A fine control of the air-to-fuel ratio with recurrent neural networks

A fine control of the air-to-fuel ratio is a fundamental issue to minimise exhaust emissions in automotive fuel injection systems. Traditional approaches have limited effectiveness since the air-to-fuel ratio is sensitive to small engine perturbations, some parts of the combustion process are unknown and some others are nonlinear. In this paper we introduce a direct neural-based control scheme which results in a performance obtainable with more classic approaches based on transient fuel film compensation.