Interactive optimization of internal combustion engine tests by means of sequential experimental design

The application of a Sequential Experimental Design technique to the design of experiments for Internal Combustion Engines is presented, in order to maximize the information achieved in experimental tests by guiding them in structured way. This technique can be used whenever the parameters of a linear or non linear model have to be estimated from experimental data, or also from the results of a computational code. The new experimental condition is determined interactively by minimizing the expected volume of the model parameter confidence region. The method has been applied to the estimation of a black-box model for the specific consumption of a spark ignition engine as a function of four operating variables, used for engine control applications. The proposed experimental strategy has been compared with conventional ones, for a data set of more than 1000 experimental conditions, and a more rapid convergence to the reference values of model parameters has been observed. The results suggest that substantial benefits can be achieved in reducing and guiding the experimental tests for the study and the design of Internal Combustion Engines, for which extensive investigations are often required