Integrating Principal Component Analysis and Support Vector Machine for MEWMA Control Chart Nonrandom Pattern Recognition

• Main Authors: Ya-Hui Liu, 劉雅慧
• Other Authors: Chau-Chen Torng
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/34620609540735818319

碩士 === 國立雲林科技大學 === 工業工程與管理系 === 102 === Control chart of statistical process control techniques is a significant and simple tool. Except for any of the point falls outside the control limits is out-of-control in control chart, when the data showed nonrandom pattern which means that the process may be caused abnormal by certain factors is out-of-control. If we can correctly identify nonrandom pattern of control charts, then we can narrow diagnostic range and shorten the time to exclude abnormal. This study proposes the integration of principal component analysis and support vector machine approach to identify nonrandom pattern for MEWMA control chart. Nonrandom pattern is based on the definition of the Western Electric Handbook, It is divided: trend pattern, shift pattern, cyclic pattern. In this study, the use of principal component analysis on the original data derived new data, then calculate statistical values of MEWMA control chart, this statistical values as input data for support vector machine to construct a model which identifies nonrandom patterns of identification model. The study also proposed adding statistical eigenvalues to improve the performance of support vector machine identification. In correlation of 0 and 0.5, rate of correct classification slightly increased after adding statistical eigenvalues. Rate of correct classification with non-random pattern parameter are proportional. Non-random pattern parameter greater the pattern more obvious, is also relatively easier to be identified.