Process Monitor for Autocorrelated Data by Integrating ICA and Neural Network

• Main Authors: Chen-Hong Yang, 楊正弘
• Other Authors: 邱志洲
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/trkqt2

碩士 === 國立臺北科技大學 === 商業自動化與管理研究所 === 95 === Process monitoring and control of a production line is often used in industry to maintain high-quality production and to facilitate high levels of efficiency in the process. However, current process control techniques, such as statistical process control (SPC) and engineering process control (EPC), may not effectively detect abnormalities, especially when autocorrelation is present in the process. This paper proposes an independent component analysis (ICA)-based image reconstruction scheme with a neural network approach to identify disturbances and recognize shifts in the correlated process parameters. The resulting image can effectively remove the textual pattern and preserve disturbances distinctly. We illustrate our approach using two most commonly encountered disturbances, the step-change disturbance and the linear disturbance, in a manufacturing process. For comparison, traditional Shewhart control charts and cumulative sum (CUSUM) charts were applied to evaluate the identification capability of the proposed approach. The experimental results reveal that the proposed method is effective and efficient for disturbance identification in correlated process parameters when disturbance is significant. Additionally, the identification rate made by proposed method is almost free from the influence of the data correlation.