Condition Detection Of Gas Turbine Blades Based On SVM Model

Modern gas turbine is very important in industry. The industrial scale of gas turbine , to some degree, can be considered as a symbol of national industrialization. However, with the complex structure, gas turbine easily occur faults in the harsh environment with high temperature, high pressure and high speed. Thus, condition detection for gas turbine has great value. Turbine blade is an important part of the gas turbine. However, at present, the research in this field focus on gas turbine health monitoring, sensor condition detection, vibration state monitoring and fault diagnosis, and gas path condition detection. Statistics shows that blades failure events occupied more than 40% of the gas turbine component failure events, therefore,it's important of studying feature extraction and condition detection technology for gas turbine blades. This paper aiming to judge faults conditions and classify them, feature extraction and compute feature vectors, based on gas turbine blades temperature data,combined with the cooling structure and failure mechanism of blades.In this paper, we introduce the research meaning and purpose at first, and also a brief introduction of this topic in our country and abroad is mentioned. Due to the Support Vector Machine(SVM) has a great advantage on small sample feature extraction, we decided use SVM to research this topic. With an introduction of SVM theory in detail, which lay a foundation for the use of SVM model.The paper analyzes temperature data and processing, mainly dealing with its periodic and simulating under various working condition,and successfully isolate all the single turbine blade temperature data. Select Ensemble Empirical Mode Decomposition (EEMD) method to deal with turbine blade temperature data for the feature extraction, and successfully get the good characteristic vector integration, which lays the foundation for the next work.In the end, we designed a condition detection classifier to classify the condition of gas turbine blade. We choose Least Squares Support Vector Machines (LSSVM) as basic SVM model By using artificial bee colony algorithm to optimize parameters of LSSVM(ABC-LSSVM), and compared with other methods, both result and efficiency of condition detection has been improved obviously. By the way, this paper has a research on uneven sample in the production, the result shows that ABC-LSSVM model has good performance in this situation.