Root Cause Detection with an Ensemble Machine Learning Approach in the Multivariate Manufacturing Process

Quality control in multivariate manufacturing processes should be applied with multi variate control charts. Although this method is sufficient, it doesn’t include the causes of uncontrolled situations. It only shows samples that are out of control. A variety of methods are required to determine the root cause(s) of the uncontrolled situations. In this study, a classification model, based on the ensemble approach of machine learning classification algorithms, is proposed for determining the root cause(s). Algorithms are compared according to predictive accuracy, kappa value and root square mean error rates as performance criteria. Results show that Neural Network ensemble techniques are more efficient and successful than individual Neural Network learning algorithms.

[1]  Tai-Yue Wang,et al.  Artificial neural networks to classify mean shifts from multivariate χ2 chart signals , 2004, Comput. Ind. Eng..

[2]  Chuen-Sheng Cheng,et al.  Identifying the source of variance shifts in the multivariate process using neural networks and support vector machines , 2008, Expert Syst. Appl..

[3]  Jian Pei,et al.  Data Mining: Concepts and Techniques, 3rd edition , 2006 .

[4]  Ruey-Shiang Guh,et al.  An effective application of decision tree learning for on-line detection of mean shifts in multivariate control charts , 2008, Comput. Ind. Eng..

[5]  Mojtaba Salehi,et al.  On-line analysis of out-of-control signals in multivariate manufacturing processes using a hybrid learning-based model , 2011, Neurocomputing.

[6]  H. Hotelling.  Multivariate Quality Control-illustrated by the air testing of sample bombsights , 1947 .

[7]  Jacob Cohen.  A Coefficient of Agreement for Nominal Scales , 1960 .

[8]  Hooshang Jazayeri-Rad,et al.  Comparing the Fault Diagnosis Performances of Single Neural Networks and Two Ensemble Neural Networks Based on the Boosting Methods , 2014 .

[9]  Nola D. Tracy,et al.  Decomposition of T2 for Multivariate Control Chart Interpretation , 1995 .

[10]  Jun Lv,et al.  On-line classifying process mean shifts in multivariate control charts based on multiclass support vector machines , 2012 .

[11]  Francisco Aparisi,et al.  Interpreting the Out-of-Control Signals of Multivariate Control Charts Employing Neural Networks , 2010 .

[12]  Abdessamad Kobi,et al.  Bayesian networks and mutual information for fault diagnosis of industrial systems , 2006 .

[13]  Yoav Freund,et al.  A Short Introduction to Boosting , 1999 .

[14]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[15]  Jin Wang,et al.  Fault Detection Using the k-Nearest Neighbor Rule for Semiconductor Manufacturing Processes , 2007, IEEE Transactions on Semiconductor Manufacturing.

[16]  Ibrahim Masood,et al.  Pattern recognition for bivariate process mean shifts using feature-based artificial neural network , 2012, The International Journal of Advanced Manufacturing Technology.

[17]  John C. Young,et al.  A Practical Approach for Interpreting Multivariate T2 Control Chart Signals , 1997 .

[18]  Pilar Rodriguez-Loaiza,et al.  Application of the Multivariate T2 Control Chart and the Mason–Tracy–Young Decomposition Procedure to the Study of the Consistency of Impurity Profiles of Drug Substances , 2003 .

[19]  Min Zhang,et al.  Multivariate process monitoring and fault identification using multiple decision tree classifiers , 2013 .

[20]  Seyed Taghi Akhavan Niaki,et al.  Fault Diagnosis in Multivariate Control Charts Using Artificial Neural Networks , 2005 .