Model-Guided Extremum Seeking for Diesel Engine Fuel Injection Optimization

Abstract

Extremum seeking (ES) is a model-free optimization method that has been successfully applied to the automotive engine calibration. However, the slow convergence speed of ES algorithms presents a concern for the online application of this approach to engine optimization where time constraint is critical. In this paper, a model-guided ES architecture is proposed and is showcased with fuel injection optimization for a diesel engine at different operating points. In particular, by integrating a computable engine model into the ES scheme and monitoring the operation point online, the optimal combustion phase under the current operating point can be estimated and this estimation serves as the init..