Pattern recognition for bivariate process mean shifts using feature based artificial neural network

Int J Adv Manuf Techno1 DO1 10.1007/s00170-0124399-2

Pattern recognition for bivariate process mean shifts

using feature-based artificial neural network

Ibrahim Masood

.

Adnan Hassan

Rwei~rd. I July 2U10 Acuepr~d: I 6 July 2012 1: Syrmp;r-\'crlag Lunllon l.irniicd 2012

Abstract

In multivariate quality control, the rutificial neu-

ral networks (ANN)-based pattem recognition schemes gen-

erally performed better for monitoring bivariate process

mean

shifts

and provided more efficient information for

diagnosing the source variable(s) compared to the traditional

multivariate statistical process control charring. However,

these schemes revealed disadvantages in

term

of reference

bivariate pattern in identifying the joint effect and exeess

false alarms in identfyhg stable process condition. In this

study, feature-based ANN scheme was investigated for rec-

ognizing bivariate correlated patterns. Featurebasal input

representation was utilized into an ANN

training

and te*

towards strengthening discrimination capability between bi-

variate normal and bivariate mean shift patterns. Besides

indicating an effective diagnosis capability in dealing with

low correlation bivariate pattern, the proposed scheme

promotes a smaller network size and better monitoring ca-

pability as compared to the raw data-based ANN scheme.

Keywords Artificial neural networks

.

Bivariate emelated

pattem

.

Process monitoring and diagnosis

.

Statistical

features

.

Pattern recognition

L Masood (B)

Faculty of Mechanical and Maouficturing Engineering, Universiti Tun Hussein Onn

Mdaysia,

86400 Pait Rajq Baru Pahat, Johor,

Malaysia

e-mail: ibrahim@uthmedumy

URL.

hnp://im.uthin.edu.my

A. Hassan

Faculty

of Mechanical Engineering,

Universiti

Teknologi Malaysia, 81310UTM Sbd&

Johor,

Malaysia

e-mail:

[email protected]

WRC.

http:l/w~\w.utm.rny

Abbreviations

ANN

Aaificial neural network

ARL

Average run length

ART

Adaptive resonance theory

BPR

Bivariate pattern recognition

CCPR

Control

chart

pattern recognition

CUSUM

Cumulative sum

EWMA

Exponentially weighted moving average

FAR

False alarm rate

i.i.d.

Identically and independently distributed

LEEWMA

Last value of exponentially weighted moving

average

LVQ

Learning vector quantization

MAT

(Mean)

x

(autocomelation)

1MGUSUM Multivariate cumulative sum

MEWMA

Multivariate exponentially weighted moving

average

MLP

Multilayer perceptrons

M S V

(Mean)

x

(mean square value)

MQC

Multivariate quality control

MRWA

Multi-resolution wavelet analysis

MSD

(Mean)

x

(standard deviation)

MSE

Mean square e m

MSPC

Multivariate statistical proeess control

YLA

k c i p l e component anaiysis

RA

Recognition a c c m y percentage

RBF

Wal basis function

SOM

Kohonen self-organizing mapping

SPC

Statistical process control

traingdx

Gradient decent

with momentum and adaptive

leaming rate

1 Introduction

It is well known that variation in manufacluring processes

has

become a major source of poor quality. W~ear and tear,

Int I Adv Manuf Teehnal

For t h e future work, we plan

to

conduct a more extensive investigation towards

heightening

the

performance of the feature-based

ANN

for identifying smaller

mean shifts.

In- vestigation also wili be extended

for

other

causable

p a t t e r n

such as trend and cyclic.

Acknowledgments The authors would like to thank Universiti Tun Hussein OnnMalaysia, Universiti Tehologi Malaysia, and M b b t r y of Higher Education of a y s i a who are sponsoring this work The constructive suggations h m the Editor in Chief of UAMT and the anonymous reviewers have greatly aided us to jmprove the contmts as well as the presentation of this paper

References

1. Oakland JS (1996) Statistical process control. Butteworth- Heincmann, Oxford

2. Shewhad WA (1931) The eecanomjc wntml of quality m f a o tured products. Van Nostrand, New York

3. Nelson LS (1989) Standardization of Shewhart control charts. J Qual Techno1 21(4):287-289

4. Cmsier RB (1986) A new two-sided cumulative sum quality control schemes. Technometrics 28[3):187-194

5. Lucas JM, Saccucci MS (1990) Exponentially weighted moving averaee control schemes: Aurnoeaies and enhancements. Techno- . rnetrii 32(1):1-12

