, with stability limits being dependent on sawdust feed rate.Data on spatial temperature distributions and composition profiles at the exit plane providedinformation on the location of combustion zones. The combustion zone can be displaced inbetween two extreme modes: one with combustion occuring primarily inside the chamber andthe other primarily in and around the exit sleeve and outside the cyclone exit.

INTRODUCTION

Sawdust, a waste product from the processing of wood, is an important source of energy

for a variety of industries. The fishmeal industry, for example, consumes approximately 1.25x 106 m3

of sa wdust annuall y for process steam generation [ 1]. In general, the processing of woodin saw mills yields 10 - 13 % of sawdust. Due to its high volatile content (typically in the rangeof 60 - 70 %) and relatively low calorific value (17 - 18 MJ/kg), sawdust has not been burntefficiently in conventional combustors. There is thus a need to develop an efficient process,withrespect to heat release and pollutant formation, for burning sawdust.

A cyclone combustor, as its name implies, is derived from cyclone dust separatortechnology. Thus one of the most remarkable capabilities of a cyclone combustor is theseparation of the ash from the gaseous products of solid fuel combustion [2]. Other distinctadvantages include the swirl effect generated by the tangential inlet of air and fuel; rapid mixingas well as recirculation of heat and active chemical species (alwalys accompanied by hightUrbulent intensity levels) ; and the very long residence times for the fuel panicles (of theorder of fifteen times the average velocity in the cyclone) [3, 4]. All these factors tend toprovide a stabilised flame and an efficient burning process. The cyclone combustor has beenused satisfactorily for the combustion of materials which are otherwise known as "difficultfuels", such as vegetable refuse, high ash content and brown coals [5, 6], anthracite, highsulphur oils [7], and waste gases with low calorific value [8]. Materials of high volatile contentare also included in this category [2, 3].

Many different types of cyclone combustor are in use, but, in general, five main typescan be considered and mainly differentiated by combustor orientation and by the relativeposition of air and fuel inlets to the combustor exit. An extensive review of the advantagesand disadvantages of each type, is provided by Gupta et al [9]. What is termed by Gupta et alas 'type l' cyclone combustor is used for fuels with high calorific value. These fuels sometimescontain large quantities of volatile matter and slag and ash generation and removal are nota serious problem. The 'type II' combustor is used for high ash content fuels when problemsdue to slag formation and fly ash carry-over are of concern. The use of this type of combustorfor burning high ash, high volatile content, poor quality coal has been demonstrated byKalishevskii [10]. Type III is more suited to low calorific value fuels, often containing highlevels of volatile matter. Type IV is often used for better-quality fuels as their particulateretention properties are far inferior to those of types I-III. Type V cyclones are used for materialprocessing.

In view of the high volatile content of sawdust and the concern for fly ash carry-over,which is characteristic of conventional sawdust combustion, the type II combustor has beenchosen for the present study. This combustor is a mere modification of standard cyclone dustseparators. The purpose is to obtain fundamental information on the flow and combustionpatterns of sawdust in this type of combustor. The parameters investigated are particletrajectories (through flow visualization) in the cy£lone, combustor load, flame stability, spatialtemperature distributions and combustion-product composition profiles.

EQUIPMENT

Cyclone Combustor

A two-inlet 'type II' [9] cyclone combustor was used throughout this study. It consistsof a vertical cylindrical chamber of 155 mm in diameter and 3 I 0 mm in length; two 28-mm-dia.tangertial inlets arranged at opposite directions on the top portion of the chamber; a 68-mm-dia.by 115-mm long exhaust nozzle at the top and an ash settling chamber at the bottom (Fig. 1 andtable 1). Although in principle a symmetrical arrangement of, say, four tangential inletsaround the periphery of the cyclone, are preferred to produce uniform flow patterns.optimum recirculation zones, and better flame stabilization and heat release characteristics [111],for simplicity, only two inlets were provided: one playing the role of primary air and fuel(sawdust) feeding, the other the role of dilution. They are both placed in the same plane to attainflow symmetry. The latter also provided flexibility in regulating the air/fuel ratio. Onedisadvantage of such a system is that much of the flow can quickly diffuse around the sleeveinto the exit and effectively by pass the main body of the cyclone. However, the injection ofsecondary air at selected positions along the length of the cyclone chamber can increase theproportion of flow reaching the end plate of the chamber [12]. Again, for simplicity in thispreliminary investigation, secondary air was not used.

