Use of thermodynamical models with predictive combustion and emission capability in virtual calibration of heavy duty engines

Abstract

Calibration ensures low fuel consumption operation of an engine within emission limits. It is usually performed in a dynamometer test environment. A cost effective alternative is virtual engine calibration which relies on fast and accurate models with emission prediction capability. NOx emission prediction is particularly important due to NOx and bsfc trade off of diesel engines. A 12.7L diesel engine model is generated and calibrated with dynamometer test data. Extended Zeldovich meachanism is used for NOx emissions calculations. A previously developed methodology is utilized to increase the emission prediction accuracy of the model, and a NOx calibration multiplier map with CA50, Tmax, EGR rate, rail pressure and F/A ratio as the inputs is generated. Model results of the final map are in good correlation with dynamometer outputs. R2 is 0.9382 and nRMSE is 4.1%. The model is used for the engine virtual calibration of five cases. A constraint optimization problem is solved with non-dominated sorting genetic algorithm (NSGA-III) for each case and optimum values of start of injection, rack position, and EGR valve position which result in minimum total operating cost are calculated. Comparison with the results of traditional dynamometer test calibration shows that; model start of injection, boost pressure and air mass flow target predictions are within 3 CA, 0.5 bar and 300 mg/str difference range. The method can be used for rapid and accurate detection of calibration parameters, calibration can be done virtually and a significant decrease in the number of dynamometer tests can be achieved.