Abstract
New computation methods are described to help increase combustion engine performance during the development process and later during the whole service life by model-based control loops. In the early development phase simplified models of the engine are used to predict its performance, such as power output, fuel consumption, gas flow, pressure and temperature curves, before the hardware is manufactured. Injection system parameters, turbocharger matching and transient response can also be varied on the virtual engine . Results that show more details can be obtained from the first test engine applying model-based analysing methods. For example, heat release evaluation has an immense effect on efficiency, stresses as well as noise and exhaust gas emissions. Simple sensors and sophisticated models, implemented on the new generation of engine control units, allow the engine performance to be optimized on-line during operation. These models contain both : mathematical structures like filters or neural networks and physical equations, such as mass or energy balances. They show the way, how to use signals from a noise transducer to get the pressure curve and the heat release function. If there is deviation from the reference heat release function, which is stored in a neural network, engine control parameters can be changed. Thus a good cylinder balancing is obtained to get the same process in each cylinder. All the above methods can reduce development time and costs and increase quality. Therefore a continous optimization of the engine performance becomes possible during operation. The same models monitor the engines health.
This ensures a high level of safety and availability.