Abstract
The combustion process of the latest Diesel Engine became able to use multistage fuel injection time and quantity of jet as a control input, and the possibility of the control performance enhancement increased markedly. However, under the present conditions, the static maps are made using DOE (Design of experiments) design which needs time, cost, and, hands, and it is mainstream to perform the feedforward control by the static map. By the control by such a static map, it is difficult to realize the most suitable engine control depending on a driver and environment. In this study feedback error learning (FEL), and let indicate online using a technique of AI (artificial intelligence), and adapt and/or learn a feedforward controller, the control system which can learn to a driver and environment positively. In order to confirm the effectiveness of designed control algorithm, we make the numerical simulation based on the Tokyo University discrete-time model and the experiment with the Tokyo University engine bench.
