Development of Malfunction Diagnosis System for Fermentation Based on Fuzzy Inference and Its Application to Industrial Scale Production of Alkaline Cellulase

A malfunction diagnosis system based on fuzzy inference was developed for industrial scale fermentation with an alkaliphilic cellulase-producing bacterium. The index J indicating the degree of malfunction in the fermentation process is calculated from the deviations of actual values from the standard values with respect to three kinds of process data obtained from on-line sensors. The diagnosis system proposed in this study is as follows. The fermentation process is automatically categorized into one of three states (normal, intermediate and abnormal) according to computation of the J value. Only when the fermentation process is categorized to an “intermediate” state, an operator should supervise the process by means of direct observation of process data. The diagnosis system was applied to 100 runs of batch fermentations conducted on an industrial scale. In all cases, it was confirmed that this system worked to offer the correct diagnosis results. The introduction of the system resulted in the reduction of over 99% in the operators' diagnosing labor because their supervising tasks could be limited to the intermediate cases.