Gear Misalignment Optimization for Electric Drive Unit Using Machine Learning Based Prescriptive Analytics

In Electric vehicle Drive Unit Gears, high mesh misalignments result in shift in load distribution of a gear pair that can increase contact and bending stresses. It can move the peak bending and contact stresses to the edge of the face width and increase gear noise as well. Lower misalignment value is often required to reduce the peak bending and contact stresses and have a balanced load distribution along the gear flank, which in turn helps in reducing noise and improving durability of drive unit.

This paper delineates Prescriptive Analytics method that combines virtual simulations, Machine learning (ML) and optimization techniques to minimize different gear misalignments for the electric vehicle drive units. Generally, the manual optimization process is carried out by sequential modifications of stiffness of individual components. However, this process is time consuming and does not account for interactions between the components. In this study, firstly, Machine learning models are developed based on design of experiments (DOE) simulations. These ML models are used as surrogates for actual simulations in generic algorithms (Differential Evolution) based optimization techniques. It finally prescribes changes in stiffness of different components to get optimum misalignment value.