An Information Fusion Method of Fault Diagnosis Based on Interval Basic Probability Assignment

In order to deal with the ambiguities in the processes of signal acquisition and fault feature extraction, each Fault template (FT) in fault mode database is modeled as a set of Gaussian membership functions by statistically analyzing typical fault data. Every real-time Fault feature (FF) is modeled as a single membership function extracted from the on-line monitoring data. The matching degree between a FF and every FT can be obtained based on random sets model of fuzzy information. An Interval basic probability assignment (IBPA) can be calculated from this matching degree by the Modified Latin Hypercube Sampling Monte Carlo (MLHSMC) technique. Several IBPAs can be fused by interval evidence combination rule. A diagnosis decision-making can be done according to the fusion results. Finally, the diagnosis examples of machine rotor show that the proposed method can enhance accuracy and reliability of fusion-based diagnosis system.