Research On Early Weak Fault Diagnosis Of Rolling Bearing Based On Variational Mode Decomposition

In the field of mechanical failure diagnosis,rolling bearings are always considered by many scholars as the focus of research.When the rolling bearings of mechanical equipment are broken down,the fault signals are often presented with non-stationary and nonlinear characteristics.Therefore,how to extract effective fault signatures from these fault status signals is the key to mechanical equipment fault diagnosis.In recent years,variational mode decomposition(VMD)proposed as a new signal processing method,it not only overcomes the shortcomings of other time-frequency analysis methods,but also can accurately decompose the state signals to obtain the signal time-domain and frequency-domain local information.Therefore,the paper applies it to the feature extraction of rolling bearings and the original fault feature set feature component of the signal is constructed.Because the fault feature set extracted from the signal components often contains a large amount of redundant information,they interfere with each other and cause confusion in the model.At the same time,it may also lead to the occurrence of “dimension disaster”problems and seriously affect the fault recognition effect.Therefore,it is necessary to perform dimensional optimization on complex and high-dimensional signal feature sets to achieve sensitive feature extraction.But,the methods of data dimension reduction have the ability to reduce the number of dimensions and extract effective features.Therefore,Combining the advantages of variational mode decomposition(VMD),the method of variational modal decomposition(VMD)combined with local tangent space alignment algorithm(LTSA)is proposed in this paper for fault diagnosis of rolling bearings.Firstly,the variational modal decomposition algorithm was used to decompose the vibration signal of rolling bearing under different conditions and some components which were most relevant to the original signals sifted out by selecting the instantaneous frequency mean and plotting the characteristic curve.Then,the time domain of vibration signal and the frequency band decomposition of the wavelet packet energy of frequency index were exracted from effective components.After forming a high-dimensional fault feature,LTSA was used to get sensitive extraction of high-dimensional features.Finally,the extracted features were evaluated by inputting them into the K-means classifier.The main contents of the subject are as follows:(1)The cylindrical roller bearing was taken as the research object.Firstly,the SG bimetal pitting machine was used for the rolling bearing to simulate the ball fault,inner fault,and themix of ball fault and the inner ring for pitting failure.Then it was installed on the BVT-5bearing vibration tester.Finally,the four state vibration signals of the rolling bearing model were collected by INV-1618 C of the Institute of Vibration and Noise of the East.(2)The concept of Variational Mode Decomposition(VMD)is introduced.the importance of selecting modal number K in VMD method is researched.and some components which were most relevant to the original signals were sifted out by selecting the instantaneous frequency mean and plotting the characteristic curve.and this method is verified by simulation data.(3)The method of based on VMD+ time-frequency domain is proposed,and the method is compared with the time-frequency domain and EMD+time-frequency fault feature extraction methods.The experimental results show that the VMD+ time-frequency domain method is effective.And no parameters need to be set.(4)Through the comparative analysis of the fault diagnosis of deep groove ball bearings and cylindrical roller bearings,it was found that: Compared with time frequency feature+LTSA?EMD+LTSA methods,the method of VMD+LTSA has more advantages on classification effect and recognition accuracy;LTSA algorithm has more advantages to extract the sensitive fault feature after reducing dimensions than these five algorithms: PCA?LPP?LE?ISOMAP and LLE.The results show that the proposed method has some advantages in the fault diagnosis of deep groove ball bearings and cylindrical roller bearings.