ICE Simulation- Calibration and Optimization |

ICE Simulation- Calibration and Optimization

Oldirich Vítek (Czech Technical University in Prague)

The document concerns application of optimization approach when dealing with thermodynamic modelling in internal combustion engine using detailed 0-D/1-D model. Three different possibilities are mentioned – calibration of ICE mathematical model, engine setting optimization and engine control from system simulation point of view (controller optimization is not meant by this term). These possibilities are listed in order of importance (based on author’s opinion).

Calibration is critical phase of engine model work-flow – only properly calibrated model can provide sound conclusions. Optimization is supposed to provide a great help due to the fact that an error between prediction and measurement can be minimized. This is in line with main goal of calibration which is to find the values of calibration parameters so that the mathematical model matches experimental data as closely as possible. On the other hand, user’s experience is invaluable as each calibration is a unique process – there is no general method to calibrate engine model properly, moreover there might be significant errors in experimental data. Optimization of engine setting under steady-state operation is a standard task. Due to increase in computer power, complex multi-variable multi-target multi-constraint problems can be solved within reasonable time – this allows to perform optimizations of higher quality as internal combustion engine features strong non-linear interaction of many input parameters.









Engine control optimization is very challenging task due to the need to simulate transient response and the fact that optimization is more difficult from theoretical point of view. Strong development is expected in this application field before it becomes an industry standard task. At present time, a common approach is to derive a model which is much simpler from complexity level point of view. Such model is significantly faster however it usually ’looses’ most of its physical meaning hence its predictive ability is limited or even lost. Some examples are briefly presented to show application of optimization in all three above-mentioned categories. Some results are shown in each example so that the quality of optimization/calibration process can be estimated.