Research On The Multiple Faults Diagnosis Of Gearboxes Based On The Morphological Component Analysis

Gearbox is an important power transmission component of mechanical equipment, its operation state greatly affects the safety and reliability of mechanical system. Therefore, condition monitoring and fault diagnosis of gearbox is practicably valuable and realistically significant. In the existing gearbox fault diagnosis techniques, the methods based on vi bration signal have higher practicability and wider application range. Extracting fault feature information from the gearbox ’s vibration signal by using various kinds of signal processing methods has been the key to the fault diagnosis for gearbox.In engineering practice, the mechanical equipment fault appears not individually, some failure often induce the occurrence of other faults, which demonstrate as multiple faults. The multiple faults in different positions, with different failure modes and degrees have different effects on the mechanical equipment. Besides, the interaction and mutual interference among different fault components bring great challenge to the comprehensive fault diagnosis of mechanical equipment, thus, the multiple fault diagnosis has become a hard problem in the fault diagnosis of mechanical equipments. Supported by the project of Natural Science Foundation of China, which named as “Research on the resonance-based sparse signal decomposition method and its application in mechanical fault diagnosis(Project Approval Number: 51275161)” and the independent research project of the s tate key laboratory of advanced design and manufacturing for vehicle body of Hunan University, which named as “Research on the early fault diagnosis and residua l life prediction techniques of key automotive components and parts(Project Number: 71375004)”, Aiming at the problems above, the signal processing methods have been used to separate multi-fault features of the gearbox in this thesis.The main research work and innovative achievements of the thesis are as follows:(1) The vibrational mechanism, fault modalities and causes of the main components in the gearbox have been analyzed, and then the local fault vibration signal model of gears and rolling bearings have been set up, Finally, the morphological differences caused by different faults are analyzed. The results of study show that, when a local fault occurs in a gear, there will be harmonic component in the vibration signal, whereas, when a local fault occurs in rolling bearings, there will be periodic impulse component with damped oscillation.(2) The morphological component analysis method(MCA) was applied to the analysis of the vibration signal with gear fault and rolling bearing fault, the decomposition results show that the MCA is closely related to the selection of the sparse dictionaries, and the optimal decomposition result is difficult to be obtained with the sparse dictionaries which were selected based on experience. Meanwhile, the MCA cannot separate useful fault feature under strong noisy environment. Aiming at these problems, an improved MCA method for the multiple faults diagnosis of the gearbox is proposed in this thesis. In the proposed method, the sparse dictionary which can best match each fault characteristic component is firstly selected based on the principle of minimum information entropy. Then, according to the morphological difference between the local fault component of the gear and that of the bearing, the original vibration signal of the gearbox with multiple faults can be decomposed into the harmonic component and the periodic impulse component. Finally, the gear fault component and the bearing fault component are subject to the Hilbert spectrum analysis respectively. The gear and the bearing faults can then be diagnosed based on their own the envelope spectra. Simulation results and application examples show that the proposed method can effectively reduce the noise influence and separate the fault features of gear and bearing.(3) When bearings have multiple faults, the vibration signals of faulty bearings contain a variety of similar form impulse components, it is easy to cause missed diagnosis, because MCA cannot separate multiple impact components with similar form. Aiming at this problem, a method based on wavelet matching pursuit for bearing multiple faults diagnosis is proposed. Firstly, a new impact function model based on the characteristics of bearing multiple faults is established through varying parameters individually. Next, the similarity coefficients between dictionaries and fault characteristics can be obtained using the waveform matching pursuit. The obtained coefficients which are bigger than the set threshold were selected by using adaptive threshold algorithm. Then, the selected coefficients were analyzed by Hilbert spectrum analysis. Finally, the bearing multiple faults can be diagnosed through observing the envelope spectrum of selected coefficients. The simulation and experiment results show that the proposed method can effectively separate the bearing multiple faults, and then avoid missed diagnosis.At present, the multiple faults diagnosis has been the difficulty in the fault diagnosis of mechanical equipments. T he improved MCA and waveform matching pursuit algorithm have been used in this thesis to analyze the multiple faults vibration signal of gearbox. Algorithm simulation and application examples indicate that the above mentioned methods can effectively extract the multi-fault characteristics from the gearbox vibration signal and then avoid missed diagnosis. Therefore, these methods have a brilliant application prospect in the fault diagnosis of mechanical equipments.