Adaptive SVDD for Multivariate Process monitoring

• Main Authors: Yu-Zi Wang, 王郁誌
• Other Authors: Chun-Chin Hsu
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/89742060341485569761

碩士 === 朝陽科技大學 === 工業工程與管理系碩士班 === 100 === Suppor Vector Data Description（SVDD）is originally developed as a one-class classifier. The objective of SVDD is to find a optimized hypersphere that can tightly envelop the training data. Recently, SVDD is widely used for several fields, such as image feature classification, mechanical fault diagnosis, linguistic recognition and so forth. Recently, SVDD is used for multivariate process monitoring which was firstly developed by Sun and Tsung（2003）and they named it as K-chart. The K-chart takes advantage from no limitation to data distribution. Eventhough, the K-chart neglects the characteristic of process autocorrelation, that is, the time trend of the data. As mentioned above, this study aims to develop an Adaptive SVDD（ASVDD）for improving traditional K-chart in order to monitor autocorrelated multivariate process.The proposed method will be implemented via a chemical case and etch process case. In which, conventional multivariate process monitoring method, such as Principal Component Analysis（PCA）and Independent Component Analysis（ICA） are also implemented to show up the efficiency of propsed ASVDD.