All the components of the combustor were made of mild steel (Fig. 2 (a)). For isothermalflow visualization studies, a transparent perspex model of comparable dimensions to themildsteel combustor was used (Fig. 2 (b)). The relevant dimensions of the model are listed intable 1.

A design parameter called 'swirl number', S, which can be used to characterize cycloneflows is also shown in table 1. The definition of swirl number (with due modification for theeffects of combustion) is given in the Appendix. The swirl numbers of the combustor and theperspex model are 6.8 and 8.8 respectively, which are considered as typical for cyclonecombustors but much higher than for swirl burners [4].

Test Rig and Experimental Apparatus

The overall experimental set-up which consisted of the combustor, air and fuel supplysystems, and measurement instruments are illustrated in Fig. 3.

The Air and Fuel Supply

The fuel (sawdust) was injected into the combustor by a pneumatic conveying system.Stored in a hopper, the sawdust dropped freely into an injector chamber in ~hich compressedair was injected. Sawdust is injected into the combustor with the aid of an air pump. The air pumpconsists of a cylinder into which air is injected through tangential nozzles. The incoming aircreates a low pressure central core that blows the sawdust particles int;, a 28 mm pipe. Thesawdust flow rate was measured using a load cell installed at the bottom of the hopper. Themaximum sawdust supply capacity was 950 g/min. The compressed air flow rate was metredusing an orifice and a differential pressure transmitter, and regulated by a needle valve.

-

Dilution air was injected via the second combustor inlet port using a 0.5 kW centrifugalblower of 0.16-Nm3/sec capacity. The air flow rate was measured using a nozzle at the blower

inlet and a digital manometer, and regulated by a butterfly-valve.

Ignition System

A pilot ignition system was used to preheat the combustor up to the point of auto-ignition

of the sawdust. The system consisted of an LPG cylinder which supplies LPG fuel into thedilution stream through an injection nozzle. The gas/air mixture was then ignited by a spark- plugigniter in the combustion chamber.

Instrumentation

Temperature Measurements

All temperature measurements were made using chromel/alumel (type K) thermocouplewires of 0.65-mm diameter, which were gas welded to form a junction with a bead size ofl.50-mm diameter. For insulation and strength, the wires were threaded through a 100-mmlength 3.8-mm diameter twin bore (0.8 mm) ceramic tube. A traversing mechanism was usedto hold 12 thermocouples, enabling point measurement of temperatures both inside thecombustion chamber and at the cyclone exit to be made (Fig. 4). The maximum temperature thatcould be measured was 1,370°C.

Data Acquisition

The data acquistion system consisted a Yokogawa 3880 Hybrid Recorder (RH 2500 E)and a Yokogawa 388252 Remote Scanner with 30 input channels. The scanner was placedclosed to the test rig while the recorder could be located up to 30 meters away from the former.The system is capable of taking 300 samples in 2 seconds, and both analog and digital signalscould be recorded with a built-in dot-matrix printer. In this study the signals connected to thescanner included 12 thermocouple voltages and output signals from the load cell indicator,digital manometer and differential pressure transmitter.

Concentration Measurements

Concentrations of O2

and CO at the cyclone exit were measured using a Bacharach300 Combustion Analyzer which operates on the priciple of electrochemical cells. Themeasurement ranges were 0.1 - 23.5% for O2

and 0 - 3,900 ppm for CO. Due to hightemperature in the flame and the need to prevent further reaction of the gases inside thesampling chain, the sampling probe was water-cooled.

Other details of the experimental set-up has been reported in reference 11.

EXPERIMENTAL TECHNIQUES

Determination of Fuel Properties

To ensure consistency of the test results, the physical, chemical and thermal propenies

of the sawdust were determined prior to the experiments. The size distribution of sawdustwas found by sieve analysis. By limiting the largest mesh size to 1.0 mm, the mean particlediameter (dp) was found to be 0.29 mm. The bulk density was 184 kg/m3. The high heatingvalue of the sawdust was determined using a bomb calorimeter, and was found to be18,209 kJ/kg. The composition of the sawdust, analysed on an "as received basis", wasfound to be : 9.9 % moisture, 0.50 % ash, 67.0 % volatile matter, and 22.6 % fixed carbon.The ultimate analysis showed 44.6 % carbon, 6.5 % hydrogen, 48.3 % oxygen, 0.05 %nitrogen, and 0.06 % sulphur.

