Research Application Of Information Fusion Technology In High Velocity Ram Machine Fault Diagnosis

In fault diagnosis practice, the researcher discover.(1)The diagnosis result basing on different sensor is conflictive sometimes.(2)The diagnosis result basing on different characteristic region is conflictive sometimes.(3)The diagnosis result basing on different reasoning method is conflictive sometimes. Because the structure of large scale equipment is complex and the running conditions are various, a mass of uncertainty is imported. So the accuracy and reliability of the fault diagnosis is very low, and the current method can not to meet the diagnosis needs. For decreasing uncertainty and improving accuracy of fault diagnosis, it is very necessarily to research the application of information fusion technology in high velocity ram machine fault diagnosis.This paper explores the application of information fusion technology in high velocity ram machine fault vibration-diagnosis form theory and practice. The multi-sensor signal, multi-characteristic region information and multi-method of fault diagnosis are integrative used. The uncertainty of fault diagnosis can reduce to minimum, so the equipment diagnosis is comprehensive and veracious. The chief research as follow:(1)By researching the information fusion technology and analyzing the uncertainty of fault diagnosis process, this paper construct the fusion diagnosis frame and establish the fusion diagnosis principle, uncertainties in fault diagnosis could be canceled out with each other to the utmost extent. So the uncertainty of fault diagnosis could be reduced in theory, and the precision diagnosis come true.(2)The principal component analysis (PCA) can treat with the linear problem, the kernel function can transform low-dimension nonlinear problem to high-dimension linear problem, so the kernel principal component analysis (KPCA) composed of PCA and kernel function can effective treat with the nonlinear problem. It’s applied to compress and pick-up the fault characteristic of mechanical equipment, the experiment proves highly effective. So the puzzle of the information overfull and redundant in multi-source information fault fusion-diagnosis is settled well.(3)The specific application method of neural network in fault diagnosis is generalized. Moreover the experimental research finds that the combination of KPCA and neural network can reduce network configuration and mitigate diagnosis calculation, so the accuracy of fault diagnosis is improved.(4)By combining the evidential theory and the weighted concept, the weighted evidential theory come into being. It makes all weighted evidence assembled to objectively embody the fact that the diverse reliability between different characteristic regions and different fault diagnoses. So it covers the shortage of the evidential theory, and lays theoretical basis for the application of the evidential theory in fault fusion diagnosis.(5)In order to make the decision fusion based on multi-regional diagnosis more effective, this paper, based on the evidence theory improved by the weighted concept, by constructing the fusion diagnosis frame and keep to the fusion diagnosis principle established in the2nd chapter, proposes a fault fusion-diagnosis method of the weighted evidential theory.Finally, an experimental analysis based on the lth high velocity ram machine in Shenyang Mint is run. There are three regional diagnosis form three independent characteristic regions:time-domain, frequency-domain and Axle Center Trail, And the diagnostic results of three characteristic regions are fusion-analyzed to make decision. The experimental result shows:the reliability of fault diagnosis upon the multi-fault characteristic information fusion is improved evidently and its uncertainty decreased remarkably as well. It proves that the fault fusion-diagnosis method is very effectively. Furthermore, the method is opening and easy implement, so its value in engineering application is hugely.