Adaptive And Sparsest Time-frequency Analysis Method And Its Applications To Fault Diagnosis For Rotating Machinery

Researches on the fault diagnosis for rotating machinery have great significance,whose key step is the extraction of fault information.However,vibration signals of the rotating machinery are mostly non-stationary.Consequently,extracting state characteristics from complicated vibration signals by an appropriate signal analysis method is always a key and hot research.As the recognized analysis processing method for non-stationary signals,time-frequency analysis method has been applied in many fields.Adaptive and sparsest time-frequency analysis(ASTFA)method,which is a new adaptive analysis method for non-stationary signals,is inspired by empirical mode decomposition(EMD)and matching pursuit(MP).The main idea of ASTFA is searching for the sparsest decomposition over a highly redundant dictionary which consists of intrinsic mode functions.Thusly,signal decomposition problem is translated into an optimization problem and the adaptive decomposition is achieved during the optimization process.Comparing with EMD,fitting the envelop of the extremas is no longer necessary in ASTFA,thus ASTFA is superior in mode mixing and end effect.Comparing with MP,ASTFA has an advantage over the adaptivity and the physical meaning of the components.Funded by the Natural Science Foundation of China(No.51375152),this paper further studies and improves ASTF-A by solving its existed problems.And a new adaptive decomposition method,adaptive sparsest narrow-band decomposition(ASNBD)method is proposed and applied to the fault diagnosis of rotating machinery.The main researches and innovations of this paper are shown as below.(1)In order to solving the disadvantages of the setting of initial values existed in ASTFA method,adaptive and sparsest time-frequency analysis based on optimized initial values(ASTFA-OIV)method is proposed.The optimized initial values are obtained by searching the initial values by reducing the resolution ratio,so that the decomposition ability of ASTFA is improved.The analysis results of simulation signals and gear fault signals show that ASTFA-OIV can not only improve the accuracy of the decomposition results of ASTFA,but also can effectively applied for gear fault diagnosis.(2)Except the disadvantage in the setting of initial values,there is another problem existed in ASTFA which is mode mixing.Therefore,ASNBD method is proposed aiming at solving these problems.Firstly,the optimization of all the parameters in the filter in ASNBD is solved by genetic algorithm(GA),so that the initial values can be set randomly.Secondly,a differential operator is used as the objective function to constrain the decomposed components to be local narrow band signals,thus the physical meaning of the components is more specific,and the mode mixing in the decomposition results can be avoided.ASNBD is utilized to analyze simulation signals and rolling bearing fault vibration signals,the results show that the proposed method can restrain the mode mixing phenomenon,as well as be applied for rolling bearing fault diagnosis.(3)Aiming at avoiding the advantages existed in GA such as it is easily trapped into the local optimal solution,has low calculation efficiency and too much input parameters,ACROA method,which is an innovative heuristic algorithm inspired by chemical reactions,is applied to solve the optimization problems in ASNBD.ASNBD based on ACROA(ASNBD-ACROA)is proposed on that basis.The analysis results of simulation signals show that compared with ASNBD based on GA,ASNBD-ACROA is more robust,needs less input parameters and calculation time.Meanwhile,the diagnosis results of rotor show the effectiveness of the proposed method in mechanical fault diagnosis.(4)Aiming at solving the problem of fault pattern recognition,fault pattern recognition method by using ASNBD and maximum margin classification based on flexible convex hulls(MMC-FCH)method is proposed.MMC-FCH is a newly proposed pattern recognition method.Flexible convex hulls are defined in MMC-FCH based on the basic theory of support vector machine(SVM)so that a more appropriate estimation for categorical distribution can be obtained.The analysis results of experimental signals show that compared with SVM,the proposed method is superior in both the robust and the classification accuracy.(5)To achieve the recognition of the degeneration states and the life prediction of rolling bearing,method based on ASNBD and Gaussain mixture model(GMM)is proposed.GMM is applied to obtain the regeneration states and the number of the states,as well as recognize the abnormal data in the training data sets.Method on the recognition of the degeneration states and the life prediction of rolling bearing based on ASNBD and GMM is proposed on that basis.The analysis results of experimental signals show that the proposed method can improve the accuracy of degeneration states recognition and life prediction.