Data-driven command shaping for improved torque control accuracy of torsional dampers

Abstract

Automotive torsional dampers are designed to absorb torque fluctuations generated by the engine and deliver a smoother torque to the components located toward the rear. During the torsion evaluation test of these dampers, torsional torque is repeatedly applied while the damper rotates. Given that torsional dampers consist of dead zone elements and multi-stage spring elements, achieving effective torque control can be challenging during these evaluations. In this study, we proposed a method to enhance the performance of the torque control system used in traditional torsional evaluation test machines. Our approach involves modifying the target torsional torque reference by adding a one-pulse signal. Then, the shape of this one-pulse signal is determined through Bayesian optimization. We validated the effectiveness of this proposed method through both simulations and an experiment conducted on actual equipment.