A hybrid support vector machine and fuzzy reasoning based fault diagnosis and rescue system for stable glutamate fermentation

Abstract

In industrial glutamate fermentation by biotin-auxotroph Corynebacterium glutamicum, biotin content variation in corn slurry greatly affects fermentation performance. To maintain the fermentation stability, a hybrid support vector machine (SVM) and fuzzy reasoning based fault diagnosis and rescue system was developed. The system uses SVM outputs as the inputs of the fuzzy reasoning classifier having a couple of production rules and condition membership functions related with SVM outputs, to categorize multiple faults. The effectiveness of the proposed system was verified in a normal fermentation run and two abnormal runs with different biotin initial-content faults with the aid of using on-line measurable data such as ammonia consumption rate and CO2 evolution rate. The results indicated that the proposed faults-diagnosis system could cluster multiple fermentation faults quickly, accurately and stably, and faults and their types could be identified at the earliest fermentation stage. Based on the diagnosis results, the proposed system was further applied for real fault-rescue in two fermentations with different biotin initial-content faults. In both cases, by immediately taking relevant rescue measures after identifying the faults and their types, glutamate fermentations with initial faults were restored to normal, and final glutamate concentrations reached a normal level of 75–80 g/L at 34h.