Fault Diagnosis Of Roller Bearings Based On Variational Mode Decomposition And SVM

Rolling bearing is an important part of rotating machinery,because it often works in complex working conditions and harsh environment,it is one of the mechanical parts which are easy to be damaged,once a fault occurs,it will bring significant security risks and economic losses.Therefore,it is very important to study the fault diagnosis method of rolling bearing.The main steps of fault diagnosis of rolling bearing are to collect the bearing operation information,to extract the fault feature from the information,and to diagnose the fault according to the fault characteristics.The feature extraction and state identification are the most important parts.When the bearing is broken,the vibration signal is non-stationary.In this paper,the method of variational mode decomposition(VMD)and support vector machine(SVM)are used to realize the fault diagnosis of rolling bearing.Variational modal decomposition is a new signal processing method,the basic principle of the algorithm is studied in this paper,and its parameters are optimized by Quantum-behaved Particle Swarm optimization(QPSO).It is proved that the variational modal decomposition is superior to the empirical mode decomposition(EMD)in the extraction of fault information by using the simulation signal and the measured signal.On the other hand,the basic principles of support vector machine and its advantages in small sample classification is introduced in this paper.The energy of the modal component processed by the variational modal decomposition is input into the support vector machine,and the fault is identified according to the output result.The validity of the method is verified by the actual signal.Finally,a method to determine the VMD parameters quickly is proposed,it is verified by the experimental data.And a rolling bearing fault diagnosis software with friendly interface based on the Matlab GUI interface,simple operation and high fault diagnosis accuracy is developed and verified by bearing experimental data.The results show that the software is correct and feasible.