Fault Diagnosis Research Of Gas Turbine Based On Support Vector Machine

Gas turbine generating system is a significant process of energy conservation and regeneration of RFCC in refinery, which can make use of the principle of cogeneration of renewable energy to achieve energy recovery purpose. However, Due to mechanical, process, electrical control and some other reasons, the gas turbine breaks down frequently, vibration kept in a high value, probably brought enormous losses to the production of economic and security at any time. So, in behalf of ensuring the normal operation of the equipment, real-time monitoring of the equipment, determining the suspicious points when failures happened in time, reducing the frequency of occurrence of the equipment, and protecting the environment at the same time, the study of the failure of gas turbine generating system is of great practical significance.This article based on the research of the parameters of gas turbine equipment information, fault information, causes and characteristics of the vibration, the current status and development trend of gas turbine vibration signal, establish the experiment of fault diagnosis , achieve the fault sample data collection at the control center equipment of the refinery; Use the wavelet methods for the pre-processing of original fault data, accomplish 6 layer decomposition processing of the simulating fault signal with the DB wavelet, the fourth layer best represent the effect of the optimal wavelet decomposition and achieve the goal of the research over singular point; Through 3 layers decomposition of the original signal, extract the feature vector of the signal, Do intelligent classification after Wavelet pre-processing of the signal based on three ways, C-SVC ,v parameter, and least-square SVM, turned out the classification with v parameter and function estimation with least-square SVM had best results.Setting up the intelligent software platforms Lib-SVM and LS-SVM for gas turbine fault diagnosis, Realized the real-time fault diagnosis of gas turbine after the pre-processing of the signal with wavelet, realized the singular point analysis with the multi-layer decomposition, and obtain the feature vectors, which would be used by SVM for classification, the sensitivity of the SVM classification for the fault data is satisfying, the accuracy reaches 100 percent, and the processing approximately last for 0.45 seconds in average, which is also satisfying, which builds a foundation for the future fault diagnosis, estimation, and the multi-faults classification.