Research On Fault Separation And Diagnosis Method Of Rotating Machinery's Multi-source Coupling Signal Under Complicated Conditions

Rotating machinery often works under complex working conditions with high speed and heavy load,which easily causes damage on the surface of the key components.Because of components interaction and other interference factors,multiple source signals tend to couple with each other in the measured signals,making measurement and diagnosis extremely challenging.Therefore,how to timely identify fault characteristics and separate source signals is key to diagnosing failures of rotating machinery.In order to solve this problem,the thesis researches into methods for the separation and diagnosis of multi-source coupling fault signals,which have been analyzed,verified and compared based on bearing vibration data.The main content is as follows:(1)The separation and diagnosis methods for weak coupling fault based on feature enhancement.Aiming at the problem of characteristic aliasing and strong interference when coupling faults occur,the feature enhancement strategy based on signal geometric transformation and information mining is developed,and the separation diagnosis methods for weak coupling fault are proposed.The vibration characteristics of fault signals are analyzed,which excited by different damage parts are emphasized.Aiming at improving the energy distribution of damaged signals,a new framework of multi-source coupling fault diagnosis based on peak geometric transformation is developed by smoothing the peak features of the signal to enhance the energy gathering characteristic.From the perspective of information mining,the characteristics of the rank and sparsity of vibration signal are studied,and the similarities and differences of which under a single damage type are revealed.A new coupling fault diagnosis strategy based on low rank and sparse decomposition is developed to achieve more accurate detection for the weak features.(2)The separation and diagnosis methods for coupling fault based on the sparse strategy in high-dimensional space.Aiming at solving the underdetermined situation caused by the limitation of sensors,and the failure of independent characteristics of signal statistics in specific scenarios,the coupled fault separation diagnosis method based on high-dimensional sparse strategy is studied,under the idea of enriching and improving blind source separation technology and extending the diagnosis approach of coupling faults under complex working conditions.The clustering effect of multi-source sparse signals is analyzed theoretically and experimentally,and the method of separation plane construction in high-dimensional and mixing matrix estimation based on SVM is developed,which has improved the recovery accuracy of coupling fault sources.Continuing the idea of projecting signals into a high-dimensional space to achieve separability,a hyperplane theoretical without considering the kernel function is established,the optimal variational mode extraction based on energy criterion is studied,and a mixing matrix estimation method based on the optimization for normal vector of hyperplane is developed.(3)The step-by-step separation and diagnosis strategy for coupling faults with uncertain number of sources.In view of the dynamic change of the number of sources under complex working conditions and the difficulty in number estimation,a study on the method for the uncertain number of sources is carried out.This thesis studies the theory of blind source separation based on the information criterion,and explores the criterion of negative entropy and mutual information for extracting independent components.The limitations of the traditional estimation method for source number are analyzed to reveal the influence of source number estimation on the results of blind source separation.The research significance of binary independent component extraction under underdetermined situations is summarized,and the internal relation between the separation signals and the sources after adding the update termination constraint is revealed.A step-by-step blind source separation diagnosis method without source number estimation is developed,whose effect on bearing fault data under accelerated degradation process is studied.(4)The separation and detection method of the nonlinear coupling signal under variable speed condition.Aiming at the complex coupling problem of multiple fault sources which may be caused under time-varying speed,a nonlinear decoupling method for the multi-source coupled signal is developed.The basic principle,characteristics and applicable conditions of the nonlinear coupling model are studied from the perspective of signal's coupling mode.Based on the vibration mechanism,the time-frequency characteristics of the vibration signal and the feature extraction method of time-varying FM signal under variable speed conditions are studied.The structure of multi-layer perceptron is improved and optimized,and the nonlinear decoupling model of time-varying signals is established,so as to eliminate the redundant time-frequency components and improve the identification ability of variable FM decomposition.The actual bearing data is used to test and optimize the effect of the method on the vibration signal under constant speed.The influence of different coupling modes on feature detection of variable-speed bearing signals is explored,the detection and diagnosis ability for the actual variable speed multi-source coupling fault signal is improved.The "complex working conditions" concerned in this paper basically belong to the typical and representative conditions in the actual industrial production activities.The research on the targeted methods for the separation and diagnosis of multi-source coupling fault signal is helpful to enhance the feature identification and separation capability of weak coupling signals,overcome the sparse diagnosis problem in the case of underdetermined situation or independent failure,get rid of the dependence on the exact estimation of the sources number in the case of unknown fault source,and solve the nonlinear decoupling problem of time-varying signals under variable speed.The research is very significant,and it can provide the theoretical support and technical means for the engineering application.