Research On Feature Extraction And State Recognition Of Vibration Signal Of DC Servo Motor

DC motor plays an important role in modern industrial production.It is widely used in many research fields,such as machinery,chemical industry,materials and so on.If a part of the motor itself is damaged,the whole operation system will be paralyzed,and the factory property and personal safety will not be guaranteed.In view of the above reasons,how to ensure the stable operation of DC motor is particularly important.The research shows that the vibration signal generated by the operation of DC motor contains rich fault information,in-depth study of the vibration signal generated by the early fault of the equipment can accurately predict the fault location,effectively identify the fault type,and repair the equipment according to the corresponding fault type.It is of great significance to ensure the safe and reliable operation of the system and reduce the loss of industrial assets.Most of the vibration signals of DC motor are continuous,nonstationary and the noise background is too large,which leads to the rich fault information contained in the signals can not be directly reflected.In recent years,experts have proposed empirical mode decomposition(EMD,1998),aggregate empirical mode decomposition(EEMD,2009),empirical wavelet transform(EWT,2013),variational mode decomposition(VMD,2014),nonlinear mode decomposition(NMD,2015),adaptive local iterative filtering(ALIF),2016)etc.Although these algorithms can extract the fault information of vibration signal in the process of decomposition,these methods have some defects that can not be solved.Based on the above research,this paper studies the feature extraction and state recognition of inv1618 vibration signal of DC motor.Firstly,the existing decomposition algorithms EMD,EEMD,EWT and VMD are discussed.Compared with other traditional methods,VMD can extract fault information better,but there is a problem that parameters are not suitable for determination.In this paper,a VMD method based on multi-scale permutation entropy(MPE)is proposed to optimize VMD,and the preset threshold method is used to determine the VMD modal parameter k.by comparing with the traditional VMD method,the mpe-vmd algorithm is proposed to avoid the phenomenon of modal overlap or over decomposition.Then,in order to improve the signal-to-noise ratio,the method of adding Gaussian white noise to the signal by reference to ceemd is used to add positive and negative white noise to the signal to be decomposed.In this paper,the positive and negative white noise is added to the signal to be decomposed.Through simulation and experimental analysis,it can be seen that the positive and negative Gaussian noise has significant effect in eliminating signal reconstruction error and background noise and promoting noise neutralization.Finally,in view of the lack of early fault samples of DC motor,this paper proposes an optimized support vector machine based on cuckoo search algorithm.Compared with particle swarm optimization(PSO),grid algorithm(GS)and genetic algorithm(GA),the cuckoo algorithm is not easy to fall into local minimum and search for global optimum.This method classifies and learns different kernel functions to find the global optimal kernel function.The experimental data shows that this algorithm overcomes the error caused by the uncertainty of parameter selection of support vector machine,and has the advantages of strong generalization ability and good model learning ability,it has strong stability and generalization for the early fault classification of DC motor Broad capabilities.