Failure Diagnosis Method Based On Auto Associative Neural Network Applied In Fermentation Process

The fermentation process is one extremely complex biochemistry reaction process, which has general nonlinear system's large inertia, relatedness, uncertainty, moreover, because fermentation process's some key parameters like matrix density and product density may not the online survey, therefore fermentation process's control is more complex than the general nonlinear system. Because fermentation process's parameter measuring technique falls behind by far the system real-time control request, causes the fermentation process the control to satisfy the timely request with difficulty. Therefore using the failure detection and the diagnosis technology to the monitoring and the optimization of the fermentation process is very essential.In recent years, along with artificial intelligence technology rapidly expand, specially knowledge engineering, expert system and artificial neural networks in diagnosis. Obtained based on artificial intelligence's failure diagnosis method has been more thorough, system's research, has had the expert system failure diagnosis method, based on the case failure diagnosis method, the neural network failure diagnosis method, the fuzzy failure diagnosis method and so on. Compares with the tradition failure diagnosis method, the breakdown intelligence diagnosis method can simulate human brain's logical thinking process, carries on the inference using the expert knowledge to solve the complex diagnosis problem, has represented the failure diagnosis development direction.The present paper has used the auto-association neural network(AANN)to carry on the online failure diagnosis to the glutanic acid fermentative process the validity and the practicability research. The auto-association neural network is one kind of belt bottleneck level,5 structure special neural network. Through carried on suitable screening to fermentative process's apparent variable, we have selected some variables finally as the training data in AANN. The different performance's fermentative process may obtain the cluster and may carry on the failure diagnosis using evaluating indicator J to the glutanic acid fermentation.In view of the glutanic acid fermentation process which has cannot online-measuring quantity, has conducted research of the soft modelling, and gives the experimental result, the experimental result indicated that Least Squares Support Vector Machines(LSSVM) has the very good estimate ability. Simultaneously this article proposed forecasts the technology ahead of time based on the LSSVM. Base on LSSVM of soft forecast technology base, take the glutanic acid fermentation as the example, did has planned the online predict that proposed some pots approved the quota which and the qualitative characteristic exceptionally judged, could be exceptionally prompt to the fermentative process gives the early warning, and using the AANN's formidable data memory ability, further determined the break down reason and took the corresponding measure.