Simplified fuzzy ARTMAP(SFAM) is a simplification of fuzzyARTMAP(FAM) in reducing architectural redundancy and computationaloverhead. The performance of individual SFAM depends on the ordering oftraining sample presentation. A multiple classifier combination scheme isproposed in order to overcome the problem. The sum rule voting algorithmcombines the results from several SFAM’s and generates reliable and accuraterecognition conclusion. A confidence vector is assigned to each SFAM. Theconfidence element value can be dynamically adjusted according to the historicalachievements. Experiments of recognizing mechanical workpieces have beenconducted to verify the proposed method. The experimental results have shownthat the fusion approach can achieve reliable recognition.