NEURAL NETWORKS FOR MODELLING AND CONTROLLING FUTURE LOW TEMPERATURE COMBUSTION TECHNOLOGIES

Modern internal combustion engines with a low peak combustion temperature suffer from instabilities of the process and a highly nonlinear behaviour. These make a closed loop control a necessity. In order to build and tune a controller a model is needed, which has to be able to reproduce the nonlinear behaviour. The paper presents the application of offline trained NNSSIF nets, a neural networks architecture with state space attributes. These are combined with an extended Kalman filter and a nonlinear model-based predictive controller to a research internal combustion engine.