The Study Of Multi-stage Statistical Monitoring For The Fermentation Process

Fermentation process is a typical batch process, for the characteristics such as nonlinear, short production cycle, internal dynamic properties changing rapidly, much more complex than continuous process, and product quality is more vulnerable to external uncertainty factor. To ensure the security of the fermentation process and control system, and improve the product yield and quality, it's necessary to carried out online monitoring and fault diagnosis on production process.Statistical process monitoring is a data-driven method, which is based on multivariate statistical methodologies, through a variety of data processing method to obtain information of normal operation and fault feature and guarantee the efficient, safe and product quality of the process. In view of characteristics of fermentation process, some improvements of tradition MPCA methods have been made at different degree. The new method is applied on the monitoring of the penicillin fermentation process, and achieved good results.The main contributions are as follows:(1) Principal Component Analysis (PCA) and its related theory are introduced, the method of Multi-way Principal Component Analysis (MPCA) is discussed in detail. Application on penicillin fermentation process monitoring and the result is not very effective.(2) In view of dynamic properties of fermentation process, a multi-stage MPCA method is proposed, it divided the batch process into several phases according to its dynamic changes, at the same time, several models are used in the penicillin fermentation process. The simulation result of penicillin cultivation process shows that the multi-stage MPCA method is more reliable than the traditional one.(3) In view of unequal length of the data between batches, an improved multi-stage MPCA method is proposed, which divides the batch process into several phases with Fuzzy C-means (Fuzzy c-Mean, FCM) algorithm, then the Dynamic Time Warping (DTW) algorithm is used to synchronize the time length of all stages of data, and several MPCA models are used in the penicillin fermentation process. Monitoring results shows that the improved method has a better performance.