A simulation-based approach for developing optimal calibrations for engines with variable valve actuation

Abstract

Variable Valve Actuation (VVA) technology provides high potential for achieving improved performance, fuel economy and pollutant reduction. Benefits of VVA stem from better breathing and the ability to control internal residual. However, additional independent control variables in a VVA engine increase the complexity of the system, and achieving its full benefit depends critically on devising an optimum control strategy. The traditional approach relying on experimentation is in this case prohibitively costly, since the number of tests increases exponentially. Instead, this work formulates the task o defining actuator set-points as an optimization problem. It identifies simulation needs for supporting development of a generic methodology, capable of handling increased number of degrees offteedom. A high-fidelity tool predicts effects of various devices being considered. Since solving an optimization problem requires hundreds of function evaluations, direct use of the high-fidelity simulation leads to unacceptably long computational times. Instead, the Artificial Neural Networks (ANN) are trained with high-fidelity simulation results and used to represent engine's response to different control variable combinations with greatly reduced computational time. The paper describes a comprehensive simulation-based methodology, provides details of high-fidelity and surrogate modeling techniques, and then demonstrates application on a prototype four-cylinder spark-ignition engine with dual independent cam-phasers. Copyright © 2007, Institut français du pétrole.