Optimization of Engine Torque Management Under Uncertainty for Vehicle Driveline Clunk Using Time-Dependent Metamodels

Abstract

Quality and performance are two important customer requirements in vehicle design. Driveline clunk negatively affects the perceived quality and must be minimized. This can be achieved using engine torque management, which is part of engine calibration. During a tip-in event, the engine torque rate of rise is limited until all the driveline lash is taken up. The engine torque rate of rise can negatively affect the vehicle throttle response, which determines performance. The engine torque management must be therefore balanced against throttle response. In practice, the engine torque rate of rise is calibrated manually. This article describes an analytical methodology for calibrating the engine torque considering uncertainty, in order to minimize clunk, while still meeting throttle response constraints. A set of predetermined engine torque profiles are considered, which span the practical range of interest. The transmission turbine speed is calculated for each profile using a bond graph vehicle model. Clunk is quantified by the magnitude of the turbine speed spike. Using the engine torque profiles and the corresponding turbine speed responses, a time-dependent metamodel is created using principal component analysis and kriging. The metamodel predicts the turbine speed response due to any engine torque profile and is used in deterministic and reliability-based optimizations to minimize clunk. Compared with commonly used production calibration, the clunk disturbance is reduced substantially without greatly affecting the vehicle throttle response.