Research On Fault Diagnosis Of Gear Based On Multiscale Fusion Dispersion Entropy

With the continuous development of the national economy and the ever-increasing improvement in the level of science and technology,there are increasing requirements of high-quality,low-cost products and safe operation of machinery equipment in modern industries.The maintenance requirement for industrial machinery equipment has also responded to that,which quickly shifted from preventive maintenance to real-time condition monitoring and intelligent fault diagnosis.As one of the most common types of mechanical equipment,rotary machinery plays an important role in industrial applications.Based on condition monitoring and fault maintenance,rotating machinery equipment can effectively reduce the downtime of unexpected accidents and even avoid catastrophic accidents in industrial enterprises.Therefore,it has important significance and value to deeply develop the fault diagnosis technology of rotating machinery for the safely and highly operation of industrial equipment.Gear is one of the most common mechanical components,which is widely used in the drive system of industrial production lines and plays an important role in almost all types of rotating machinery.Its running status is directly related to the normal production efficiency and equipment safety.Based on the stated above,the paper takes gear as the research object and conducts in-depth study on the identification of weak fault features of gears,noise suppression,and the compound faults detection in gearbox system.The Resonance Sparse Decomposition(RSSD)method is introduced in the paper to the application of processing the gear fault in the gearbox system with non-linear and non-stationary characteristics.Different from other traditional signal processing methods,RSSD can decompose complex signals into several resonance components based on oscillation property,which is especially more suitable and adaptive to analyze fault vibration signals with obvious nonlinear and non-stationary characteristics.However,the RSSD method is sensitive to noise and has a poor anti-noise ability.Therefore,to effectively suppress noise,the paper develops and improves the RSSD,and separately proposed different methods from the perspective of each resonance component to the application of gear fault diagnosis.The main research contents of the paper are as follows:(1)From the viewpoint of the theory of Prognostics and Health Management(PHM)and fault diagnosis technology of rotating machinery in practical engineering application,the research background and meaning of the selected topic are separately discussed.Firstly,overseas and domestic research status of fault diagnosis,feature extraction,and condition monitoring in the rotating machinery system has been reviewed in the paper.Then,the content of this paper has been determined based on the summary of various signal processing methods.(2)This paper proposes a feature extraction method based on Multiscale Dispersion Entropy with rms(MDErms)from the viewpoint of the faulty gear characteristics.The proposed method can quickly quantify the feature richness and complexity of series in different scales,which conbines with manifold learning to fuse the multidimensional features and reduce redundancy.Compared with traditional information entropy,the proposed MDErms effectively avoid the disadvantages of feature insensitivity and information redundancy.Besides,a new gear fault diagnosis method is put forward by making full use of the proposed MDErms,and RSSD.The high resonance components of RSSD has the waveform characteristic of continuous oscillations,which is apply to the resonance demodulation of cracked gear.Firstly,the vibration signal is segmented into a series of local short-time sequences.And the proposed method focuses on analyzing the local short-time series by MDErms.Then combined with a FIR filter design method,a corresponding band-pass filter is proposed by making effectively use of the impulse feature mentioned above.Lastly,the proposed method is verified by the simulation and test results in identifying the weak features of gear faults from the low resonance component effectively.(3)Considering the non-stationary characteristics of gear signal in variable load,the signal sparse representation method based on an adaptive dictionary is proposed for gear fault diagnosis.The paper discusses the waveform characteristic and shock feature of the low resonance component.Besides,the sparse representation algorithm shows good ability in recognizing and extracting the transient pulse with strong impact characteristics,which is adequate for the weak fault feature extraction from low resonance components.The proposed method mainly includes two aspects:adaptive dictionary design and atomic decomposition.Firstly,MDErms and delay segmentation methods are applied to the atom dictionary construction by signal itself rather than primary function.Besides,the K-SVD algorithm is also introduced to train and update atom,which can effectively reduce the dictionary redundancy.Compared to traditional analytic dictionaries,the adaptive dictionary constructed by signal itself show better flexibility and adaptability and can effectively achieve a high similarity degree between atoms and fault impulses.Then the atomic decomposition algorithm is introduced to identifying and extracting the gear fault impact characteristics.On the other hand,in order to acquire the optimal sparse representation of gear signal,the self-constained iterative termination condition is established on the basis of the above filtering method from the high resonance component of RSSD.Then the optimal iteration times can be flexibly determined according to the noise level of the target signal itself.Finally,compared with other traditional signal decomposition methods,the proposed method shows good superiority and reliability in gear fault diagnosis.(4)Aiming at the compound faults detection and decoupling of gear and rolling bearing in gearbox system,a method based on meshing resonance is proposed.This method makes full use of gear meshing resonance phenomenon that the gears with local faults will produce some high-order characteristic frequency modulation sidebands and gear structure resonancein the meshing process.The paper proposes the method with the meshing kurtosis index and translation window filter technology.Combined with fast Kurtogrtam spectrum analysis,the proposed method can effectively separate the different resonance frequency bands corresponding to fault gears and bearings,and identify the natural frequencies of the gear and gearbox systems.Based on the FIR band-pass filter method,the optimal filters with each resonance frequency as the center frequency are separately designed to analyze the vibration signal.Then the separation of compound fault impacts of gears and bearings can be effectively achieved.Besieds,the effectiveness of the proposed method in separating and identifying compound faults of gear and bearing is separately verified by simulation signal and experimental test.