Research On New Method Of Volterra Kernel Function Feature Extraction For Rolling Bearing Fault State

At present,in the field of mechanical equipment fault diagnosis,there are many signal processing methods to extract and monitor the fault of mechanical equipment,such as wavelet transform,EMD decomposition and spectral analysis only on the widely used fault diagnosis method based on signal processing in engineering.Most of the methods are to extract the characteristic information from the vibration signals in the equipment,too much noise,phase difference of transmission path,frequency nonlinear coupling and other unavoidable practical engineering applications are introduced inevitably.Therefore,a new method of fault feature extraction is urgently needed to avoid these problems.To solve these problems,widely used rolling bearing parts were taken in the actual production equipment as an example,the method of fault feature extraction of rolling bearing based on Volterra kernel function was put forward,and the input vibration signal of the system was introduced into the fault diagnosis algorithm.Firstly,the Volterra series mode was established by using the input and output signals of the rolling bearing.Secondly,the feasibility of the method was analyzed theoretically on the basis of the dynamic model of the rolling bearing.Finally,the Volterra kernel function method was proved to be used to extract the fault feature of the rolling bearing.The main contents are divided as follows:(1)According to the deep groove ball bearing as an example,the dynamic model of rolling bearing failure was established,and the model was solved by Runge-Kutta method.The time domain waveform and frequency spectrum diagram of its displacement and velocity were analyzed and compared with the theoretical characteristic frequency,the correctness of the established dynamic model was verified and to lay the foundation for further research and analysis.(2)The basic content of Volterra series theory was studied.The difference of its expression form was analyzed in time and frequency domain.The method of multi pulse excitation was used to solve the low order Volterra kernel function.The correctness of the multi pulse excitation method was verified by an example simulation,and the concepts of high order spectrum and slice spectrum were expounded.Theoretically,the validity of the high-order spectrum and slice spectrum method in the characterization of Volterra kernel was analyzed.(3)The difference between the expression of Volterra kernel function under different truncation forms and different memory lengths was analyzed.The difference between the original output signal and the system output signal expressed by the Volterra kernel was compared through the dynamic model of rolling bearing,and the effect of the truncation form and the memory length on the solution of the kernel function were explained.So that the order of kernel function and memory length could be correctly selected to achieve the best identification effect.(4)The Volterra kernel function was solved by using the data of the sampled rolling bearing fault experiment table,and the second order and third order kernel functions of the rolling bearing Volterra with different fault degrees and positions were directly characterized by the bispectral and trispectral slice method.The frequency information of the inner ring,outer ring and rolling element of the rolling bearing were quantitatively extracted.On the basis,the fault feature information was normalized by kernel assurance criterion and compared with the existing normalization index.Finally,the superiority of the Volterra kernel function method was verified from the experimental point of view.