Feedback error learning neural networks for air-to-fuel ratio control in SI engines

摘要

A controller is introduced for air-to-fuel ratio management, and the control scheme is based on the feedback error learning method. The controller consists of neural networks with linear feedback controller. The neural networks are radial basis function network (RBFN) that are trained by using the feedback error learning method, and the air-to-fuel ratio is measured from the wide-band oxygen sensor. Because the RBFNs are trained by on-line manner, the controller has adaptation capability, accordingly do not require the calibration effort. The performance of the controller is examined through experiments in transient operation with the engine dynamometer.