A Research on I.C. Engine Misfire Fault Diagnosis Based on Rough Sets Theory and Neural Network

Abstract

A method for diagnosis of misfire fault in internal combustion engine based on exhaust density of HC, CO2, O2 and the engine’s work parameters are presented in this paper. Rough sets theory is used to simplify attribute parameter reflecting exhaust emission and conditions of internal combustion engine and in which unnecessary properties are eliminated. The engine’s work parameters, exhaust emission with misfire fault and without fault are tested by the experimentation of CA6100 engine. A diagnosis model which describing the relationship between the misfire degree and the internal combustion engine’s exhaust emission and work parameters is established based on rough sets theory and RBF neural network. The model reduces the sample size, optimizes the neural network, increase the diagnosis correctness. The model is also trained by test data and MATLAB software. The model has been used to diagnosis internal combustion engine misfire fault, the result illustrates that this diagnosis model is suitable. This system can reduce input node number and overcome some shortcomings, such as neural network scale is too large and the rate of classification is slow.