Reduction of the experimental effort in engine calibration by using neural networks and 1D engine simulation

In this study, two effective methodologies are proposed to overcome some critical issues concerning the base calibration of engine control parameters. Specifically, Neural Networks and 1D CFD simulation were alternatively adopted to reliably calibrate specific ECU functions starting from a reduced number of experimental data. The calibration performance fall within acceptable limits even when significant cuts are made to the experimental data usually acquired for calibration purposes, demonstrating that the proposed methodologies can be useful to significantly reduce the dynamometer tests and their related times and costs.