Mean shifts identification model in bivariate process based on LS-SVM pattern recognizer

作者：Zhi Qiang Cheng^*; Yi Zhong Ma; Jing Bu

来源：International Journal of Digital Content Technology and Its Applications, 2010, 4(3).

DOI：10.4156/jdcta.vol4.issue3.16

• Abnormal patterns
• Bivariate
• Classification accuracy
• Excellent performance
• Hotelling's T
• Identification model
• Least squares support vector machines
• Mean shift
• Multivariate control charts
• Multivariate statistical process control
• Pattern recognizers
• Process mean
• Process mean shifts
• Quality diagnosis
• Shewhart
• Small Sample Size

This study develops a least squares support vector machines (LS-SVM) based model for bivariate process to diagnose abnormal patterns of process mean vector, and to help identify abnormal variable(s) when Shewhart-type multivariate control charts based on Hotelling's T 2 are used. On the basis of studying and defining the normal/abnormal patterns of the bivariate process mean shifts, a LS-SVM pattern recognizer is constructed in this model to identify the abnormal variable(s). The model in this study can be a strong supplement of the Shewhart-type multivariate control charts. Furthermore, the LS-SVM techniques introduced in this research can meet the requirements of process abnormalities diagnosis and causes identification under the condition of small sample size. An industrial case application of the proposed model is provided. The performance of the proposed model was evaluated by computing its classification accuracy of the LS-SVM pattern recognizer. Results from simulation case studies indicate that the proposed model is a successful method in identifying the abnormal variable(s) of process mean shifts. The results demonstrate that the proposed method provides an excellent performance of abnormal pattern recognition. Although the proposed model used for identifying the abnormal variable(s) of bivariate process mean shifts is a particular application, the proposed model and methodology here can be potentially applied to multivariate SPC in genera.

• 出版日期2010
• 单位南京理工大学