The Studing Of Fault Diagnosis Methods For Vulnerable Key Components Of Rotating Machinery

Due to the special status of rotating machine in various types of mechanical equipment. Fault theories and diagnosis methods has always been a research focus of experts and scholars home and broad. In rotating machinery equipment, the rotor and its related components are the major fault sources, and the most of its vibration signal appears unstable, dynamic and random, and the fault signal becomes more complex when the rotating equipment works in the variable speed. So the signal extraction and fault recognition method has always been a complex and interesting subject. With the continual emergence of new technologies and methods of a variety of signal processing and pattern recognition, new vitality has been injected in solving these problems. This paper aims to introduce some new methods of signal processing and pattern recognition technology in fault diagnosis of rotating machine, and some useful researches and explorations of fault diagnosis methods for fault-prone bearings, gears and rotor.(1) For the high noise ratio but unstable characteristics of resonance signal caused by rolling bearing fault at the constant speed conditions, EMD method is employed so that the resonance signal can be decomposed and fault information that contains smooth signal can be extracted, after demodulation by envelope the characteristics in reflecting the bearing fault state vector is constructed. Diverse training methods of DHMM are researched and the status of the bearing is also identified. Experimental result shows that the proposed method is capable in carrying out satisfactory result in testing both single and multiple faults.(2) Pointing out that the resonance signal of the bearing caused by the failure of the antifriction bearing is not a simple bias AM signal, utilizing the fault models to test the outcome after demodulated by CAF. It is proved that the morphology of CAF could report the failure status of the bearing based on research, comparison and analysis. Combining the ability of CHMM to recognize the dynamic signal, a method of antifriction bearing fault detection is proposed, and the validity of the method has been verified.(3) The fault characteristic frequency disappears during the starting procedure of the antifriction bearing due to the change of the rotate speed. The fault resonance signal is figured out by the method of wavelet analysis, the vibration signals of the antifriction bearing are analyzed by the method of estimation of instantaneous frequency and ratio tracking, so that the concept and algorithm of the fault characteristics and ratio coefficient are raised, and the feasibility of the algorithm has been verified.(4) The change of the rotate speed is closer to real situation of the application of the rotating machinery. This article indicates the dual frequency conversion characteristics of variable speed antifriction bearing fault signal. Computational formula of OCAF is proposed, and research on the demodulation capability on dual-frequency signal has been done. A method of fault diagnosis on the rolling bearing is carried out based on OCAF-CHMM and the characteristic rector of the.multidate information under the specific rotate speed, and the method has been verified to be effective.(5) The feature information of the gear is obtained via DCS, and the failure of the gear has been identified via DHMM model. The experiment indicates that the proposed method can recognize the broken tooth, corrosive pitting and sort of status correctly. The specific method is the combination of the three features, such as the high demodulation SNR of DCS, simplicity of expression and the reduced quantity of computation, which can improve the diagnostic rate.(6) Fully taking the advantage of HMM model and SVM model on the sequence behavior classification and small sample, a rotating machine faults diagnosis method is proposed based on SVM-HMM model. Comparing with the SOM-HMM method, SVM-HMM method reduces the sample numbers of the information loss and model training, decreases the time of model training and increases the rate of identification. A satisfactory outcome on the test of diagnosis on the damage of rolling bearing has been obtained, and the feasibility of SVM-HMM model on the diagnosis of the rolling bearing has been verified.