Modern Diesel engines require a model based optimisation of the engine control to fully exploit the additional degrees of freedom of modern engines. For identification of combustion engines, different experimental model structures are presented and compared to each other. The local adaptive model approach LOPOMOT is derived from the the local linear model approach LOLIMOT and an adaptive polynomial approach. Further regarded model structures are the in automotive industry well known look-up tables and the individual approximators kernel models. The model structures are generally presented and are rated with regard to applications in an electronic control unit. For the identification of the combustion engine, the combustion outputs NO_x, soot and the engine torque are regarded. Experimental models are presented for measurements from the engine test bed. Stationary and dynamic effects are modelled separately, to avoid the influence of measurement dynamics. Thus, stationary measurements can be applied to identify the combustion models. The connection of these stationary combustion models to a dynamic air path model enables a dynamic overall simulation of the Diesel engine. The stationary and the dynamic model qualities are demonstrated using measurements from the engine test bed. The models are then applied for a stationary and a dynamic optimisation of control functions for the engine control unit. At first a local optimisation is presented for the stationary optimisation, which shows the Pareto front of the emissions NO_x and soot. The subsequent global optimisation minimises the fuel consumption over a test cycle and formulates the emission limits as constraints. Initial values for the global optimisation are taken from the results of the local optimisation. Finally, a robust global optimisation is presented, which regards model uncertainties and variations due to series tolerances. For the dynamic optimisation, the trajectories of the air path actuators are optimised for a typical acceleration event. Because of the high computationally effort, such an optimisation can not be performed during engine operation, but it enables conclusions about suitable control structures. Thereafter, a smoke limitation based on the soot model is presented. This model based smoke limitation requires no additional calibration effort, but the model parameters are difficult to interpret. Therefore, a simplification to an open loop control structure with look-up tables is shown, which enables a manual fine tuning of the maps. This dissertation contributes to the model based optimisation of engine control functions and presents new modelling and optimisation approaches. Furthermore, new model structures are compared to the in automotive industry well known look-up tables and assets and drawbacks are discussed.
展开▼
