Modellbasierte Optimierung der Steuerung und Regelung eines Automobilmotors

The present work discusses model-based concepts for the control optimization of a spark-igmted engine The concepts are tested on a dynamic engine test bench which is fully Computer controlled The control of an SI engine can be divided into several blocks The stationary Operation requires maps for the spark advance and the metered fuel as a function of speed and load These maps guarantee the fuel-optimal Operation of the engine taking exhaust emission hmitations into account Determimng them requires a high number of engine tests To reduce this bürden, a model-based concept for the evaluation of the fuel-optimal maps is presented, although without emission constraints (The evaluation of fueloptimal maps without emission constraints is always the first Step, even with arbitrary exhaust emission limits) A large number of the engine tests are thus replaced by Computer Simulation In a few reference points certain Parameters are identified For these parameters rules for their Variation dependmg on changes in the variables Operation point, spark advance, and air-to-fuel ratio must be given With a cyclic process calculation the dehvered engine power within a specified area around the reference point can be determined With a maximum search the optimal values of the spark advance and the air-to-fuel ratio are therefore found in the Simulation On the engine test bench, only a few venfication measurements remain and perhaps a shght reoptimization, lf necessary A next block is the dynamic compensation Conventional motor management Systems contain several Controllers which are activated accordmg to the circumstances The model-based multivariable Controller presented here always controls the engine speed and the exhaust air-to-fuel ratio Its mampulated variables are the demanded throttle angle, the spark advance, and the air-to-fuel ratio For the design of this multivanable Controller, different engine modeis with vanous levels of complexity are denved The nonhnear structure used first contains discrete-time as well as continuous-time Subsystems All modeled effects have been venfied by measurements on the engine test bench With discretization and lineanzation, a model is then denved which takes the delays of the System into account This model can be used for the Computer Validation of any Controller designed The model-based multi¬ vanable Controller designed here is based on a model which neglects the