Control chart pattern recognition using an integrated model based on binary-tree support vector machine

Abstract

Control charts are important tools in statistical process control for determining whether a process is in control. Furthermore, effective recognition of abnormal control chart pattern (CCP) can greatly narrow the set of possible assignable causes, significantly shortening the diagnostic process. In this article, an integrated model in which binary-tree support vector machine (BTSVM) is applied for abnormal CCP recognition is proposed. The integrated model consists of three stages. In the first stage, five statistical features and eight shape features are extracted. In the second stage, a binary-class support vector machine is used to detect abnormal CCPs. In the third stage, the BTSVM is applied to classify the detected abnormal CCPs. Additionally, the Fisher Ratio method is utilised at each node of the binary tree to design the architecture of the BTSVM. Simulation experimental results show that the proposed model is able to effectively identify the type of CCPs and that the classification accuracies of the second and the third stages are up to 100% and 98.5%, respectively. A series of contrast experiments prove that the classification accuracy of BTSVM outperforms that of one against one and one against rest for abnormal CCP classification.

Keywords:

• control chart pattern
• control chart pattern recognition
• support vector machine
• binary-tree support vector machine
• Fisher Ratio method

Acknowledgements

The authors gratefully acknowledge the financial support from the National Hi-Tech R&D Program (863) 2012AA041306.

We are grateful to the editors and all anonymous referees for helpful comments. Additionally, we thank our friends for their contributions to this paper.