Statistical Feature-Based Evidential Reasoning for Equipment Health State Assessment

In the health state assessment of complex equipment, there are some issues such as high-dimensional data or correlation variables. Therefore, a statistic-based feature fusion method for equipment health state assessment is proposed, which contains advantages in indicator decorrelation and multisource information fusion. Specifically, principal component analysis (PCA) is introduced to extract uncorrelated principal component features. Considering that the principal components have no definite physical meaning, a statistic-based feature transformation method is developed to achieve conversion from the principal component feature to the evidence belief degree. Furthermore, the evidence weight for feature fusion can be calculated from the principal component contribution rate. Finally, the equipment health state can be assessed based on the evidential reasoning rule. Numerical simulations are performed to show that the proposed method can reduce the fusion uncertainty. The practical application is validated with case studies of the turbofan engine (TE) and the inertial measurement unit (IMU), which demonstrates the implementation process and assessment results.