fundamental study of corrective abrasive machining...

26/05/2011

1

Fundamental Study of Fundamental Study of Corrective Abrasive Corrective Abrasive Machining TechnologyMachining Technology

Dr Xun ChenDr Xun Chen

Advanced Machining Technology GroupAdvanced Machining Technology Group

Centre for Precision TechnologiesCentre for Precision Technologies

School of Computing and EngineeringSchool of Computing and Engineering

CIRP UK Meeting 8th May 2009

Centre for Precision Technologies

Precision machining technology

Current Research ThemesCurrent Research Themes� Precision machining

� Abrasive machining

� Diamond turning

� Finite element analysis and molecular dynamics

� Intelligent process monitoring and control � Contact detection using acoustic emission

� Abrasive machining monitoring using AI techniques

� Knowledge support systems Database / Knowledge warehouse

Unique pioneer challenge in developing

next generation machining technology

Educating tomorrow’s professionals

Electrical & Electronic

Engineering

Mechanical & Manufacturing

Engineering

Computing & information

Technology

26/05/2011

2


Basic principle of precision corrective machining

Sa = 344.8nm Sa = 9.7nm

Pelvis

Acetabulum cup

Cup holder

Stem

Tribological analysis

AE pream-plifier 1& 2

Acc. coupler

Dynamo-amplifier

Thermo-amplifier

Data Acquisition card

SCB100

Spindle power

PAC signal card

On-line monitoring

3


4

Framework of a grinding process control

Control methods

• adaptive control,

• constraint control.

Monitoring & control; Surface integrity; Coolant delivery; Wheel dressing

Environmental factors:

•Machine

•Workpiece

•Coolant

Wheel preparation:

•Wheel type

•Dressing conditions

•dressing depth

•dressing lead

Grinding process:

•Force

•Power

•Vibration

•Temperature

•Acoustic emission

•Material removal

•Displacement

•Wheel wear

•Grinding burn

Control parameters:

•Grinding conditions

•grinding speed

•workspeed

•feedrate

•Coolant delivery

•Grinding cycle

Wheel conditioning:

•Wheel balance

•Wheel topography

Control strategies

Quality inspection:

•Surface roughness

•Surface wave

•Surface integrity

•crack

•hardness change

•residual stress

•phase changes

•Form error

•Size accuracy

Process parameters Quality parameters

Outputs

Three major measures of

grinding qualities

• accuracy of size and form,

• surface roughness,

• surface integrity.

Sensors:• Power and torque• Forces• Size gauging• Grinding force• Vibration monitoring• Temperature sensing• Eddy current sensing • Acoustic emission• Image sensing

26/05/2011

3


From the macroworld to the nanoworld (The Hitchhiker’s Guide to Nanotechnology, 2006)

MicroMicro--machiningmachining is considered to cover the production of minute components and features from a wide range of materials, generally in the size range of 200 microns to a few nanometres and may also be known as micro drilling, micro cutting, micro milling, micro grinding and micro etching.

5


Challenge of nanometre scale machiningChallenge of nanometre scale machining

Grain boundary indicates the limitation of the conventional turning operation limitation.What can be done?

Aluminum (Al) Sa = 0.742 nm Nickel (Ni) Sa = 0.924 nm

• Spindle RPM: 2000

• Finish Feedrate: 7.5 mm/min

• Finish Depth of Cut: 2 µm

• Coolant: Odorless Mineral Spirits

• Spindle RPM: 3000

• Finish Feedrate: 5 mm/min

• Finish Depth of Cut: 4 µm

• Coolant: Odorless Mineral Spirits

Sa < 1 nmSa < 1 nmForm error < 0.1 Form error < 0.1 µµmm

It was claimed that grinding has no minimum depth of cut.It was claimed that grinding has no minimum depth of cut.6

26/05/2011

4


7

grinding forces wheel wear

grinding wheel topography

surfaceintegrity

surfaceroughness

grindingtemperature

grindingvibration

size error

Outputs of grinding process

grinding power

grindingwheel

workpiecedressing

tooldressing

kinematicsgrinding

kinematics

Inputs of grinding process

chip geometry

Grinding process

single grain load

environment

grinding

deformation

form error

Basic Relationships of a Grinding Process and Simulation

Input : specifications of the wheel and the workpiece, dressing and grinding conditions.

