The subject of this master course thesis is to investigate on model-based calibration, in particular on the modelling stage. The first phase has been dedicated to the DoE theory, the first step of an engine calibration work: very essential to reduce the tests number and increase the model quality at the same time. Then an alternative method was found for outlier detection. These theoretical considerations have been applied on experimental diesel engine data: gaussian and polynomial models are compared on different tests, with different engines and different strategies in order to evaluate their precision in fitting and predictivity. Model-based calibration(MBC) toolbox is the employed software for the activity.