The Study Of Fault Diagnosis Based On Information Fusion Technology

In recent years,with the gradually improving of the modern automation, the expanding of the system scale and the rapid increasing of the complexity, potential fault also will increase, so it is more and more difficult for us to detect, locate and exclude fault. At the same time the investment is also gradually increasing, thus cause a series of problems. Maintenance cost is increased, breakdown loss is huge and so on. Because a symptom may be caused by a variety of faults and a fault can also show a variety of symptoms. So to diagnose the system rapidly, in time and accurately becomes particularly important.The fault diagnosis system uses a variety of status information and abundant expert knowledge of the diagnosis object system that is running, processes comprehensively the information, and then gets a comprehensive assessment of the system operation status and fault status eventually. In this sense, the fault diagnosis is the typical data fusion processing actually, so the fault diagnosis method based on information fusion technology is an effective method.This paper describes in detail the information fusion technology,the basic principles of fault diagnosis, and the relationship between them, and then present a new fault diagnosis model and method based on information fusion technology. We get conclusions with deviation by different diagnosis methods, and then we further fusion these conclusions by decision level fusion. The final result is more accurate than a single result. On one hand we can process the signal by the method based on data field such as the wavelet analysis method and the cross-correlation analysis method. On the other hand we can process the signal by the method based on analytical model such as the scalar Kalman filter algorithm. Then we fusion the preliminary diagnosis conclusions by decision fusion method, so achieve a more accurate fault diagnosis.The methodology that is given by this paper is used to the fault diagnosis of the oil pipeline system, including pump fault, sensor fault and pipeline leakage fault in this system. We use MATLAB to simulate the fault pressure data acquired from the experimental platform. The results of simulation show that the result by this method is consistent with the actual situation. And show the method is feasibility and the validity. So the information fusion method is better than a single detection method.The work that in this paper is the research of basic theory and the algorithm analysis, it must be further tested, modified and improved if it want to be developed and applied in the actual fault diagnosis system.