Neural Network Parameter Adaptation for a Fuel Injection Control System

When the dynamic sliding mode control (DSMC) is applied to maintain the stoichiometric value of air-fuel ratio for automobile engines, the model-plant mismatch and some time-varying parameters cause negative influence on the control performance. This paper proposes a neural network parameter adaptation method for two immeasurable control parameters and to compensate the model uncertainty, so that the air-fuel ratio is regulated within the desired range. The adaptive law of the neural network is derived using the Lyapunov method, thus the stability of the whole system and the convergence of the networks are guaranteed. Since the model-plant mismatch caused by mechanical wear of parts and batch error in production are compensated and no initial values needed, the proposed technique has a strong potential to find industrial applications. Computer simulations based on a mean value engine model show the effectiveness of the technique