MSFF-CBR: case-based reasoning technology for adaptive multi-information fusion fault diagnosis

Abstract With the transmission of massive, high-dimensional, low-value density data, measurement systems are able to capture extensive multi-sensor data. However, challenges such as high-dimensional incompatibility and granularity destruction pose significant issues for existing multi-sensor fusion theories. In this study, case-based reasoning (CBR) is applied to fault diagnosis using multi-information fusion. MSFF-CBR, a two-layer information fusion reasoning system embedding a multi-sensor feature fusion layer (MSFF), was designed. By incorporating the novel F2-Apriori algorithm and an attribute importance measurement model based on the multi-granulation rough set model, MSFF demonstrates exceptional feature fusion efficiency. In the decision fusion layer, three distance-based similarity measurement modes were developed to enable case retrieval and demonstrate the adaptability of MSFF to various sensitivity metrics. The model exhibits efficient multi-sensor information fusion for fault diagnosis under various operating conditions of multi-stage reciprocating compressor.