Study Of Fault Diagnosis Of The Rolling Bearing Using GMM-HMM

With the rapid development of modern high-tech, modern industrial processes are maturing in the modernization of the plant machinery. Equipment requirements are also increasing. But with the operation of machinery and equipment, mechanical equipment failures also appear inevitable. The Rolling Machinery is one of the most widely used parts. So what the pros and cons of rolling bearings operating status directly affects the performance of the whole machinery. Rolling bearing is a part of rotating machinery which is the most widely used and most prone to failure, and it’s running state will directly affect the performance of the whole unit.Traditional pattern recognition methods(such as neural network identification method) has been stuck on a static pattern recognition problems, this paper proposes preclude the use of a recently developed fast dynamic pattern recognition technology in voice recognition technology- Hidden Markov Models to rolling element bearing fault diagnosis. When HMM statistical modeling is dynamic information on a time span, especially for a large amount of information, non-stationary, poor reproducibility of diagnostic characteristic signal classification. Under normal circumstances the dynamic process of the sequence behavior change performance, fault diagnosis of rolling bearings is also true. If a short time signal is defined as a frame, then transfer each fault type-specific frames are different, so you can use HMM to do statistical processing of the transfer and the presence of a specific frame between frames. In addition, less use of HMM when training the model used in a sample, faster, and diagnostic accuracy is high, a strong pattern classification ability, so it is very suitable for vibration signal of rolling bearing fault modeling and classification. Therefore, this article as the Rexnord ER10 k Rolling models for the study and as the Rolling pattern recognition research purposes. Pattern recognition is proposed based on the research and application of rolling bearing fault diagnosis on Hidden Markov Models. Related parts for rolling bearing fault monomer fault simulation study were made. Verificated the experiment for the Rolling related parts. Its main contents are as follows:Extraction of signal based on wavelet packet decomposition and reconstruction.Data collected by the experiment,Due to the mechanical operation of the factors or external environment,Bring some noise to the experimental data signal is inevitably.Because of the rolling bearing vibration signals associated with different parts of the frequency domain energy.So the rolling bearing fault signal after wavelet packet decomposition and reconstruction got all the wavelet packet decomposition coefficients.Then wavelet packet coefficients were reconstructed after the decomposition.According to different failure modes vibration signals in the frequency domain energy distribution of the differences, the wavelet coefficients reconstructed Feature Extraction.Then through the feature vector wavelet packet decomposition and reconstruction after the fault signal is normalized, using this method not only improves the resolution of the signal, but also played a very good noise cancellation effect.Research on Modeling and Simulation of fault rolling bearing based on single ADAMS.Rexnord ER10 establish Rolling model targeted at solidworks inside.Through simulation, the experimental parameters set to obtain the rolling bearing monomer fault signal under ideal conditions.The Rolling monomer fault signal simulation and experimental data obtained as the establishment of a Gaussian input samples of each failure mode hybrid hidden Markov model, Using experimental data and simulation data as a Gaussian mixture HMM input samples also good adapted to the later stages of rolling bearing fault patterns are different for each degree of injury to provide a good basis for discrimination.Fault diagnosis used in Gaussian mixture Hidden Markov model. Because hidden Markov model was first used in the field of speech recognition,Fault diagnosis rarely considered in the rolling bearing.this paper analyzes the characteristics of rolling bearing fault diagnosis and discussed the feasibility of hidden Markov model in rolling bearing fault diagnostic applications.Several different types of application methods for Hidden Markov Models were discussed, it concluded that the Gaussian mixture explore hidden Markov model for fault diagnosis of rolling bearing applications than other types of methods have an advantage.Bearing Fault Diagnosis based on modeling and experimental validation of Hidden Markov Models.Rolling fault signal gets through wavelet packet decomposition and reconstruction of energy to get the energy of the band’s characteristic signal using a signal consisting of feature vectors as sample of input for GMM-HMM model.Gaussian mixture Hidden Markov Model for GMM-HMM model training.The GMM-HMM model get a different failure modes.Finally, the test data for GMM-HMM model is set up to test.By calculating the current state of the monitoring data in the probability of GMM-HMM model library appears, and the maximum probability of the failure mode in which to assess.The rolling bearing fault signal simulation monomer obtained as a sample input to the hidden Markov model.Hidden Markov model to determine the effect of diagnosis.The study found which can be well applied to hidden Markov model after wavelet packet decomposition and reconstruction for rolling bearing fault signal. The results are displayed as a high recognition rate.And the number of samples to establish Hidden Markov model requires less calculation and not complicated.Therefore, Fault Diagnosis Method were well applied for the Rolling Bearing Based on Wavelet Packet Decomposition and GMM-HMM.