Voting Algorithm of Fuzzy ARTMAP and Its Application to Fault Diagnosis

Simplified fuzzy ARTMAP (SFAM) is a simplification of fuzzy ARTMAP (FAM) by reducing the computational overhead and architectural redundancy. SFAM is powerful neural network in the application of prediction and classifier. Individual SFAM performance depends on the ordering of training sample presentation. A multiple classifier combination scheme is proposed in this paper to obtain reliable and accurate fault diagnosis. SFAMs input the data vectors in the complement code format and output fault diagnosis results. The sum rule voting algorithm combines the SFAMs outputs and generates final conclusion. A weight is assigned to each SFAM according to its historical achievements. Case study has shown that the proposed method is effective in fault diagnosis application.