Generating the cutting surface of the grain

Simulation based on a single grain in order k, j, i.

Distributing the grain centres

X (i = 1, 2, •••)

Y (j = 1, 2, •••)

Z (

k =

1, 2,

3)

Simulating the grinding action of the grain

All grains simulated?

no

Output results

yes

Accumulating the actions of the grains

workpiece surface before grinding

workpiece

Z

X

Simulating the wear of the grain


8

Fundamental Mechanism of Grinding

wheel

vw

vs

workpiece

a

a

chip formation ploughing sliding

grithm

l k

vs

Three stages of chip generation

AE signals would have certain features that present three stages of chip generation.

26/05/2011

5


9

Rigs for single grit grinding tests

Workpiece (2 workpieces 180° apart from each other)

Single grit holder

Force sensor

AE Sensor and protective housing

Material

v

Str

oke

v

Grit tip

Scratch

Stroke

v

Grit tip

Scratch

Stroke


10

AE signals of single grit scratch

0 0.1 0.2 0.3 0.4-0.2

-0.1

0

0.1

0.2Test 212 SG Scratches Channel 1

Time (S)

0 0.1 0.2 0.3 0.4-0.2

-0.1

0

0.1


Time (S)

Am

plit

ud

e (

V)

0 0.1 0.2 0.3 0.4-0.2

-0.1

0

0.1


Time (S)

Am

plit

ud

e

(V)

0 0.1 0.2 0.3 0.4-0.2

-0.1

0

0.1


Time (S)

0 0.2 0.4 0.6 0.8 1 1.2

x 10-3

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

Time (S)

Am

plit

ude (

V)

SG4 Experiment Test 212 Hit 2 (normalised)

26/05/2011

6


Single grit scratch testvv

Scratch

Stroke towards

page

Grit Tip

Steel plate

0 5 10

x 105

0

1

2

3

4 Cutting FFT

Norm

alis

ed A

mplit

ude

0 5 10

x 105

0

0.5

1

1.5 P loughing FFT

Frequency (Hz)0 5 10

x 105

0

0.1

0.2

0.3

0.4 Rubbing FFT

10 20 30-3

-2

-1

0

1Cutting P rofile

depth

of

cut

(um

)

0 20-1

-0.5

0

0.5

1 P loughing profile

c ross sec tion length (um )

0 20 40-0.1

-0.05

0

0.05

0.1 Rubbinging profile

401µm

11


12

Grinding mechanism classification using Neural Networks

0 10 20 30 40 50 600.5

1

1.5

2

2.5

3

3.5

NN Input Vectors Verification Test Set

1:

Cutt

ing,

2:

Plo

ugh

ing

and 3

: ru

bb

ing

NN Results for STFT SG4 AE Signals

x : classification

x : misclassification

x & x : Unseen Data

Network classification: 92%

Unseen Data classification: 83%

0 10 20 30 40 50 600.5

1

1.5

2

2.5

3

3.5

NN INput Vectors Test Set

1:

Cutt

ing,

2:

Plo

ugh

ing

and 3

: ru

bb

ing

NN Results for STFT SG4 AE Signals

x : classification

x : misclassification

x & x : Unseen Data

Network classification: 93%

Unseen Data classification: 87%

Wheel rotational speed = 4000 rpm, Feedrate = 4 m/s.

Workpiece materials: CMSX4

26/05/2011

7


13

Feature extraction of AE in grinding

Wheel rotational speed = 4000 rpm,

Feedrate = 4 m/s,

Workpiece materials: CMSX4.

