A Method for Learning the Varying Parameters of Gasoline Engine Control Systems Based on δ-Rule

Abstract

This paper proposes a method for learning the varying parameters of the gasoline engine control system. The method is based on the δ-rule which is a basic learning law in neural networks. Since parameters change frequently according to engine running conditions, they are not directly learned. Instead, the data of a table which stores parameter values of various engine running conditions are learned based on the δ-rule. This enables frequently varying parameters to be learned. A simulation shows that the proposed method is effective to achieve accurate control in a comparatively short period of time.