| Rolling bearing is an indispensable key component in mechanical equipment.It is easy to break down in the working process.When the rolling bearing is in the early fault state,under the interference of complex transmission path and background noise,multiple excitation sources and their corresponding characteristic signals are coupled to each other,and then superimposed with weak fault impact characteristics,making it difficult to effectively extract the fault characteristics of rolling bearings.In recent years,various mode decomposition methods have become a research hotspot for bearing fault diagnosis.Dynamic mode decomposition algorithm is driven by data itself,which can highly summarize the spatio-temporal evolution process of complex dynamic systems and extract the characteristic patterns representing the original dynamic system.In this thesis,a series of problems in the application process are studied and corresponding solutions are proposed.The main work of this thesis are as follows:1.The dynamic mode decomposition theory is deeply studied,and describe the transformation process of the original dynamic system information in the process of onedimensional mechanical vibration signal fault diagnosis.Based on the different types simulation dynamic rolling bearing vibration signal,the characteristics of DMD algorithm are studied,and it is found that it has excellent noise reduction ability.However,in the implementation process and practical application of the algorithm,there are some key problems that need to be solved.2.Aiming at the noise sensitivity and the establishment of truncation rank parameter of DMD algorithm,after a systematic comparative study of the information enhancement algorithm,the MOMEDA algorithm is proposed to preprocess the signal,enhance the implied impact information from a global perspective,which is most consistent with the characteristics of the highly summarized dynamic space-time system of DMD algorithm,and then combine with the optimal truncation rank DMD for post-processing.The weak fault information in the original signals is fully extracted,and a complete theoretical system of information enhanced DMD algorithm is constructed.The feasibility of improved DMD algorithm is proved by simulation signal.3.The information enhanced DMD method is applied to the single fault and compound fault signals collected by experiments.Different MOMEDA parameter optimization methods are proposed for the different types of faults.For the single early fault,an adaptive parameter selection method based on joint fitness function and improved particle swarm optimization algorithm was proposed.Aiming at the compound early fault signals situation,in order to solve the problem of information decoupling,a MOMEDA multi-component parameter optimization method based on harmonic significance index combined with variable step search method was proposed,which has higher efficiency.Experimental results show that the proposed method has excellent feature information enhancement effect,decoupling ability of coupling information and feature extraction ability.4.Aiming at the problem of low efficiency and lack of coupling decomposition method using DMD algorithm for multi-channel signals,on the basis of the research in the previous chapters,an information enhancement dynamic mode decomposition multi-source extension algorithm is proposed.The original high-dimensional matrix is projected into low-dimensional space by a Full-QR decomposition and a Thin-QR decomposition,because the matrix has the same orthogonal basis as the original high-dimensional matrix,so the low-dimensional matrix is used to represent the space-time evolution characteristics of the entire mechanical system.and then the common characteristic mode calculated by Koopman theory is mapped back to the original channels to obtain the recovered signals after noise reduction.Experimental studies show that this method can efficiently extract the dynamic characteristics of the early fault signals of each channel. |
PDF Full Text Request