Study On Mechanical Fault Diagnosis Of G4-73No8D Centrifugal Fan

The fan is diffusely used in many fields，such as the chemical industry,petroleum industry, metallurgical industry, power industry and so on, which belongsto general-purpose machine. Therefore, it has great significance for the progress ofnational economy. The fan is one type of large rotating machines in power plant,which also is the power supply of the air/gas system. It’s easy to malfunctionbecause that its operation process will produce strong vibrations and the huge noise.Once it fails, may cause the unit stop resulting in serious economic loss, what’smore, even in some cases, the failure can trigger the environmental pollution, thedamage of personal safety and other more serious consequences. Therefore, researchon fault diagnosis of fan is great significance to ensure the security and economicoperation of power plant.In this paper, the vibration signals are acquired by fan fault simulation test. Inorder to accurately diagnose the mechanical failure of fan, the different featureextraction methods of the vibration signals are fully carried on the deep research.The main contents of this paper are as follows:(1) The diagnosis method of fan machinery fault based on the complexityanalysis of vibration signals is proposed in this paper. The sample entropy featureand the symbolic dynamics entropy feature of vibration signals are respectivelyextracted. The improved BP neural network with additional momentum and adaptiveadjusting learning rate is used to respectively establish the fan machinery faultdiagnosis model corresponding to the sample entropy feature vector sets andsymbolic dynamics entropy feature vector sets, and the accuracy of two models aretested.(2) The diagnosis method of fan machinery fault based on the wavelet packetanalysis of vibration signals is proposed in this paper. The wavelet packet energyfeature and the wavelet packet singular value feature of vibration signals arerespectively extracted. The improved BP neural network is used to respectivelyestablish the fan machinery fault diagnosis model corresponding to the sampleentropy feature vector sets and symbolic dynamics entropy feature vector sets, andthe accuracy of the models are tested.(3) A new diagnosis method of fan machinery fault based on the symmetrizeddot pattern (SDP) analysis of vibration signal is proposed in this paper, for that theSDP graphics corresponding to different fan running states have obvious differences.The vibration signals can be converted to graphics by the SDP analysis, and then the SDP graphics of vibration signals are transformed into numerical matrix. For thenext step, the similarity degree among SDP graphics of vibration signals is analyzedby the method of correlation coefficient. That is, the operation state of fancorresponding to test signal can be diagnosed by comparing the similarity degreesbetween the SDP diagram of test signal and the SDP graphic models correspondingto all kinds of operation states of fan.