6. Hotelling HH (1947) Multinkte aualiW control. In: Eisenhart C,

Hastay kW, W ~ U ~ ~ ' W A (eds) ~'hni~nes of statistical analysis McGraw-Hill New York

7. Cmsier R3 (19881 Multivariate eeneralizations of cumulative sum

-

quality coniol schemes. Technornehics 30(3]:291-303

8. P~gnafieUo U, R u m GC (1990) Conpansom of multivariate CIJSUM ch&s. J

Qual

Techno1 22(3):173-186

9. Lowry CA, Wwdall WH, Champ CW, Rigdon SE 11992) M a - variate exponentially weightedrnoving average control chart. Technomdcs 34(1):46-53

10. Prabhu SS, Runga GC (1997) Designing a multivaziate EWU4 control chart I Qual Teehnal 29(1):8-I5

11. Alt FB (1985) Multivariate quality control. In: Johnson NL, Kotz S (eds) Encyclopedia of statistical sciences 6. Wiley,

New York

12. Jwkson JE (1991) A user's guide to principle component. Wiley. New Jersey

13. F u c h C, Benjamini Y (1994) Multivariate profile c h w for sta- tistical oroms controL Technamehim 36i231182-195

14. ~ a s a n - k , Tracy ND, Young JC (1995) D&mpositian of

15

for mulhmate eontrol chart mntemrefauon. J W l - m P ..

.

108

15. Mason RL, Tracy ND, Young JC (1996) Monitoring a multivariate SteD DrOCeSS.

. .

J Oual Te~6nol2811):39-50

16 Srpnil\cda A. N.xhl..r JA (199-J A ..in~ulauon apprdnch lo multi- \dndtc~sualit) c.>nrrol. ('urnpat lnd Fne 33(1- 2):l 13-1 lo 17. Lowry CA Montgomery DC ( 1 9 9 5 ) ~ review of multivariate

eontror charts. llE Tmos 27[6):81)(31810

18. KO& T. MacGreeor SE (1996) Multivariate .

.

SPC

methods for

proce,.; 3nJ p r d u ; ? ,non!lonng. J C)L.IJ 1'cchnl,l 2P(1):J1I1J426 IQ. Mxwn 1U. Tnry ND, Yourl* Ji t1YL)7) A ~racti;~l appruach lur

inte'preiing mliivariate

f

control chart &n&. J Q&I Techno1 29(4):39W06

20. Bersimis S, Psarakis S. Panaretos J (2007) Multivariate statistical process control charts an o-iew. ~ual'&liab EngInt23:517- 543

21. W6hler C, Aolauf JK (1999) A time delay neural network algo- rithm for &mating image-pattern shape and motion. - - Image Vis Comput 17-281-294

22. Singh V, Mishra R (2006) Development a machine vision system for spangle classification using image processing and artificial neural nehvmk Surf Coat Technol201:2813-2817

23. Liang SF, Lu SM, Chang JY, Lin CT (2008) A novel two-stage imoulse noise removal echniaue based on n e d networks aid

fu& decision. IEEE T m F& Syst 16(4):863-873

24. Chdstodoulou CI, Pattichis CS (1999) UnsuDenised ~ a t t e m rec- ognition for t h e classificationof &G Signals. ~ E E Trans Bimned Eng 46(2):16%178

25. Giiven A, Kara S (20061 Diagnosis of the macular diseases fiom pattern electroretinogmphy signals using artificial neural netwo~ks. Expal Svst ADD] 30:361-366

26. ~ l t u n s ~ S, Teiatarz, Emrogul 0, Aydur E (2009) Anew approach to uninary system dynamics problems: evaluafion and classification of uroflo&neter iienals usine artificial neural networks. Exoert - -

Syst Appl36:4891>895

27. Hr 0, Yumusak N, Temvrtas F (2010) Chest diseases d~agnosis .

.

using artificial neural networks. Expert Syst Appl 37:764b

7 K C C _, w--

28. Chen CW (2009) Modeling and control for nonlinear structural systems via a M*T-based approach. Expert Syst Appl 36:4765- 4772

