Sub-stages Auto Regression Principal Component Analysis Fault Monitoring For Fermentation Process

The fermentation process is the most promising branch of biological field, Bio fermentation technology has played an increasingly important role in modern food, medicine and other high value-added processing industry. Bio fermentation industry will become one of the leading pillars of China's national economic development in the next few years. But the development of technology is a double-edged sword, many safety problems also highlight one by one, while the vigorous development of fermentation technology bring considerable changes for our production and life, this forcing people to pay more and more attention to the safety and reliability of the production process. Therefore, in order to improve the maintainability and security of fermentation process, and improve the quality of products, the production process is in urgent need for fault monitoring, capturing the change of each detection variables immediately, feeding the abnormal situation up the operator, making disposal timely, guaranteeing the continuity, stability and safety of fermentation process.This topic analyzed multi stage characteristics and dynamic characteristics of fermentation process deeply, and for the defects of traditional methods for process monitoring, to study a novel online monitoring algorithm for fermentation process, to reduce the leaking alarm rate and nuisance alarm rate of process monitoring.(1) Implementation of the batch weighted soft classifying based on Affinity Propagation ClusteringFor the multiphase property and slow time-varying characteristics inherent in the fermentation process, analyzing the relationship between stable phase and transition process deeply, on the basis of AP realize hard Division for stage based on single batch, fusing multiple batches data by introducing Inverse Distance Weighted, avoiding the limitation of a single batch as the input of AP cannot represent the stage characteristics of the entire production process, to achieve a reasonable division of the transition phase.(2) Research on the stage attribution of real time sampling points based on information transmissionStudy on the selection of the optimal model of the new time sampling point for online monitoring, information transmission is introduced to determine the stage attribution of real time sampling points, and to solve the problem of optimal model selection for unequal length batch, realizing the new time sampling points can fall into the corresponding actual operation stage, and to select monitoring model corresponding to the stage to realize the monitoring of real-time sampling point.(3) Extraction of sub-phase Auto Regression-Principal Component Analysis fault monitoring method for fermentation processThe time series of single variable process analysis method is extended to the multivariate case, distinguishing the stable stage and transition process with strong dynamic property. After that AR-PCA and MPCA model was established for the transition phase and the stable phase respectively, while eliminating the dynamic of transition phase, can effectively reduce leaking alarm and false alarm.(4) Field experiment study on the fermentation of Escherichia coliApplying the proposed method in this article to the actual production process, and to validate the rationality and validity of this method with the help of Escherichia coli fermentation experiment. The result indicated that this method can effectively reduce the leaking alarms and nuisance alarms than the traditional method, having more reliable monitoring performance, and can be a good practice guide for the operator to find and remedy fault in a timely and effective manner.