An approach for physical model adaptation based on transient measurements

In order to cope with the ever increasing complexity of the control of internal combustion engines, model-based calibration processes have become well established in recent years. Physical engine models are often used in early development stages, when no measurement data is available. At that stage, physical models can give a good, even though rough, overview over the overall engine behavior. Later on in the development process, however, an adaptation to a particular engine is usually required in order to match that particular engine´s output as good as possible. In contrast to data-driven statistical models, physical models have the disadvantage that one has to decide how to calibrate them, since they have many parameters, which are in addition often difficult to measure. One way of adaptation to measurement data, is by using statistical correction models. These can accommodate modeling shortcomings such as unknown or non-representable physical effects.