29. Cheo CW, Yeh K, Liu KFR (2009) Adaptive iiizzy sliding mode contml for seismically exited bridges mth lead rubber beariDg isolation. Int J Unce~tain Fuzziness Knowl-Based Syst 17(5):705- 777

30. Chen CY. Lin

JW,

Lee WL Chen CW 12010) F ~ w control for an wcdnic ruucrwi.. 3 cfis: S R I . ~ ~ in rirrl:-dsIay 1'1 I' ~ I I C ~ . J Vib C.unllol lh(lJ.147 Ifll

31. Amin A (2000) Rewenition of minted Arabic text based on elobal

.

,

-

feahlres and decision tree learning techniques. Pattern Recogu 331309-1323

32. B h a U a c M a K, Sarma KK (2009) ANN-based innovative seg- mentation method for handwritten text in assam-. Int J Cornput Sci Issues 5 9 1 6

33. Marcu T, Seliger BK, Stiicher R (2008) Design of fault detection for ahydmnlic looper using dynamic neural networks. Conhol Eng Pract 16:192-213

34. Demetgul M, T m e l IM, Taskin S (2009) Fault diagnosis of oneumatic svstems with ardficial neural network alearithms. Ex-

k n

Sknt h i 1 36:11)512 -1U51Y

35. Ji.4 LY, M a JN', Nang FJ, l.iu N (ZUIG, ('h~r.~;tr.riu~;, t'.,r.x?slicc of hydraulic valve based on grey ro~~elation and AtWS

E X ~ G

Syst Appl37:125&1255

36. H a v k S (1999) Neural networks: a commehemive foundation PAtice H& N ~ W Jersey

37. Hwarng

HB,

Hubele NF (1993) Back-propagation pattern reeog-

- E k i + h f 2 G b ~ & ~ c m e M ~

-Comput Ind Eng 24:219-235

38. V e l w T (1993) Back propagation aaii~cial neuralnetworks for the analysis of quality control c h a t s Comput Ind Eng 25(L-

4):397-400

39. Pham DT, Oztemel E (1993) Control chart pattem rewgnition using combinations of muhilayer pemptroos and leaning vector qwanthtion -neural nelworks. Pros Inst Mecb Eng 207: 11 3-1 18 40. Pham DT, Oztemel E (1994) Control chart pattem r-gnition

using learning vector quantization networks. In1 J Prod Res 32721-729

41. Zomassafine F. Tanneck JDT 11998) A review of n&

.

, oetwmks for statistical pmcess control. J Intell Manuf 9:209-224

shier Goin n.ubivrise P e h u l \ig.nuls. Co~nput Inti Fng 47:195--205

24. Niaki SIX, A h b h 3 ( 2 ~ 0 5 ) I:3ult disgfiusli in rnn~l~lvnrjarr. c711- -01 charts ushe mircial neural nehvorks. Oual Reliab Ene Int

.

-

2 1 :82&840

45. Cbeng CS, Cheng HP (2008) Identifying the saurce ofvariance shifts in the multivariate process using neural networks and sup- port vector machines. Expert Syst Appl35:19&206

46. Yu JB. Xi LF. Zhou XI (2009) ldentifvine

.

, >

-

source(^) , , of out-of- control signals in multivariate m a n u b r i n g prmesses using se- lective new1 nchvork ensemble. Ene Avvl Artif Intell 22141-152

iate pracem. ~ o r n ~ u t Ind Eng 4 4 3 8 5 4 8

48. G u h R S (2007) On-line identification and quantification of mean shifm in bivariate processes using a NN-bmed approach. Qua1 Reliab Eng Int 23:367-385

49. Yu W, Xi LF (2W9) A neural network ensemble-based model for on-line monitoring and diagnosis of out-ofantto1 signals in mul- tivariate manufacturing processes. Expert Syst Appl36:909-921 50. EI-Midany TT, El-Baz MA, Abd-Elwahed US (2010) A nrmsed

