基于分形理论的涡轮叶片特征提取  

Feature extraction of turbine blades based on the fractal theory

作　　者：冯驰[1] 胡杨[1] 王兆丰[2] 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001 [2]西安航空发动机(集团)有限公司,陕西西安710021

出　　处：《应用科技》2015年第4期64-69,共6页Applied Science and Technology

基　　金：黑龙江省自然科学基金资助项目(F201413)

摘　　要：采用了叶片温度这一叶片质量的重要指标,对数据预处理并进行特征提取,为涡轮叶片建立起特征模型。基于分形理论提取叶片温度信号的3种分形维数特征,结合K-means聚类分析和Relief F算法计算各特征值的权重,从而建立起涡轮叶片的温度特征模型,实现对涡轮叶片故障的早期预警。统计结果表明,该特征模型能够较好地反映出处于故障状态的涡轮叶片的状态。The data is preprocessed and features are extracted to establish a feature model of turbine blades by blade temperature,which is an important indicator of the turbine blade＇s quality. Based on the fractal theory,the three fractal dimension features of blade temperature signals are extracted,the weight of each feature is calculated by combining with K-means clustering analysis and Relief F algorithm,and thereby establish the temperature feature model of turbine blades,achieving early warning of the turbine blade failure. Statistical results show that the feature model can effectively reflect the state of turbine blades with failure.

关 键 词：涡轮叶片 特征提取 分形理论 K-MEANS RELIEFF 

分 类 号：TK473[动力工程及工程热物理—动力机械及工程]