0 10 20 30 40 50 600

0.5

1

1.5

2

2.5

3

3.5

NN Input Vectors (STFT of Hit 14)

1:

Cutt

ing,

2:

Plo

ughin

g a

nd 3

: R

ubbin

g

Hit 14 of 1um grinding cut with VIPER 60 Wheel

0 10 20 30 40 50 60 700

0.5

1

1.5

2

2.5

3

3.5

NN INput Vectors (STFT of Hit 15)

1:

Cutt

ing,

2:

Plo

ughin

g a

nd 3

: R

ubbin

g

Hit 15 of 1um cut with VIPER 60 Wheel

0 10 20 30 40 50 60 700.5

1

1.5

2

2.5

3

NN Input Vectors (STFT of Hit 20)

1: C

utt

ing,

2: P

loughin

g,

3: R

ubbin

g

Hit 20 of 0.1mm grinding cut with VIPER 60 Wheel

C% P% R%

1 µm cut 45 29 26

1 µm cut 45 47 8

0.1 mm cut 57 40 3


Laser irradiation imitating grinding thermal behaviour

Laser machine Lumonics:JK704 Nd:YAGWave length 1.06 ηm

Pulse energy 1.36J

Maximum peak power 2.5kW

Laser irradiation time 0.06 ms

Focal length 120 mmLight beam diameter 12 mm

Off-focal length 34~46 mm

14

26/05/2011

8


15

Features of thermal stress induced AE signals


Comparison of AE signals generated by laser and grinding

0 1 2 3 4 5 6 7 8 9 10

x 105

0

0.1

0.2

0.3

0.4

0.5

Frequency(Hz)

No

rma

lize

d E

nerg

y

Test No.: Laser23H8

Pulse Time: 0.6msPulse Energy: 3.4J

Off-focal: 15mmT

m=781°C

Thermocouple: K-type

867KHz

835KHz

820KHz

508KHz

461KHz

336KHz

0 1 2 3 4 5 6 7 8 9 10

x 105

0

0.1

0.2

0.3

0.4

0.5

0.6

Frequency(Hz)

Norm

alized E

nerg

y

Pulse Time: 0.6ms

Pulse Energy: 3.5J

Temperature: 108°COff-focal: 30mm

Thermocouple: K-type

Test No.: Laser37

F1=266KHz

F2=258KHz

F3=141KHz

F4=391KHz

F5=836KHz

0 1 2 3 4 5 6 7 8 9 10

x 105

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Frequency(Hz)

Norm

aliz

ed E

ne

rgy

Test No:Fd2-4

Wheel: XA60F13VRP

Material:CMSX4vs=40 m/s

vw

=0.5 m/min

ap=0.074

Direction: down

Coolant: NoT

m=784°C

258

141

391 766

0 1 2 3 4 5 6 7 8 9 10

x 105

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Frequency(Hz)

Norm

alis

ed e

nerg

y

Test No:F155010

Wheel:XA60F13VRP

Material:CMSX4vs=50 m/s

vw

=1 m/min

ap=1.3 mm

Direction: down

Coolant:48barT

m=611°C

391 141

406

156

438

Laser

Grinding

Low temperature High temperature16

26/05/2011

9


Grinding thermal behaviour monitoring

by using thermal AE signatures

The NN created using AE signatures from laser irradiation

Grinding burn identification using the NN from thermal AE signatures

17


18

Pattern recognition for

grinding defects

A Real Grinding Process (input)

Sensors (AE signal)

Impedance match & Amplifier

Data digitized & logging

Quantification, Normalization, & Denoise etc

Feature extraction: joint-time-frequency domain (wavelet packet)