_{. .}

f i i w w u r l ~ L I cdnnc l rlj.ln pmcni recagn~lhrr in mullivlrifi plo-

cci, winp ~niricljl ~new.il~~r.tworks. I.xpm S)a Appl37:1035 ,012 51 Park S-J. ~ ~ L e \I-$. Shir. S-Y. Chr, K-H. Lint J- 1. Cho U-S. Y-H Jci M-K, park G H (2005) ~un-&-mn overlay eo&l of

stqpers in semiconductor m m i a d g systems based on hitor- ical data analysis and neural network modeling. IEEE Tram Semi- c o n Manuf 18(4):60%13

52. hlnnm RL, Chou 1%. Yani: JC {2lIJI) .Applying 1Llr.lling 7' slatirrs w hitch prorrssc,. J Qua1 T<ihnr,l 33(4):40M7:, j3. Parrd h.l(iL, L I M J ~ !'K (20~3-7U01, Ap~lkdtion of nul1iv~vi;;le ,

.,

control chai and ~a&n-Tmcy dmmpasition procedure to the study of the consistency of impurity profiles of drug substances. Oualitv Eneineerine 16(1):127-142

54 &gbk&lbIeathth~ b b 2 ) A real-time information system far mulfivariatte staiisticalprffiess contml. lot JProdEeon 75161-172 55. Miletic I, Quinn S, ~ i d d c M, Vaculik V, Champagne M (2004)

An industrial perspedive on implementation on-line applications o f multivariate statistics J Pmeess Control 14:821-836

5 6 Lehman RS (19771 Computer simulation and m o d e l i i an Intro- duction. Lawrence Erlbaum, London

57. B m e U H, Eriksson J (2000) Feature reduction for classification o f multidimensional data Pattern Recog 33(12):1741-1748 58. Klosgen W, Zytkow

J

M

(2002) Handbook of data mining and

knowledge discovery. Oxford University Press, London

In1 i Adv Manuf Techno1

59. Pham DT, Wani MA (1997) F a r e - h e d wm01 c Mpartem rewpition IatJ Prod Res 35(7):1875-1890

60. Gauri SK Chakrabor@ S (2006) Feature-based recognition of control chart patterns. Comput Ind Eng 51:72&742

61. Gaui SK, Chakrabmty S (21248) b r o v e d recopnition oicontrol chchart paltern using artificial neuafnetworks. &t J Rdv kf Techno1 36:1191-1201

62. Hassan A, Nabi Baksh MS, Shaharoun MA, Jamaludin H (2003) Improved SPC chart pattern recognition using sfatistical featllres. Int JPrcd Res 41(7):1587-1643

63. Guh RS (2010) Simultaneous pmcess mean and variance monitor- ing using artificial neural networks. Comput Ind Eng 58:739-753 64. A1-hsaf Y (2004) Recognition of control chart pan- using

multi-resolution wavelefs analysis and neural corks. Comput Ind Eng 47:17-29

65. Assaleh K, Al-AssafY (2005) Features extraction and analysis for classifying musable pattern in control charts. Comput Ind Eng 49:168-181

66. Chen Z, Lu S, Lam S (2007) A hybn'd system for SPC con-nt pattem recognition. Adv Eng Inform 21:30>310

67. Wang CH, Kuo W, Qi H (2W7) An integrated approach forproms monitoring ming wavelet analysis and competitive neural netwn~k I n t i Prod Res 45:227-244

68. Montgomery DC (2005) Introduction to statistical quality control. Wiley, New Jersey

69. Demuth H, Beale M (1998) Neural network twlbox, user's guide. The Mathworks, Natick

70. Bishop C (1995) Neural nehvork for pattern recognition. Oxford University Press, New York

71. Cheng CS (1995) A multi-layer neural network model for detect- ing changes in the process meaa Comput Ind Eng 2 8 5 6 1 72. Cheng CS (1997) A mural network appraacb i m the analysis of

contml chart pattern. Int Pmd Res 35(3):667497

73. Guh RS, Hsieh YC (1999) A neural network based model for abnormal pattern recognition af control charts. Comput Ind Eng 3697-108

74. Guh RS, Tannock JDT, O'Brien C (1999) Intellispc: a hybrid intelligence tool for on-line economical statistical process mntml. Expert Syst Appl17: 19S2 12

75. Guh RS, Z o h t i n e 8, T w m k JDT, O'Brien C (1999) On-line control cbm pattern detection and discriminatio-a neural net- work approach. Artif Intel1 Eng 13:41%425