Fault Diagnosis Based on MFICA-FFRLSSVM for Batch Process

For the sake of surmount problems of hatch process precision of single fault diagnosis methods and low efficiency of the traditional, a new ensemble approach based on multi-way fast independent component analysis (MFICA) and recursive least squares support vector machines with forgetting factor (FFRLSSVM) is proposed. Firstly, MFICA is used to abstract rapid information which belongs to non-Gaussian hatch process. Secondly, the faults are sorted by FFRLSSVM rapidly. Owning to the forgetting factor application, history data are forgotten which reduce the complexity of computational. Experiment shows, compared with conventional single fault diagnosis methods, the accuracy and the adoption of MFICA-FFRLSSVM algorithm is higher.