Model Studies

For flow visualization study of the panicles, the perspex model was illuminated using astroboscope. When the desired flow condition was established, a photograph of the flowpattern was taken. The changing patterns were also recorded by a video camera.

Experiments were carried out first by keeping the sawdust flow rate constant whilevarying the air flow rate. Then the sawdust feed rate was varied while keeping the air flow rateconstant.

Combustion Studies

ignition

Ignition of the sawdust was achieved by preheating the combustor to the sawdustignition temperature. This was done by injecting LPG fuel into the primary air stream andigniting the mixture with the spark plug. When the flame temperature at the centre of thecyclone exit reached about 1,3000C, sawdust was injected. An initial temperature drop wasobserved followed by a temperature increase to about 1,400 0C, at this point, LPG feedwas terminated. The time-histroy record of the flame temperature during this process isshown in figure 5. At the very beginning only LPG burns with air, the temperature then riseswith time because of the heat released by the LPG - air reaction. At 1, T = Tvolatilisation,i.e., the temperature reached the point at which the sawdust particles can vaporize. Between

1 and 2 , T decreases slightly because of the endothermic vaporisation of the sawdustParticles. At 2 enough gaseous fuel has been released to allow the ignition of the sawdust.Between 2 and 3 the sawdust and the LPG are burning with air, T then rises again as a resultof heat released from the above mentioned reactions. At 3 , LPG is turned off and only thesawdust burns. The temperature first decreases till equilibrium between heat release and lossesis achieved at approximately 1,0000C. Equilibrium is not attained until about 30 minutesafter ignition.

The radial distributions of O2, and CO concentrations in the plane of the nozzle exitwere determined by traversing the gas sampling probe radially from 0 - 1r

o, for mixture ratios

of 1.2, 1.4, 1.5, 1.6 and 2.2.

Flame Characterization

Flame stability was investigated by observing the effect of combustor loading andmixture ratio on the flame behaviour. For a given sawdust flow rate, the dilution air flow wasadjusted until a stable flame was obtained. The air flow rate would then be reduced graduallyuntil the flame was no longer stable, that is, the flash-back condition. To determine the blow-off limit for the same loading, the initial flame state was re-established and the air floowgradually increased until the blow off limit was reached. The changing characteristics of theflame was monitored at each stage and the limiting air flow rates noted. The same procedureswere then repeated for different sawdust flow rates.

Temperature Distributions and Composition Analyses

Temperature distributions both inside the combustor chamber and in the vicinity of thecyclone exit were measured by traversing the 12 thermocouples simultaneously along theradial direction. The positions of the thermocouples are shown in Fig. 6. Detailed measurementswere made at a fixed sawdust flow rate and varying mixture ratios (<o> which is defined as

Inside the chamber combustion is fuel rich allowing for the escape of unburnt fuel thatburns in a premixed fashion at the exit of the chamber.

RESULTS AND DISCUSSION

Isothermal Flow Patterns

Figs. 7 (a) to (h) show photographs of the sa wdust particle trajectories in the cyclonemodel under various flow conditions. It is obvious that the sawdust particles, due to its size(and hence weight) being too large, do not follow the air flow. Instead they form a descendingspiral against the chamber wall to the bottom of the cyclone. Only very fine panicles wereobserved in a video film to follow the ascending central vortex to the exit.

Considering pictures (a) and (b), and then (c) and (d), it seems despite the fact that ma

(air flow rate) increases, the residence time remains constant for a given m, (fuel flow rate).

For pictures (a), (e) and (f) where m is constant, the residence time appears shorter asm, increases. This is due to the fact that they are more panicles per unit air flow rate, so thatthe particles fall more rapidly under gravitational force. The same result is observed forpictures (g) and (h).

Also observed is the tendency, at higher sawdust flow rates (pictures (c) - (h), for

lobules of particles to break away from the outer wall boundary layer - a phenomena calledsaItation' [14], causing instability to the flow.

