Research On Fault Feature Extraction Of Rotating Machinery Based On EEMD And Entropy Characteristics

In the research of intelligent diagnosis of rotating machinery faults,the effective extraction of fault features is a key step in the implementation of mechanical diagnosis.The accuracy and reliability of fault identification are affected by it.Therefore,this is a key issue in the research of mechanical intelligent fault diagnosis.In this dissertation,a set of empirical mode decomposition adaptive signal processing methods are combined with singular value decomposition and entropy feature to extract the fault features and study the pattern recognition of vibration signals of different states of bearings and rotors in rotating machinery.The main research contents are as follows:(1)This paper introduces EEMD decomposition algorithm for the deficiency of EMD.Compared with EMD decomposition principle and decomposition step,this paper analyzes that EEMD decomposition has stronger modal anti-aliasing performance than EMD decomposition and summarizes the advantages and disadvantages of both.In addition,three classical principles of entropy feature algorithms—information entropy,fuzzy entropy and singular value entropy—are introduced.Three entropy-adapted sequences are described,and corresponding mathematical expressions and definitions are given in combination with their physical meanings.(2)In order to make full use of the vibration signal for fault identification,a rolling bearing fault diagnosis method based on ensemble empirical mode decomposition and singular value entropy criterion is proposed.Firstly,the EEMD decomposition of the vibration signal of the rolling bearing is carried out to obtain the intrinsic mode functions.According to an evaluation index(ie,kurtosis,mean square error and Euclidean distance)of the IMF components fault information content,the components that can characterize the original signal state are selected for signal reconstruction.Then the singular value entropy is used to deal with the reconstructed signal,and the singular value entropy is obtained by combining the information entropy method.Finally,the fault category of the rolling bearing is judged by the size of the singular value entropy.The results show that compared with the traditional EMD singular value entropy fault diagnosis method,this method can more clearly distinguish the different characteristics of the rolling bearing different types of work characteristics of the interval.It has a higher fault diagnosis accuracy.(3)Combining the SVD with EEMD,a sensitive feature extraction method for the description of weak state of rolling bearing is proposed.In order to improve the quality of signal failure,the Hankel matrix is reconstructed by phase space reconstruction of the collected signal.According to the singular value difference spectrum of the matrix,the order of noise reduction is determined.Using EEMD to decompose the signal after noise reduction,we can obtain 11 IMFs and one R.According to the established kurtosis-mean square error criterion,we can select one of the most effective states Sensitive IMF,and calculate its corresponding Teager Energy Operator,Fourier transform of TEO,thus realizing the effective identification of weak failure mode of rolling bearing.The new method is compared with the traditional EEMD-Hilbert method and EEMD-TEO method by the opening rolling bearing fault signal of the US West Reserve University.The results show that the sensitive features extracted by this method can accurately identify the cycle frequency of rolling bearing fault and accurately identify the fault type,which provides an effective method for weak feature extraction of rolling bearing.(4)According to the fault recognition problem of low accuracy,rotating machinery fault diagnosis method is proposed for EEMD and fuzzy information entropy integration.The method combines the advantages of EEMD decomposition and fuzzy information entropy in feature extraction and constructs a feature set which can finely measure the complexity of the fault probability of different classes of vibration signals.Firstly,the original vibration signal is decomposed by EEMD to obtain a number of IMFs;The fuzzy information entropy of the first 5 high frequency IMF components is calculated;Using LPP to reduce dimensionality of high-dimensional winner,we eliminate redundant irrelevant features;Finally,the reduced sample set is input into the KNN classifier to identify the faults.The method is validated by the data collected from a two spans rotor test rig,and compared with the EMD fuzzy entropy,EMD fuzzy information entropy and EEMD fuzzy entropy method,the results show that this method can effectively extract the fault characteristics of rotor vibration signals and has higher fault recognition rates.