Fault degree evaluation for rolling bearing combining backward inference with forward inference

A novel evaluation method combining backward inference with forward inference was proposed to quantitatively evaluate fault degree of deep groove ball bearings with localized defects. The separation-integration strategy was used for modeling the forward infrence. A detailed multi-body dynamic contact model was developed with the assumptions that rolling elements-inner race had multi-DOF and outer race was deformable in radial direction, which was modeled with finite elements (FE). In this model the nonlinear stiffness was obtained by the application of Hertzian elastic contact deformation theory and the localized defects were modeled through surface profile changes. In order to study the influence of the transmission path, the bearing housing was taken into account in bearing model. The backward inference was translated into comparison of geometric distance, and the defects location and size can be determined by inputting the measured features. In order to reduce the relative error, a concept of damage range was introduced to characterize the fault degree by utilizing range code. Finally, the effectiveness and accuracy of the proposed method in quantitative evaluation was verified.