Research On Intelligent Fault Diagnosis Of Turbine Shafting Based On Support Vector Machine

With the development of large-scale, complex turbine of high parameter, people pay more and more attention to technology of state monitor and fault diagnosis to ensure safely running of the equipment, and the technology of fault diagnosis begins developing to intelligent diagnosis. Support Vector Machine (SVM) is a new kind of intelligent learning machine built on Statistical Learning Theory (SLT), and offers an effective means to solve the problem of fault classification with small samples. The application of SVM in turbine fault diagnosis is able to effectively improve veracity of fault diagnosis, and is significative to avoid incalculable loss of accident and enhance economic and social benefit.The paper adopts method of SVM to solve the problems of fault classification and prediction according to familiar vibration fault of shafting, and suggests basis for means of fault diagnosis of turbine. The paper mainly works on data pre-processing, fault feature extraction, fault classification, fault modeling and prediction, turbine fault diagnosis system with method of intelligent fault diagnosis based on SVM. The main research achievements are showed in follows:1. Method of feature extraction based on Principal Component Analysis and kernel function are imported to achieve feature extraction for turbine shafting fault, which uses fuzzy K-L transform to comprese feature dimensions of fault data to reduce compution complexity of SVM's classification arithmetic. Simulation experiment shows that this method is able to effectively improve veracity of fault classification.2. The paper discuss application of SVM in turbine fault diagnosis and constructs SVM muti-classification model to accomplish fault diagnosis with once operation according to several kinds of faults.3. Simulation experiment of the application of Support Vector Regression (SVR) in fault modeling and prediction shows that fault diagnosis based on SVM is more effective than other intelligent method.4. According to actual vibration fault data of turbin shafting, SVM methonds are used to fault classification and fault trend prediction, which validates the validity of the intelligent fault diagnosis based on SVM and offers reference for practicality of SVM.5. A kit of turbine shafting vibration fault diagnosis system is designed and the method of SVM and fuzzy diagnosis are imported to the diagnosis software, which uses SVR's ability of modeling and prediction to analyse the trend of shafting vibration signal. The diagnosis system is able to collect fault data on-line and analyse the fault off-line, which achieves classification of turbine shafting vibration fault and detect early slight fault.