A Research To The Fault Analysis And The Trend Predicting Of Large Rotary Sets Based On Virtual Instrument

The large rotary sets are commonly important equipment in the industry production process. Ensuring them safely and steadily running is crucial to ensure produce. Today, because the equipments become more and more advanced and complexed, the running state, technical parameter and working condition of the equipment need be considered. The simple and convenient diagnosis and predicting technique is not enough to help the state maintenance of the equipment, so the fault analysis and the trend predicting of large rotary sets need be studied, which adopt to modern production, to realize state maintenance to the sets when the sets hasn't gone wrong. This article expatiates the theory and method of the large rotary sets fault diagnosis. The fault diagnosis of the large rotary should include on-line monitoring on the sets working condition, selecting the characteristic which can indicate the sets state, analyzing the characteristic and identifying the sets working condition. The vibrating signal can indicates the running state more directly, quickly and correctly, so it is selected characteristic of fault diagnosis. The methods to obtain and pre-process the vibrating signal and how to calculate the intensity are studied. Spectrum analysis, cepstrum analysis, short time Fourier analysis and wavelets analysis are used in the process of vibrating intensity signal. The running state of the sets is discovered. The virtues and shortcomings of each analysis method are compared. The method of digital filter of wavelets analysis is researched and used to filter the acquired vibrating signal. Because of themselves and abroad factors, the running large rotary sets always run at non-stationary state. The conventional predicting method using stationary time series is less effective, so the predicting method of non-stationary state is researched. The abnormal data and the aberrant data among acquired vibrating signal are pre-processed. The database that storage predicted sample has been built. The intensity trend predicting using the method of Hidden Markov Model (HMM), which has been successfully applied to sound signal processing, is discussed. The HMM based on vibrating intensity has been built. The parameter restimate algorithm and optimization state search algorithm of the HMM are studied. The predicting result is prior to the one using Grey predicting model. The system of fault analysis and trend predicting of the large rotary sets based on Virtual Instrument has been built. The hardware adopts PC-DAQ and the software is developed by LabWindows/CVI language compared with Visual C++.