Font Size: a A A

Fault Diagnosis Method Based On Sparse Representation

Posted on:2015-11-22Degree:DoctorType:Dissertation
Country:ChinaCandidate:H F TangFull Text:PDF
GTID:1222330476453949Subject:Mechanical design and theory
Abstract/Summary:PDF Full Text Request
With the development of science and technology, the mechanical equipment has developed to the direction of complexity, precision, integration and intelligence. However, the operating conditions have become increasingly harsh. So it put forward higher requirements on the reliability of the equipment. In order to avoid economic losses and human injury caused by unpredictable equipment failure, development of condition monitoring and fault diagnosis is very necessary. Advanced condition monitoring and fault diagnosis technology is helpful in maintenance and improving the efficiency and reliability of machine.Taking rolling bearing and gear as the main object, we carried out the research of fault feature extraction based on the latest research on signal sparse representation theory. Our work focuses in three aspects: the time frequency distribution of fault signal based on the reconstructed sparse atoms, morphological component analysis and weak feature extraction method based on shift invariant sparse coding theory. Several methods about noise cancellation, feature enhancement and multi fault analysis are presented in the paper, and these methods applied to various simulation and experimental data to verify its effectiveness. The main contents are as follows:(1) From the viewpoint of condition monitoring theoretical analysis and practical engineering applications, the background and significance of the paper are elucidated. The present status of mechanical fault signal processing technology is introduced from three aspects: timed domain analysis, frequency domain analysis and time frequency domain analysis. The application and research potential of sparse representation theory in fault diagnosis is discussed, and the contents and technical framework are presented.(2) The sparse representation theory is introduced, including the mathematical model, sparsity metrics and signal reconstruction conditions etc.. The optimization problem about sparse representation is presented in details, and coefficients solving algorithms such as basis pursuit, FOCUSS, matching pursuit etc. are described and compared. In addition, the concept of construction of redundant dictionary, the typical analytic dictionaries and the self learning dictionary is introduced, which is the theoretical background for the subsequent chapters.(3) The association between sparse representation and time frequency distribution of signal energy is introduced. The characteristics of time frequency distribution of sparse atoms reconstruction is concluded through comparative analysis. According to the structure and kinematics characteristics of rolling bearing, the method for calculating characteristic frequency of bearing is derived and the fault simulation signal is presented. A method combined with AR pre-whitening and sparse time frequency distribution is proposed in the paper. Experiments prove that the method can effectively suppress the inherent components of the signal and the noise, and then strengthen the impact component.(4) The basic concept, principle and algorithm of morphological component analysis are introduced and its characteristics are further expounded by examples. According to the structure and operation characteristics of gear transmission, the simulation model of gear fault signal is proposed. The addictive mixed simulation signal of gearbox is analyzed by morphological component analysis method. Finally, the feasibility and effectiveness of the proposed morphological component analysis method is validated by experiments.(5) The reason why shift invariant sparse coding provides a direction to solve the weak fault feature extraction challenge is introduced firstly. Then the mathematical model of shift invariant sparse coding and its fast algorithm are presented. A weak feature extraction method based on SISC idea is proposed in the paper, that the recurring weak feature can be effectively extracted. Through the analysis of simulation signal with heavy noise, the algorithm implementation process and the parameters selection principle are introduced, and the effectiveness is also proved. Through three typical experiments analysis, its ability of early fault detection, weak bearing fault feature extraction in gearbox and weak gear fault extraction are validated.(6) How to use shift invariant sparse coding idea to solve the problem of multi faults signal analysis is studied. Combining the characteristics of machinery fault signal, a method for single channel blind source separation is proposed based on shift invariant sparse coding and adaptive clustering. The source number is estimated by minimize the structural correlation among source signals. Through simulation, the process and the effect of the method are introduced. The analysis results show that the proposed method can effectively separate different source signals caused by different type of fault.
Keywords/Search Tags:Fault diagnosis, Condition monitoring, Gear, Bearing, Sparse representation, Sparse metric, Time frequency distribution, AR pre-whitening, Morphological component analysis, Shift invariant sparse coding, Latent components
PDF Full Text Request
Related items