ISSN 1753-9439 Paper 5 Vol 2 No 1 pages 122 – 138
Statistical Process Control using Support Vector Machines: A Case Study
Control limits in traditional Multivariate Quality Control Charts (MQCC), such as Hotelling's T2 control chart, Multivariate Cumulative Sum (MCUSUM) and Multivariate Exponentially-Weighted Moving Average (MEWMA) control charts are based on multivariate normal assumption. This assumption is not usually satisfied, especially in real life applications. The purpose of a control chart based on support vector machines (SVM) can alleviate the dependence of multivariate control charts to normality assumption, and provides an efficient Statistical Process Control (SPC) tool, with a high ability to deal with real data problem. Consequently, Sun and Tsung (2003) proposed a "kernel-distance-based control chart", also known as the "k-chart". Thus, the k-chart, drawn from Statistical Learning Theory, could be seen as a new approach in SPC. This paper focuses on the application of the k-chart to an industrial process. An assessment of the k-chart is analyzed by comparing it to T2 control chart. This comparison is based on the Average Run Length (ARL) criterion.
Keywords: k-chart, industrial process, Hotelling’s T, ARL.