Multivariate process monitoring and fault identification model using decision tree learning techniques

Abstract

A multivariate process monitoring and fault identification model using decision tree (DT) learning techniques is proposed. We Use one DT classifier for process monitoring and other p (p is the number of the variables) DT classifiers for fault identification. The Mahalanobis distance contours based method for selecting model training samples is proposed to decrease the number of training samples. Numerical experiments based on bivariate process show that the proposed model works well in different conditions considered. The results also show that the sample sizes have obvious effect on the performance of the model. The correlation coefficients have nearly no effects on the performance of the DT classifier for process monitoring, while have obvious effects on the performance of the DT classifiers for fault identification.