Study On Fault Diagnosis System Of EFI Engine Based On Neural Network &Virtual Instrument

Engine is the heart of an automobile, and 40 percent of the car troubles come from this assembly. During the long-time utilization, for its complex structures and poor working conditions, the technical parameters of the engine will be changed by different laws and different intensity, which eventually led to failure, and bring a serious threat to traffic safety. So doing the research on fault diagnosis of engine have a very important and realistic significance. However, the current engine fault diagnosis systems still lack of a common platform, which integrates with data acquisition, data processing and diagnosis reasoning. At the same time, there are also some limitations when analyzing the nonlinear relationship between fault phenomena and fault causes.Therefore, based on study and comparing with the various diagnosis technologies, the neural network theory and virtual instrument technology are applied to EFI engine fault diagnosis. The results show that the reliability of the engine running is improved and the significant economic benefits are received by guaranteeing the engine performance, reducing spare parts and shortening maintenance time.Firstly, the basic principle, the model structure and the algorithm design of BP neural network are introduced in this paper; some improved algorithms to BP neural network and its training effect are studied either. The feasibility is certified by simulation which applying BP neural network into fault diagnosis. In addition, an assessment and forecasting model of EFI engine performance is created with data fusion technology, to solve the problem that training procedure can't convergent when the number of samples is excessive. The model is validated by examples such as idle speed unstable fault and start-up difficult fault.Then the basic concept, the composition and the LABVIEW programming technology of virtual instrument are introduced; a universal platform for the intelligent fault diagnosis system is designed with the research on parameters which characterize EFI engine performance. The platform combines with signal acquisition technology, signal processing technology, database technology and other intelligent fault diagnosis technology including the new fault diagnosis model in this paper.Finally, taking the emission test as an example, a fault diagnosis subsystem for EFI engine emission based on neural network theory and virtual instrument technology is designed. This subsystem overcomes many difficulties such as the data acquisition and storage, the signal analyzing and diagnostic reasoning of neural network. Besides, the subsystem also achieves analysis functions on-line, improves the diagnostic accuracy and speed up the development progress of the system.