.

Flame Characteristics

Flmne Behaviour

Visual observations of the flame behaviour were made for a nominal sawdust flowrate of 100 g/min. with different air flow rates. The results are displayed in Fig. 8. At low airflow rate, there is a lack of air, ie. fuel rich combustion, the flame is shon, and it is close tothe extinction limit. Flame flash back was observed. It is a premixed flame with a lot ofunburnt fuel as the ash mixture coming out of the settling chamber was black at the surfacebut brown inside, and spontaneous combustion of the mixture could be observed if left outsidethe chamber. Since it is close to extinction the flame is highly unstable.

As ma increases, the flame appears more stable at first. There is enough heat to vaporizethesawdust, and the combustion is still fuel rich. The flame becomes wider and wider and moreluminous outside the combustion chamber. The nearly underventilated flame moves out of thechamber to suck the air needed to bum all the fuel.

Then as ma continues to increase, the flame appears more and more luminous. It isprimarily located inside the combustion chamber, while the external flame shrinks.

However as ma increases funher, the flame finally reaches the other extinction limit(fuel lean) or blow-off limit. There is an extinction of the flame around the sawdust particles,that is, the particles are panly burning outside and the blow-offlimit will be reached when theyare quenched at higher ma.

Stability Limits

Fig. 9 shows the effect offuel flow rate and mixture ratio on flame stability limits. Theresult corresponds perfectly to the explanation given above. For a given fuel flow rate, theblow-off limit is reached when ma or the freestream air velocity relative to the sawdustParticles causes the extinction of the flame surrounding the particles.

The lower limit is called the flash back limit or fuel rich extinction limit. Due to the largeheatrelease, lots of volatiles are produced but not burnt. The flame then acts as a premixed flameand flashes back.

Effect of Mixture Ratio on Temperature Distributions

Figs. 10 (a) - (e) show the spatial distribution of temperature for fuel flow rates of100 g/min. is +5+_ at o = 1.2, 1.3, 1.5, 1.8 and 2.2. Assuming that the flame front coincides withthe greatest temperature gradient, that is where the isotherms are closest together, it can be seen

that the flame fronts are generally located along the 700 - 750°C isotherms. Thus it is seenthat combustion occurs both inside and outside the combustion chamber in all cases (Figs. 10 and11 ). Inside the combustor, an annular flame front located close to the inner wall is established.The penetration of the flame inside the combustor depends on the mixtUre ratio. In generalas oo increases, the combustion becomes more and more air rich (or fuel lean), and the flamezone moves downward into the combustion chamber, where more and more fuel is burnt. At

Oo = 2.2 for example, the flame zone extends deep down to the end plate with a region ofhighly uniform, high temperature distribution covering the main body of the chamber, indicatingthat a very stable combustion condition was established.

The swirl numbers under combustion conditions for O = 1.2, 1.3, 1.5, 1.8 and 2.2 areestimated to be 1.55, 1.57, 1.52 1.55 and 1.94 respectively, which are much smaller thanthe non-combustive swirl number of 6.8. However, the higher combustive swirl number forO = 2.2 in comparison to those for other mixture ratios suggests that the higher degree ofswirl was responsible for stabilizing the combustion condition as the sawdust particle residencetime was found to be independent of air flow rate for a given sawdust flow rate.

The highest temperature recorded is 1,225°C for O = 1.5. After due corrections weremade for radiation heat loss from the thermocouple probe, the maximum temperature achievedis more likely to be about l,275OC. 1

Effect of Mixture Ratio on Product Composition

Radial distributions of the composition of combustion products (O2 and CO) at thenozzle exit plane are shown in Figs. 12 (a) - (d) for O = 1.2, 1.4, 1.6 and 2.2 at sawdust flowrates of l00+_5 g/min. Generally, the o

2

concentration varies from a low level at the combustorcentre1ine, where intense combustion occurs, to a higher level at r/ro = 1. The o 2

concentrationin the radial direction increased most significantly for o = 2.2 where the flame zone is welldefined and compact. As o increases from 1.2 to 1.6, the centreline o

2

concentration increasesfrom 4.1 % to 10 %, but falls back to 6.5 % for o = 2.2 due to high o 2

consumption rateinside the chamber.

