Weighted Evidential Fusion Method for Fault Diagnosis of Mechanical Transmission Based on Oil Analysis Data

Condition monitoring (CM) and fault diagnosis are critical for the stable and reliable operation of mechanical transmissions. Mechanical transmission wear, which leads to changes in the physicochemical properties of the lubrication oil and thus severe wear, is a slow degradation process that can be monitored by oil analysis, but the actual degradation degree is difficult to evaluate. To solve this problem, we propose a new weighted evidential data fusion method to better characterize the degradation degree of the mechanical transmission through the fusion of multiple CM datasets from oil analysis. This method includes weight allocation and data fusion steps that lead to a more accurate data-based fault diagnostic result for CM. First, the weight of each evidence is modeled with a weighted average function by measuring the relative scale of the permutation entropy from each CM dataset. Then, the multiple CM datasets are fused by the Dempster combination rule. Compared with other evidential data fusion methods, the proposed method using the new weight allocation function seems more reasonable. The rationality and superiority of the proposed method were evaluated through a case study involving an oilbased CM dataset from a power-shift steering transmission.