On-board monitoring and diagnosis for spark ignition engine air path via adaptive neural networks

Abstract

This paper presents a new method for on-board fault diagnosis for the air path of spark ignition (SI) engines. The method uses a radial basis function (RBF) neural network to classify predefined possible faults from engine measurements, reporting fault occurrence as well as the type and size of a fault. After diagnosing faults in each sample interval, the weights and widths of the RBF fault classifier are updated with the measurements and appropriately selected target outputs. Consequently, the network can adapt to the time-varying dynamics of the engine and environment change so that the false alarm rate is greatly reduced and the required network size is also reduced. The developed scheme is assessed with various faults simulated on a mean value engine benchmark model and compared with a fixed-parameter RBF classifier. Simulation results demonstrate the effectiveness of the proposed method.