The distribution of CO concentration is relatively uniform in the flame zone for o = 1.2to 1.6, and falls sharply outside the flame zone. However, for o = 2.2, the CO concentrationis very low at the centreline due to high oxidation rate inside the chamber and the narrow flamezone at the nozzle exit (Fig. 10 (c). The highest CO concentration recorded is 3,196 ppmfor the low mixture ratio case of O = 1.2.

CONCLUSTIONS

1. Isothermal flow studies provided information on sawdust particle flow behaviour.

Most significantly, the residence time of the particles seems to be dependent on sawdustflow rate only, but not on air flow rate.

2. In general, stable flames can be obtained for the cyclone combustor tested.

3. The mixture ratio (o) is seen to have a strong influence on the location of the flame zoneRich mixtures tend to produce long and divergent flames at the exhaust and combustionzones in and around the exit sleeve, while lean mixtures tend to push the flame zonedeeper inside the combustion chamber and shorter and compact flames are obtained atthe exit. High swirl in the chamber tend to stabilise the flame and produce a uniformcombustion zone in the chamber.

4. With the highest CO concentration observed at 3,196 ppm and 2,173 ppm for o = 1.2and 2.2 respectively, the level of CO emission is seen to be moderate.

5. The measurement of NOx concentration has been omitted in this study due to lack ofrelevant instrumentation. It is recommended however that NOx formation should bemonitored in future studies since temperatures exceeding 1,2000C have been observed.

1. Pitakarnnop, N. and Teptaranon, S., Proc. ASEAN - SCNCER-T Seminar on EnergyConservation in Industry, Ubon Ratchathani, 1991.

2. Agrest, J., Institut Francais des Combustibles et de /'Energie, 543 (1964).

3. Agrest, J., J.lnst. Fuel, 38, 344, August 1965.

4. Syred, N. and Beer, J. M., Combustion in Swirling Flows: A Review, Combustion GillFlame, Vol. 23,1974.

5. Baluev, E. D. and Troyankin, Y.V., Thermal Eng., 14(1),84,1967.

6. Glebov, V. P., Motin, G. 1., Yakhilevich, F. M. and Fedyunin, L. A., Thermal Eng.22(3),8, 1975.

7. Tsarev, V. K. and Troyankin, Y. V., Thermal Eng., 18(12), 1971.

8. Tager, S. A., Molin, G. 1., Talumaa, R. Y., Kalmark, A. M. and Galyaenko, A. B.Thermal Eng., 18(4),80, 1971.

9. Gupta, A. K., Lilley, D. G. and Syred, N. Swirl Flows, Ch. 5, Abacus Press, Kent, 1995

10. Kalishevskii, L. L. and Ganchev, B. G., Thermal Eng., 12 (6), 75, 1965.

11. Kalishevskii, L. L. and Ganchev, B. G., Thermal Eng., 13 (21), 22, 1963.

12. Katsnelson, B. D. and Bogdanov, B. A., Thermal Eng., 17 (4), 82,1970.

REFERENCES

Forthe characterisation of swirling flow which is generated by tangential entry of fluids,the most commonly used parameter is the 'swirl number', S [15],

where Go = the axial flux of angular momentumGx = the axial flux of linear momentumDe = the exit diameter of swirl generator

APPENDIX

It has also been shown [4] that the swirl number as defined is a suitable parameter forcharacterising cyclone chambers. Here, based on the input angular momentum and exit axialmomentum, S becomes

Swirl Number

where AT

is the cross-sectional area of the tangential inlet. Furthermore, S can be modified toaccomodate for temperature effect under combustion conditions as [4]

where Ti = average inlet temperature of the input fluidT = average outlet o temperature of the gas

13. Jittrepit, P., The Burning Characteristics of a Sawdustjired Cyclone Combustor,M. Eng. thesis, King Mongkut's Institute of Technology North Bangkok, 1993.

14. Cheromisnoff, N. A., Encyclopedia of Fluid Mechanics, Vol. 4 (Solid and Gas/SolidFlow), Guff Publishing, Houston, 1986.

15. Beer, 1. M. and Chigier, N. A., Combustion Aerodynamics, Applied SciencePublishers Ltd., 1972.