Transient A/F Estimation and Control Using a Neural Network 970619

A new estimator for IC engine A/F ratio is described. A/F ratio is important for engine operation since it determines the quantities of engine emissions, such as HC, CO, NOx, the conversion efficiency of catalyst systems, and the engine combustion stability. The A/F ratio estimator described in this paper is based on a fundamental metric that relies on inducing and detecting crankshaft speed fluctuations caused by modulating the engine's fuel injection pulse widths. Fuel pulse width modulation varies the instantaneous combustion A/F ratio crankshaft velocity. Synchronous measurement of crankshaft velocity provides a metric that, when used with other engine state variables as inputs to a conventional neural network, can accurately estimate A/F ratio. The estimator provides A/F information when a physical sensor is not available. This paper discusses the design of the A/F ratio neural network estimator, the transient performance during cold start operation, and the transient performance of a closed loop A/F ratio control system using the estimator for A/F ratio feedback.