Application to microbiological fermentation fault diagnosis for relevance vector machine

Microbiological fermentation is a kind of complicated batch process that is severely nonlinear and time variant. There are many process parameters needed to monitor and control in order to make the microbiological fermentation process successful. So it is important to fault diagnose for microbiological fermentation sensor. In this paper we provide a new method for fault diagnosis of microbiological fermentation, i.e. two relevance vector machines which are viewed as observer and classifier respectively. The observer is applied to estimate release rate of carbon dioxide to gain residual sequence. The classifier is applied to the residuals. The results of experiments demonstrate that this method can effectively diagnose instrument faults.