Fault Diagnosis Method And Experimental Research Of Centrifugal Fan Blade Based On EMD And SVM

As a kind of rotating machinery for conveying gas,centrifugal fan has been gradually applied to various fields of production and life.The impeller of the centrifugal fan is the most important core component of the fan system,which plays a decisive role in the overall system safety and safety performance.The blade is the core component of the impeller and bears the interaction of multiple forces.The blade fracture is one of the serious hidden troubles in the fan system.At present,the diagnosis of fault fracture degree of centrifugal fan blades and related experimental research are relatively insufficient.This paper takes the centrifugal fan of a high-speed train cooling system as the research object,and carries out the blade fault diagnosis and experimental research from the practical point of view.The main research contents of this paper are as follows:(1)In view of the insufficiency of fault data of centrifugal fan blades,this paper presents a method of combining finite element simulation with experiment to find out the failure modes of blades sequentially and carry out the failure experiments of blade fracture degree,so as to build the database of blade failure.The position of fatigue fracture of the blade is derived by mutual verification of the static analysis of the centrifugal fan,the static strength finite element analysis,the test data of the stressstrain test bench and the real blade fracture samples.According to the derived fracture location,the samples are simulated in four states with different degrees of fracture.Then we build a vibration signal acquisition test bench for the centrifugal fan.The vibration signals of four statuses of the blade at rated rotating speed are collected from six measuring points on the volute,the bearing base and the pedestal base respectively.And the fault database of the blade is established.(2)In this paper,the fault characteristics of centrifugal fan blades are extracted by Empirical Mode Decomposition(EMD).Firstly,according to the vibration signal of the test,the time domain and frequency domain analysis are carried out,and the vibration measuring points which can better reflect the fault characteristics are initially selected.Then the vibration signal is decomposed by empirical mode decomposition method.The effective Intrinsic Mode Function(IMF)component is selected by energy distribution,correlation coefficient method and K-L divergence method.The kurtosis,instantaneous energy ratio,complexity entropy and fine generalized multi-scale entropy of the effective IMF component are extracted to form the fault eigenvector of each measurement point.At the same time,the discrimination of the fine generalized multi-scale entropy of the IMF component is verified by comparing the multi-scale entropy,generalized multi-scale entropy and fine generalized multi-scale entropy between the original signal and the IMF component,.Furthermore,the optimal vibration measurement points are obtained by optimizing the vibration measurement points based on the principle of minimum redundancy,which lays a foundation for the pattern recognition of centrifugal fan blade failure.(3)The main parameters of the Support Vector Machine(SVM)are optimized by the particle swarm optimization(PSO)algorithm.Then the SVM is used to diagnose the blade fault of centrifugal fan.The feature vector of the optimal point is used as the input vector of the pattern recognition.The optimal support vector machine model is obtained by selecting the kernel function and optimizing the parameters,so the recognition accuracy reaches 97.5%.At the same time,the effectiveness of the preferred method of measuring points is verified by sample points.The fault feature vector composed of multiple features is compared with a single fault feature input.And the optimized SVM is compared with the standard SVM and BP neural network respectively to identify the advantages of the optimization model.It also shows the application value and feasibility of the centrifugal fan fault diagnosis technology.The method described in this paper can provide reference and guidance for fault diagnosis and experimental research of centrifugal fan,and is of great significance for the safe operation of centrifugal fan and the safe operation of high-speed train cooling system.