建構多重輸入多重輸出之製程管制系統－整合應用SPC/EPC模式與柔性演算法

本研究主要目的在於針對系統中多重輸入多重輸出 (Multiple-Input Multiple-Output；MIMO) 的製程型態，以整合統計製程管制 (Statistical Process Control；SPC) 與工程製程管制 (Engineering Process Control；EPC) 的概念為基礎，並使用在柔性演算法 (Soft Computing；SC)中的類神經網路技術建立製程模型，再以基因演算法與類神經網路技術之整合模式進行最佳參數之搜尋；藉此模式化複雜的製程輸入與輸出關係以獲得較佳的產出結果，期以提昇產品品質。故本研究建構一預應與參數控制模式之多重輸入多重輸出製程管制系統，其包含MIMO製程子系統、量測子系統、偏移偵測與分析子系統、類神經網路製程輸出預測子系統、基因演算法參數設計最佳化搜尋子系統、參數調整子系統等六部分；並依此模式詳細探討其流程步驟相互之關係，達成多重輸入多重輸出的製程型態在偏移干擾之下能正確地預測與診斷，除可有效控制製程的輸入與輸出外，且能達到製程最佳化之目的。

關鍵字

工程製程管制；多重輸入多重輸出；統計製程管制；柔性演算法

並列摘要

The main objective of this study aims at Multiple-Input Multiple-Output (MIMO) process mode. Based on the integrated concepts of SPC and EPC, Soft Computing (SC) technique and statistical analysis technique are combined to modularize the relationship between process output and process input, so optimal yield can be derived and process quality can be improved. The study intended to construct a MIMO process control system with soft computing methods for prediction and parameter control. The system included MIMO process sub-system, measurement sub-system, deviation detection and analysis sub-system, artificial neural network process output prediction sub-system, genetic algorithm parameter design optimization search sub-system and parameter adjustment sub-system. The study detailed the internal operation for each sub-system and relationship among each other. Compared to the past study, the greatest difference is that the soft computing in the study is to integrate artificial neural network and genetic algorithm applications, which was more effective than a single artificial neural network system according to evaluation result. Besides correct prediction and diagnosis for the noise due to system deviation, it effectively controls process input and output as well as achieves process optimization.

並列關鍵字

Multiple-Input Multiple-Output ； Engineering Process Control ； Statistical Process Control ； Soft Computing

國際替代計量

建構多重輸入多重輸出之製程管制系統－整合應用SPC/EPC模式與柔性演算法

未授權

主題瀏覽