Research On Rolling Bearing Fault Diagnosis Method Based On Adaptive Mode Decomposition

The efficient,accurate and simple detection of bearing faults has been a popular research topic in the manufacturing sector.The extraction of fault characteristics from vibration signals is a crucial step in vibration analysis methods and is a prerequisite for condition identification and fault diagnosis.However,traditional time-frequency domain analysis methods are inadequate to handle the non-linear and non-smooth vibration signals in practical operating conditions,thereby resulting in an inaccurate evaluation of the bearing’s operating condition.This paper focuses on the exploration of rolling bearing fault methods for challenging problems such as strong background noise and early faint faults by introducing signal adaptive mode decomposition techniques.Furthermore,this study investigates the construction of an intelligent diagnosis method that effectively combines the adaptive mode decomposition method with the pattern recognition method.The specific research content is as follows.(1)In the context of rolling bearing vibration signal fault diagnosis under strong background noise,selecting sensitive modal components and setting proper parameters for the Variation mode decomposition(VMD)method is a challenging task.To address this issue,this paper presents an adaptive fault diagnosis method based on the Marine predator algorithm(MPA).The proposed method involves adaptively selecting the VMD parameters using the MPA,filtering sensitive modal components using the Gini Index of Squared Envelope(GISE),and performing envelope demodulation analysis.The effectiveness of the improved method is validated using two bearing datasets.(2)In this paper,a novel approach is proposed to address the challenge of extracting fault features from rolling bearing vibration signals in the early fault stage,where weak fault features and severe background noise interference pose a significant obstacle.The proposed approach combines the component-weighted symplectic singular mode decomposition(CWSSMD)and 1.5-dimensional envelope derivative energy operator(1.5D-EDEO)demodulation,and features an adaptive process that eliminates the need for manual parameter setting.Specifically,the original vibration signal is first decomposed using the symplectic singular mode decomposition method to obtain multiple initial symplectic singular components.Then,the fault impulse sparsity(FIS)is used to measure the amount of fault information in the ISSC,which is subsequently weighted and reconstructed to obtain the final denoised symplectic singular component.Finally,the denoised symplectic singular component is demodulated and analyzed using 1.5D-EDEO to further enhance the fault characteristics of the bearing and reduce noise interference,thereby enabling accurate early bearing fault diagnosis.(3)To meet the requirements of intelligent and real-time bearing fault diagnosis in practical industrial systems,a bearing fault classification and identification method based on CWSSMD and logistic chaotic mapping extreme learning machine(Log ELM)is proposed.The method leverages the noise reduction ability of CWSSMD and the fast classification performance of Log ELM to provide a reliable solution with high accuracy and real-time performance.The proposed method has been experimentally proven to be excellent in terms of diagnostic accuracy and efficiency.