Development of a model-based auto-calibration methodology with genetic algorithms focusing on fuel efficiency and emission compliance over real-world driving cycles

The aim of this study is to develop a methodology for the optimization of a diesel engine calibration in terms of fuel consumption while being emission compliant for a prede ned route by using model-based calibration techniques with genetic algorithms. A 2 liter diesel engine of light commercial vehicle data was used to develop the proposed methodology. The model-based calibration environment was structured in Simulink with a combination of an engine control unit (ECU), internal combustion engine (ICE) and engine aftertreatment system (EAS) models. Accuracies of the models were examined and considered as adequate for this methodology development study. After the models were created, the calibration domain for a prede ned route was computed in Matlab by weighting the engine operating points based on frequency and fuel consumption. These points in the ECU model were de ned as the optimization domain for the genetic algorithm in Matlab. The cost function of the optimization consists of fuel consumption, performance parameters, mechanical limits of ICE, and tailpipe emission limits de ned by regulations. After iterating with di erent genetic algorithm con gurations, the fuel consumption over the prede ned RDE route was decreased by 3.07% at the end of simulations performed in Simulink. The smooth calibration maps were obtained while not violating any of the limits de ned in the cost function. In addition to model accuracy improvements, for further study, the models can also be embedded into an ECU and optimization can be performed automatically according to the de ned route, weather and driver information.