| Summary: | 碩士 === 輔仁大學 === 統計資訊學系應用統計碩士班 === 100 === For industrial processes, the recognition of the control chart patterns (CCPs) have become one of the indispensable monitoring technologies. Most studies assumed that the monitoring process observed value is the single types of unusual patterns. However, in practice, the observed process may be concurrent patterns where two patterns may exist together and happened to different process noise distribution, such as Normal, Gamma, and Uniform. In order to recognize the concurrent CCPs, this study integrates the ICA and SVM to construct an effective model for recognizing concurrent CCPs. In the proposed model, the ICA is applied to the concurrent patterns for generating independent components (ICs). Then the ICs used to represent the concurrent patterns are identified. The ICs are served as the input variables of the SVM model to recognize the concurrent CCPs. Simulations results showed that the proposed ICA-SVM is able to effectively recognize concurrent CCPs with different process noises.
|
|---|
