Bayesian Fault Diagnosis On Electronic Controlled Fuel Injection System In A Gasoline Engine

This paper describes the diagnosed principle and application of the engine fault self-diagnosis system, and introduces that the engine self-diagnosis system and the traditional fault diagnosis methods have been unable to meet the requirements of the electronically controlled gasoline injection system fault diagnosis. Based on this, This paper describes Bayesian networks development and application, and expatiates on the topology, the mathematical models and the characteristics of Bayesian networks. Through the fault tree and Bayesian networks compared, this paper shows the feasibility of the fault tree translation into Bayesian networks, and gives the translation algorithm.When the electronically controlled gasoline injection system fault, the paper analyzes the four common engine fault: engine not starting, engine lacking power, engine idling bad, and engine accelerating bad, then gives their fault trees. According to the translation algorithm of the fault tree translation into Bayesian networks, the fault trees are translated into Bayesian networks one by one. According to current equipment in laboratory, the fault simulation tests of typical faults have be tested on the electronically controlled engine on Hafei Saima. The experiments find that a number of engine faults are not obvious and these faults can not be diagnosed by the existing instruments of the engine diagnosis. According to the fault simulation tests, the prior probabilities of the nodes what is needed by Bayesian networks fault diagnosis are obtained. Then the paper gives the fault probability of the electronically controlled fuel injection system components when engine idling bad. Finally, fault diagnosis on Hafei Saima proves that Bayesian networks fault diagnosis of the electronically controlled gasoline injection system is convenient, fast, intuitive and high-accuracy.