The Machinery Fault Diagnosis Method Based On Adaptive And Sparsest Time-Frequency Analysis

Condition monitoring and fault diagnosis are important for ensuring the safety of equipment.With the development of science and technology,modern machinery equipment is becoming more and more complex.Therefore,the study of the new technologies and methods for the mechanical fault diagnosis has an important significance.The core of machinery fault diagnosis technology is extracting the fault characteristics.The vibration signals generated by the operation of the equipment contain a large number of condition information.Therefore,the machinery fault diagnosis based on the vibration is effective.However,the mechanical vibration signals are usually non-stationary,multi-component,multi-modulation and often disturbed by noise.Therefore,the study of a suitable signal processing method which can accurately extract fault features is critical.Adaptive time-frequency signal analysis methods select basis functions or arguments automatically according to own characteristic of the signal in decomposition.They can extract the essential characteristics of a mechanical fault vibration signal effectively.Therefore,they have been widely used in machinery fault diagnosis.Adaptive and sparsest time-frequency analysis(ASTFA)is a new adaptive time-frequency analysis method,which takes the least number of components as the optimization target and takes the instantaneous frequency of the component has the physical meaning as the constraint condition.The ASTFA method decomposes the signal into the sum of several intrinsic model functions by solving the optimization problem.The ASTFA method combines the advantages of empirical mode decomposition(EMD),which decomposes the signal into the sum of several intrinsic model functions,and the sparse decomposition which obtains the signal's sparse decomposition by optimization in the over-complete dictionary.Supported by National Natural Science Foundation(No.51375152),the theory of ASTFA method is investigated and improved,and the ASTFA method is applied to the machinery fault diagnosis in this dissertation.The main research of this dissertation is as follows:(1)The ASTFA method is introduced and investigated.The ASTFA method is compared with the EMD,local characteristic-scale decomposition(LCD)and local mean decomposition(LMD).The index of orthogonal and accuracy of the decomposition is superior to EMD,LCD,LMD.The decomposition ability of the ASTFA is examined.The ability of avoiding model mixing is investigated.The simulation and experimental analysis shows that the ASTFA method has the obvious superiority in avoiding model mixing.(2)The theoretical explanation of the ASTFA is investigated from the perspective of the instantaneous frequency calculation methods.The ASTFA and EMD,LCD,LMD have common theory.They all decompose the complex signal into a single component which has AM and FM product form adaptively.However,the ASTFA can calculate the instantaneous frequency directly,opponent to the instantaneous frequency calculation based on the pure FM signal obtained by normalization.Therefore,the ASTFA method can avoid fluctuations and estimation error attached to the extreme points.The ASTFA method has the obvious superiority.(3)The explanation of the ASTFA is addressed from the perspective of the filter design.The ASTFA method translates the signal from time domain to phase domain and designs an adaptive filter based on the phase function of the signal itself achieving the adaptive decomposition.The ASTFA method designs the filter which meet the requirements by selecting the appropriate initial phase function and smoothness control parameters or filtering the original signal in order to achieve the decomposition of the signal.(4)Due to the selection of the initial phase function of the ASTFA method,the ASTFA method based on one-dimensional refine search and the ASTFA method based on genetic algorithm(GA)are proposed.The two improved methods are proposed to solve the selection of the initial phase function.(5)The application of the ASTFA to gear fault diagnosis is studied.The instantaneous amplitude spectrum and instantaneous frequency spectrum method on the ASTFA is raised.The partial mean multi-scale fuzzy entropy(PMMFE)method for gear fault diagnosis based on the ASTFA is proposed.The application of the ASTFA to gearbox complex fault diagnosis and planetary gearbox fault diagnosis is investigated.(6)The application of the ASTFA to mechanical malfunction diagnosis in shifting conditions is suggested.The order tracking method based on the ASTFA is considered.Simulation analysis and experimental analysis shows the order analysis method based on the ASTFA can accurately extract the fault feature information of the gear.The generalized demodulation method based on the ASTFA is employed.The demodulation phase function needed by the generalized demodulation method can be provided by the ASTFA.The ASTFA method solves the selection of the demodulation phase function of the generalized demodulation method.Simulation and experimental analysis shows that the method is very suitable for processing multi-component frequency modulated signal,and can effectively extract the damage characteristic information of the rolling bearing damage in shifting conditions.