A Case-based Reasoning with Feature Weights Derived by BP Network

Case-based reasoning (CBR) is a methodology for problem solving and decision-making in complex and changing environments. This study investigates the performance of a hybrid case-based reasoning method that integrates a multi-layer BP neural network with case-based reasoning (CBR) algorithms for derivatives feature weights. This approach is applied to fault detection and diagnosis (FDD) system involves the examination of several criteria. The correct identification of the underlying mechanism of a fault is an important step in the entire fault analysis process. The trained BP neural network provides the basis to obtain attribute weights, whereas CBR serves as a classifier to identify the fault mechanism. Different parameters of the hybrid methods were varied to study their effect. The results indicate that better performance could be achieved by the proposed hybrid method than that using conventional CBR alone.