An On-line Automatic Feeding Control And Fault Diagnosis System For Glutamate Fermentation Process

This paper researched the relationship between ammonia consumption amount and glucose consumption amount. Then according to the linear relationship of them, a glucose automatic feeding system for glutamate fermentation was designed and built. The system could monitor ammonia consumption amount on-line, and then reckon the glucose concentration in fermentor automatically and real-timely. Through the intercomparison of reckoned glucose concentration and setting glucose concentration, the system could reckon the optimal feeding amount in any control interval, according to which glucose concentration could be control at any setting level steadily. In batches of fermentation, effectiveness of the system was reflected. Under the condition of high setting glucose concentration (100g/L), the last glutamate concentration was only 37g/L. When culture time was same and glucose concentration was set at 10g/L, the last glutamate concentration could reach 80g/L, which was about two times of the former. The fermentation performance significantly improved. Because of the fluctuation of the relationship between ammonia consumption amount and glucose consumption amount in different batches of fermentation, the system may lead to serious operation fault (glucose exhaustion) when the setting glucose concentration was very low (1-5g/L). The fermentation can't normally proceed.According to the relationship between oxygen consumption amount and glucose consumption amount, another glucose automatic feeding system was designed and built. There was great fluctuation of the relation coefficient between oxygen consumption amount and glucose consumption amount in batches of fermentation. In a batch of fermentation, using the system with a less relation coefficient, led to the fault (glucose exhaustion) happened frequently, and the last glutamate concentration was only 62g/L. This system was less effective than the former.For improving the former system, taking the later system as the research object, a BP neural network was trained with normal and faulty fermentation data, and then built an automatic fault (glucose exhaustion) diagnosis system, which can recognized the fault above accurately.The automatic glucose feeding system based on ammonia consumption amount and the automatic fault diagnosis system could be combined, forming an advanced on-line self-adaptive control system hopefully. When fermentation fault (glucose exhaustion) happened, BP recognition system can recognize the fault in 5 minutes, then in next minute the fault is eliminated by modifying the reckoned glucose concentration and the relation coefficient between ammonia consumption amount and glucose consumption amount, which makes the control system more robust and reliable.