发酵过程多阶段统计监控方法的应用研究-检测技术与自动化装置专业论文.docxVIP

Abstract Abstract Fermentation process is a typical batch process ，for the characteristics such as nonline叽 short production cycle，intemal dynamic properties changing rapidly，much more complex than continuous process ，and product quality is more vulnerable to extemal uncertainty factor. To ensure the security of the fermentation process and control system，and improve the product yield and quality，its necessaryωcarried out online monitoring and fault diagnosis on production process. Statistical process monitoring is a data-driven method，which is based on multivariate statistical methodologies ，由rough a variety of data processing method to obtain information of normal operation and fault feature and guarantee 由e 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.ηle new method is applied on 伽 monitoring of the penicillin fermentation process，and achieved good results. 丁he main contributions are as follows: Principal Component Analysis (PCA) and its related 伽eory are introduced ，the method of Multiway Principal Component Analysis (MPCA) is discussed in detail. Application on penicillin fermentation process monitoring and the result is not very effective. 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 penicil1in fermentation process. The simulation result of penicillin cultivation process shows that the multi-stage MPCA method is more reliable than the traditional one. 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 phωes with Fuzzy C-means (F山zy c-Mean，FCM) algorithm ，then the Dynamic Time Warping (DTW) algorithm is used to synchro