Feature optimization

Non-burn

Burn

Signal condition

Signal processing

Pattern recognition

Fuzzy classifier

Minimum distance Fuzzy c-mean cluster

Feature extraction by wavelet packetand Joint Time-Frequency Analysis

-Feature calibration

-Transitive closure calculation

-Feature equalization

-Feature normalization

More than

512 features

Similarity

measure &clustering

Less than

10 features

=

mnmjm2m1

iniji2i1

2n2j22

n1j12

x ..., x ..., x x

x ..., x ..., x x

x ..., x x x

x x..., ,x x

n x mX

,,,

,,,

,...,,,

...,,,

)(

21

111

MM

MM

=

nnnjn2n1

iniji2i1

2n2j22

n1j12

K

r ..., r ..., r r

r ..., r ..., r r

r ..., r r r

r r..., ,r r

R

~,~,~,~

~,~,~,~

~,~...,,~,~

~...,,~~,~

21

111

MM

MM

Selection &

Optimisation

Training &

Classification

0

1

2

3

4

5

6

0 5 10 15 20 25 30 35

Burn status classification

Fuzzy recognition procedure

Fuzzy recognition block diagraph

• Success rate 92.3%

26/05/2011

10


Support

vector

machine

classificationMaximum margin

Optimal Hyperplane

Class I

Class II

Support vectors

1-5

me

asu

rem

en

ts a

cro

ss w

ork

pie

ce

2 4 6 8 10 12 14 16 18 20 22 24

0

2

4

6

Ra measurements for each cut (3 Trials separated by lines)

Test number

19


20

SVM classification

�Small amount of training data required �Small amount of classifying time

Benefits of SVM:

26/05/2011

11


SVM Classification

Machine: Makino A55 CNC machine centre. Workpiece: Inconel 718.

Wheel: VIPER wheel.

Depth of cut: 1 mm;

Grinding speed: 35 m/s;

Workspeed: 1000 mm/min.

Machine: Makino A55 CNC machine centre. Workpiece: Inconel 718.

Wheel: VIPER wheel.

Depth of cut: 1 mm;

Grinding speed: 55 m/s;

Workspeed: 1000 mm/min.

21


+

+*

Y X0.707 1.5

*

+*

Y Y

0.121

*X

X0.707Y + X – 1.5 XY (Y+0.121X)

+

+

X 1.5

*

*

YX

Y+0.121X+ X + 1.5XY(0.707Y)

*

Y0.70

7

+

Y

0.121

*

X

5A

7

A

6A

1

A5B

7B

9

B

8

B

6B

1B

2B

3B 4B

2

A3A 4A

X20

X8

X10

X14

X8

X10

X8

X10

X14

X14

X8

X13 X20

mydivide

plus

mydivide

mydivide

plus

plus

plusX20

mydivide

plus

mydivide

plus

plus

mydivide

-400 -350 -300 -250 -200 -150 -100 -50 0 50 100-100

-50

0

50

100

150

200

250

300

350

400

Distance

Dis

tance

Burn data & cluster centre

Chatter data & cluster centre

Normal grinding data & cluster centre

-400 -350 -300 -250 -200 -150 -100 -50 0 50 100-150

-100

-50

0

50

100

150

Distance

Dis

tance

Burn data & cluster centre

Chatter data & cluster centre

Normal grinding data & cluster centre

MultiMulti--classification of normal grinding, grinding chatter and burn using the GPclassification of normal grinding, grinding chatter and burn using the GP

