HSMM為基礎的錯誤診斷於批次操作系統

HSMM Based Fault Diagnosis for Batch Processes

江衍澄(Yan-Cheng Jiang)

指導教授 ： 陳榮輝

中原大學/工學院/化學工程學系/碩士(2009年)

https://doi.org/10.6840/cycu200900368

摘要

本研究在MPCA(Multi-Principal Component Analysis)結合HMM(Hidden Markov model)與HSMM(Hidden semi-Markov model)為基礎的線上即時監控與診斷的模式。HMM與HSMM為具有時間的序列關係的隨機模式，能夠直接利用其訓練出的分佈來建立管制界線，改善傳統PCA建立管制界線需假設分佈為常態分佈的情形，並且利用MPCA來壓縮變數，提昇HMM與HSMM訓練的速度及準確性。此外批次程序中常有多個stage所串連而成，而真實stage切換時間是不固定的，由於HSMM能夠表示狀態的持續分佈的特性，能夠表示出stage切換時間分佈情況，改善HMM狀態持續分佈不符合真實的情形。

另外傳統的批次監控方式，都假設批次與批次之間獨立的情況之下，在實際情況下批次與批次之間會有存在交互關聯性，因此本研究提出利用MPCA的方式來萃取批次與批次間關聯性的數據，在利用HSMM來訓練時間序列之關係，來建立線上即時監控系統。在本文中將以盤尼西林發酵製程為測試例子，測試本文所提出的兩個問題：(1)stage-based批次製程的線上即時監控與診斷，(2)批次與批次之間的線上即時監控系統，並與過去的監控方式去做比較。

關鍵字

批次製程 ； 隱藏式半馬可夫模型

並列摘要

A novel and flexible of on-line batch monitoring and diagnosis approach is proposed based on hidden Markov model (HMM), hidden semi-Markov model (HSMM) and multiway principal component analysis (MPCA). HMM and HSMM have temporal and stochastic properties. They can be building up control limit by model distribution. The main purpose of the MPCA is used to reduce the number of variables, accuracy of model and speed up the convergence of HMM and HSMM. Beside, process has different stages and uncertainty of stage switching time. HSMM use the distribution of state durations to represent the stage switching time. It’s batter than HMM which state durations are not considered.
Traditional batch process monitoring system assumed the correlation between each batch is independent, but real batch processes are not. Consequently, the temporal property of HSMM and the batch-to-batch dynamic characteristics of MPCA, the proposed on-line batch monitoring method are better than conventional methods. This research is used simulated fed-batch penicillin cultivation process which demonstrated two problems: (1) stage-based on-line batch monitoring and diagnosis, (2) two dimension on-line batch monitoring, to illustrate the advantage of the proposed method in comparison to some conventional methods.

並列關鍵字

HSMM ； MPCA ； batch monitoring ； diagnosis

參考文獻

1. Baruah P. and Chinnam R. B., ”HMMs for Diagnostics and Prognostics inMachining Processes” International Journal of Production Research, 43, (2005) 1275

2. Dong M. and He D., “ A Segmental Hidden semi-Markov Model (HSMM)-Based Diagnostics and Prognostics Framework and Methodology” Mechanical Systems and Signal Processing, 21, (2007) 2248

4. Flores-Cerrillo J. and MarcGregor J. F., “ Multivariate Monitoring of Batch Processes Using Batch-to-Batch Information” AIChE J. 50, (2004) 1219

5. Ge M.; Du R. and Xu Y. “Hidden Markov Model based fault diagnosis for stamping processes” Mechanical System and Signal Processing 18 (2004) 391-408

6. Hu J.; Geklim S. and Brown M. K.,”Writer independent on-line handwriting recognition using an HMM approach” Pattern Recognition 33(2000) 133-147