Microbiological fermentation fault diagnosis based on multi-layer support vector machine

Microbiological fermentation process is a purebred culture process. Sometimes because of bad operation, some bacteria in the microbiological fermentation process may pollute mycelia. In order to reduce the loss arising by such bacteria, it is very important to diagnose the abnormal states in time. We provide a new method of fault diagnosis, i.e. combination the nonlinear principal component analysis with support vector machines, which may not only extract the main monitor variables from many monitor variables, but also obtain decision function with excellent generalization performance from limited samples of fault. In this paper we provide the algorithm for the method of fault diagnosis. The experiment demonstrates this method is very valid.