Development of a Design System for Wheel Bearings Including Prediction Functions for the Finite Element Method Using Lasso Regression and Bayesian Optimization

Wheel bearings play a critical role in providing smooth rotation when vehicles move in straight line and turning motions. Automotive electrification continues to accelerate, emphasizing specific market demands such as lightweighting, lower torque, and quietness. In addition to the above requirements, reduced development timing for automotive programs is required. Recently, the number of bearing manufacturers that utilize Model-Based Development (MBD) have been increasing in order to reduce development time. NTN has developed an integrated calculation automated system which is called Axle Bearing Integrated Calculation System (ABICS) that automates each step of the design processes for third generation hub bearings. After ABICS was released, man-hours per development project were reduced by 80 percent compared to previously used design flows in which each step of the design processes had been performed by a human. In order to further reduce development timing, even more focus has been placed on reducing man-hours. More recently, the utilization of surrogate models to approximate various physics models has been considered. An efficient use of surrogate models can bring about considerable savings in computational resources and time. Therefore, embedding surrogate models in ABICS was considered. In this research, developing a surrogate model using a database in ABICS was completed. Lasso regression which corresponds to one of the Finite Element Method (FEM)-based calculations included in ABICS was utilized and provides the designer with the reduction in the hub ring inner diameter when the inner ring is press-fit. Moreover, ABICS was modified so that the designer can confirm whether the predicted value using a surrogate model meets customer requirements or not, developing a function which combines surrogate models based on the Lasso regression with Bayesian optimization. Further modifications to ABICS may make it possible for man-hours to be reduced even further.