Method for monitoring faults in fermentation process based on MICA-OCSVM

Abstract

The invention discloses a novel method for achieving real-time fault monitoring on a penicillin fermentation process. In order to guarantee safety and stability of the penicillin fermentation process, it is necessary to establish an effective process monitoring scheme to detect abnormal phenomena timely. The method comprises two steps of off-line modeling and on-line monitoring. The step of off-line modeling comprises the steps of firstly carrying out processing on three-dimensional data of the fermentation process, then extracting independent element information of the data by adopting ICA, and finally carrying out modeling by utilizing OCSVM to structure monitoring statistical magnitude and determining a control limit by utilizing a kernel density estimation method. The step of on-line monitoring comprises the steps of carrying out processing on newly collected data according to a model, calculating the statistical magnitude of the data, and comparing the statistical magnitude and the control limit to judge whether the fermentation process runs normal or not. According to the method for monitoring the faults in the fermentation process based on the MICA-OCSVM, the assumption that a fermentation process variable subjects to Gaussian distribution or non-Gaussian distribution or other distributions is not needed, and the accuracy rate of fault monitor is high.

Description

technical field [0001] The invention relates to the field of data-driven fault diagnosis technology, in particular to a fault diagnosis technology for batch processes. The data-driven method of the present invention is a specific application in fault monitoring of a typical batch process—penicillin fermentation process. Background technique [0002] In recent decades, batch processes have attracted much attention because they can meet the demands of producing high value-added products. However, its mechanism is complicated, its operation complexity is high, and its product quality is easily affected by uncertain factors. As a typical batch process, the penicillin fermentation process has the characteristics of strong nonlinearity, dynamics, and mixed Gaussian distribution. In order to ensure the safety and stability of the fermentation process operating system, an effective process monitoring scheme is established to detect timely Anomalies are necessary. [0003] At pres...

AI Technical Summary

Problems solved by technology

However, the actual industrial process is mostly a mixed distribution of Gaussian and non-Gaussian, so the above method is limited by the assumption that the process variable obeys a specific distribution

Some scholars use PCA and ICA step-by-step modeling to solve this problem, and propose a combined method MICA-PCA to monitor non-Gaussian information and Gaussian information separately, but this method cannot clearly select the number of main independent components, and cannot be effective. Differentiate between non-Gaussian and Gaussian information

Images