Model Based Control with Online Automatic Adaptation by Neural Network for Advanced Diesel Combustion

Model based control with physical models are proposed as an alternative to conventional control methods to improve engine performance under real driving conditions including various transient condition. Even if models are built based on physical rules, the models still have several parameters which is desirable to adapt in real time according to driving condition. Therefore, the authors developed an online automatic adaptation method for model-based control of diesel engines, which is based on neural networks. The predictive accuracy of the adapted model has been evaluated by simulation, and the performance of the feed-forward controller based on the model is evaluated by experiment under actual engine.