22

26/05/2011

12


Classification of grinding anomalies using genetic programming

No. GP fitness function Data set Function Nodes Test Score Accuracy %

1 sum diff fitness ICA Chatter and burn +,' -, '/, '* 32/40 80

2 sum diff fitness *reduction: burn and no burn +,' -, '/, '* 36/40 90

3 sum diff fitness *reduction: burn and no burn =<, '=>, if 36/40 90

4 sum diff fitness *reduction: burn and chatter +,' -, '/, '* 36/40 90

5 sum diff fitness *reduction: burn and chatter =<, '=>, if 38/40 95

6 classes overlap ICA Burn and no burn +,' -, '/, '* 33/40 82.5

7 classes overlap ICA chatter and no chatter +,' -, '/, '* 32/40 80

8 classes overlap ICA chatter and no chatter +,' -, '/, '* 40/40 100

9 sum diff fitness ICA chatter and no chatter +,' -, '/, '* 36/40 90

10 classes overlap ICA Burn and no burn +,' -, '/, '* 40/40 100

11 classes overlap *reduction: burn and no burn +,' -, '/, '* 40/40 100

12 classes overlap *reduction: chatter & no chatter +,' -, '/, '* 40/40 100

*reduction: is based on the statistical window n-dimensional reduction technique

23


Micro and nano scale corrective abrasive machining

Process modelling

Two-body and three-body errosions

Grinding Lapping Polishing

Models Suitable conditions On machine probing

GrolishingGrolishing

Process monitoring24

26/05/2011

13


Grolishing Process Tools


Models of material removal

rp pvCdt

dz=Preston material removal rate model

Archard material removal volume model

H

sFKV n

w =

v

pHf

pfVS

)( +=

X. Wu, Y. Kita, K. Ikoku (2007) 26

26/05/2011

14


Grolishing material removalα

Hertzian contact theory

Real result

Modelling of Zeeko grolishing toolsSimulation of Deformations and Stresses

27


Taguchi test resultsLevels

Representation Control Factors 1 2

A Head speed (rev/min) 1000 2000

B Tool Angle (o) 0o 10o

C Grit Size (µm) 7 25

Test

No.

Control Factors Mean

Sa(av)

S/N

ratioA B C

1 A1 B1 C1 0.350 9.12

2 A1 B2 C2 0.470 6.45

3 A2 B1 C2 0.350 9.12

4 A2 B2 C1 0.221 13.12

The best result is A2, B2, C1

Further improvement could be achieved by using different pad.

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

0.400

0.450

0.500

0 1 2 3 4 5

Su

rfa

ce R

ou

gh

ness

µm

No. of Inconel 718 test workpiece

Mean Sa(av)

Mean Ra(av)

28

26/05/2011

15


Error compensation On-Machine Metrology enhanced

Machine partwith DIFFSYS® 3DX,Y,Z program

Gather data from X,Y,Z by point-to-point analysis or by scanning Output data file

to DIFFSYS®Auto-correct to desired tool path

Re-machine 3D part to desired profile

Output data file

to 3rd party softwareQuality inspection

2D and 3D Metrology

Utilizing UltraComp’s™ LVDT, 2D and 3D data

files of the part profile can be gathered including

the machine axes

DIFFSYS® MC3 performs data import, graphically

displays and auto-corrects the 3-dimensional part

profiles from the LVDT measurement data to the

desired part profile.

29


AE detection

of contact

Voltage(mV) vs Time(us) <2>

0 2000000 4000000 6000000 8000000 10000000 12000000

-180

-160

-140

-120

-100

-80

-60

-40

-20

0

20

40

60

80

100

120

140

160

180

Before contact After contact

Contact AE signals

30

26/05/2011

16


Manufacturing Knowledge Warehouse Development

Framework of the warehouse

• date Input module

• database module

• problem solving module

• knowledge discovery module

• knowledge warehouse

• knowledge analysis module.

Database

Decision support Questions & answers Statistical Analysis Query/Reporting

Knowledge Acquisition

Knowledge Analysis

Cases Collection

Knowledge Warehouse

(Storage)

Learning Knowledge

Discovery

External sources

Grinding Cases

Cases collection interface

Knowledge engineers

CoPs

Knowledge

Acquisition

Interface

Problem-solving

Knowledge Discovery

Contents of communication in CoP

• Questions and answers.

• Passing documents or web links.

• Calling for events or conference.

• Sending news box.

• Others.

31


Opportunities of further developmentOpportunities of further development

� Ultra precision corrective machining at nanometre scale (micro material removal).

� Intelligent machining monitoring and optimisation.

� Precision machining knowledge support system.

Thanks for listeningThanks for listening

32

fundamental study of corrective abrasive machining...

Documents