A New Evidential Reasoning Rule Considering Evidence Correlation with Maximum Information Coefficient and Application in Fault Diagnosis

The evidential reasoning (ER) rule has been widely adopted in engineering fault diagnosis, yet its conventional implementations inherently neglect evidence correlations due to the foundational independence assumption required for Bayesian inference. This limitation becomes particularly critical in practical scenarios where heterogeneous evidence collected from diverse sensor types exhibits significant correlations. Existing correlation processing methods fail to comprehensively address both linear and nonlinear correlations inherent in such heterogeneous evidence systems. To resolve these theoretical and practical constraints, this study develops MICER-a novel ER framework that incorporates correlation analysis based on the maximum mutual information coefficient (MIC). The proposed methodology advances ER theory by systematically integrating evidence interdependencies, thereby expanding both the theoretical boundaries of ER rules and their applicability in real-world fault diagnosis. Flange ring loosening fault diagnosis and flywheel system fault diagnosis cases are experimentally verified and the effectiveness of the method is demonstrated.