應用獨立成分分析與統計製程管制圖於多變量製程之監控

Multivariate Process Monitoring with Independent Component Analysis and Statistical Process Control Chart

楊欣娟(Hsin-Chuan Yang)

指導教授 ： 鄭春生

元智大學/工程學院/工業工程與管理學系/碩士(2008年)

摘要

多變量製程管制中，當微量的變異且具有相關性的品質特性在彼此相互作用下，會使製程出現穩定的假象或製程干擾現象被誤判，導致無法有效的偵測變異來源，進而影響不良率的提升。因此，如果能將品質特性之間的相關性去除後，再利用統計製程管制圖進行監控，將可避免因為相關性所造成的誤判。管制圖中常見之非隨機樣式包括：偏移樣式、趨勢樣式與週期性等樣式。快速偵測出製程是否具有非隨機樣式，對於管制製程有所助益。

本研究之主要目的是使用獨立成分分析中分離的技術，建構一個能偵測出多變量管制圖中之非隨機樣式的監控系統，作為實施矯正措施及改善產品品質的重要依據。

關鍵字

多變量製程管制 ； 統計製程管制圖 ； 非隨機樣式 ； 獨立成分分析

並列摘要

Quality characteristics of a multivariate process are usually correlative with each other. It is hard to detect the non-random patterns of the multivariate process, and furthermore to recognize the source of the variation by Shewhart control chart. If the correlation of quality characteristics can be eliminated, the effect of statistical process control will increase greatly. Some non-random patterns in manufacturing processes are investigated in this research, including shift, trend and cycle type. It is beneficial for process control if we can detect those non-random patterns as soon as possible.

The main purpose of this research is using the independent component analysis skill to establish a monitor system which can recognize the non-random patterns of multivariate process control. The control procedure provides a more effective statistical process control system in monitoring the non-random patterns of production processes.

並列關鍵字

multivariate process control ； statistical process control chart ； non-random patterns ； independent component analysis

參考文獻

1. Bakshi, B. R., “Multiscale PCA with application to multivariate statistical process monitoring,” AICHE Journal, 44(7), 1596-1610 (1998).

2. Crosier, R. B., “Multivariate Generalizations of Cumulative Sum Quality Control Schemes,” Technometrics, 30, 291-303 (1988).

3. Dong, D., McAvoy, T.J., “Nonlinear principal component analysis based on principal curves and neural networks,” Comp. Chem. Eng., 20(1), 65-78 (1996).

4. Duncan, A. J., Quality Control and Industrial Statistics, 5, Irwin Book Company, Illinois (1986).

5. Girolami, M., Self-Organising Neural Networks: Independent Component Analysis and Blind Source Separation, Springer-Verlag, London (1991).