Real-Time Estimation of Backlash Size in Automotive Drivetrains
Prithvi Reddy, Mahdi Shahbakhti, Maruthi Ravichandran, Jeff Doering
Abstract
The presence of backlash in automotive drivetrains causes the so-called clunk (a.k.a. shunt) phenomenon during reversals in the sign of the actuator torque. This clunk manifests as an audible noise when the gears make contact at the end of the lash traversal, and thus, affects the drive comfort of the vehicle. To mitigate the clunk, automotive OEMs employ a variety of actuator torque shaping strategies, which require knowledge of the size of the backlash in order to be effective. Furthermore, since the size of the drivetrain backlash is expected to vary significantly over the lifetime of the vehicle and/or from vehicle-to-vehicle (due to manufacturing variations), there is a requirement to estimate the backlash size in real-time so as to maintain the effectiveness of these strategies. To this end, the current work develops an innovative Kalman filter-based lash size estimator that uses readily available speed and torque signals from the vehicle CAN bus. As part of the development, we evaluate the efficacy of the proposed estimator using both simulations and test vehicle data. The evaluation also includes a study of the robustness of the estimator to variations in the actuator torque trajectory and the calculated road load torque, presence of CAN jitter in the measured speed signals, and variations in backlash size, driveshaft compliance, and tire-road interaction. Furthermore, we analyze the computational feasibility of the estimator using processor-in-loop simulations in a dSPACE prototype controller. Both the performance and robustness studies prove the effectiveness of the proposed backlash size estimation system.