Fault Detection of Multi-phase Batch Process Based on Adaptive FCM

Batch processes have the characteristic of more operation phases in nature. The standard Fuzzy C-Means (FCM) algorithm for phase partition of batch processes needs to a given phase partition number beforehand, initialize clustering centers randomly, and is sensitive to noise and outliers. For the above problems, the adaptive FCM algorithm using clustering validity function is proposed to achieve the adaptive partition of batch process operation phases. The method obtains the initial clustering center set on the basis of maximum minimum distance rule, through adaptive iteration way determines the optimal clustering number by introducing the clustering validity function. The MICA model based on the improved phase partition method is applied to fault detection of industrial penicillin fermentation process and the experimental results verify the effectiveness of the proposed method.