At IAV, the Design of Experiments (DoE) method is normally used to provide the basic calibration for combustion-engine control units. Just a few measured values provide the basis for creating a statistical model that predicts the effect the controlled variables have on the output variables. This model can be a polynomial, a neural network or any other approximator. With further measurement data, the model’s quality can be assessed and improved.
By applying the DoE process in balancing maps with several degrees of freedom, we can significantly shorten the time taken to calibrate combustion engines. Using the statistical models we obtain, we can optimize emissions, consumption and other parameters offline at the desk while taking account of trade-off limits. Combined with test-bench control, the method can be used in automatic mode.