Application Of Nonlinear Mode Decomposition In Mechanical Fault Diagnosis

This dissertation is supported by the Project supported by the National Natural Science Foundation of China(Grant No.52075236),the Project supported by the Natural Science Foundation of Jiangxi Province,China(Grant No.20212ACB202005),the Project supported by the equipment Pre-Research Foundation of China(Grant No.6142003190210)and the Project supported by the Aviation Science Foundation(Grant No.201946030001).Based on the advantages of blind source separation and adaptive processing method,this method is introduced into fault signal processing and proposes a method based on the Application of nonlinear mode decomposition in mechanical fault diagnosis.The proposed method is deeply studied,and has achieved some innovative results through simulation and experimental verification.Each chapter are as follows.Chapter 1: Discusses the background and research significance of the topic,and discusses the signal decomposition theory and nonlinear mode decomposition methods.On this basis,the main contents and innovation points of each chapter of this paper are introduced.Chapter 2: The basic principle of Nonlinear Mode Decomposition(NMD)is discussed.This method is an adaptive selection method based on the combination of wavelet transform and window Fourier transform.Then the maximum path function method is used to extract the ridge line,and then the ridge method and direct method are used to reconstruct the signal.The reconstruction results have practical physical significance and can better solve the frequency aliasing problem.Then,the superiority of the nonlinear mode decomposition is compared with the empirical mode decomposition and the variational mode decomposition.Finally,the proposed method is applied to the bearing fault signal processing,and the experimental results verify the effectiveness of the proposed method.Chapter 3:In order to solve the problem of too many iterations and slow running speed of the nonlinear modal decomposition method,a nonlinear modal decomposition-probabilistic amplitude demodulation method is proposed and applied to the bearing fault diagnosis research.The method can adaptively adjust the frequency range of harmonics,reduce the number of iterations of wavelet transform and windowed Fourier transform in time-frequency representation,speed up the running speed,and then perform probability amplitude demodulation on the reconstructed components.The simulation results show that this method can reduce the number of iterations,speed up the running speed,and can extract signal features.Finally,the proposed method is applied to the bearing composite fault signal processing,and the experimental results further verify that this method can be used in practical working conditions.Chapter 4: In order to solve the shortcomings of time-varying and strong noise signals in the processing of nonlinear modal decomposition,synchronous compression is embedded in nonlinear modal decomposition,and the synchronous compression nonlinear modal decomposition method is proposed and applied to bearing fault diagnosis.The proposed method can suppress the spectral leakage caused by the windowed Fourier transform and the wavelet transform in the nonlinear modal decomposition method,which is conducive to energy concentration and improves the temporal resolution and frequency resolution of the frequency domain signal.Simulation results show that the proposed method is obviously better than the traditional nonlinear mode decomposition can effectively handle the time-varying signal of strong noise interference.Finally,the proposed method is applied to the bearing composite fault signal processing,and the experimental results demonstrate the practicability of the synchronous compression nonlinear modal decomposition method.Chapter 5: Combining nonlinear modal decomposition and attention convolutional neural network,an intelligent diagnosis method based on nonlinear modal decomposition is proposed and applied to bearing fault diagnosis.The method first extracts high-dimensional characteristic modulus from the vibration signal of bearing fault by nonlinear modal decomposition;then,the characteristic modulus is composed of multi-channel samples and input into the attention convolutional neural network for training.The trained network fuses the feature moduli and adaptively selects the feature to improve the self-adaptability of fault diagnosis;finally,the trained network is used to diagnose the fault of the rolling bearing,and at the same time,it is based on empirical mode decomposition,variational A comparative study of attentional convolutional neural network diagnostic methods for modal decomposition is carried out,which verifies the superiority of the proposed method.Chapter 6: Mechanical fault diagnosis methods based on nonlinear mode decomposition are summarized,and the problems that deserve further study in the field of mechanical failure are discussed.