基于主成分分析的叶片温度分布一致性研究  

Research on consistency of turbine blade temperature distribution based on principal component analysis

作　　者：孟宪文 冯驰[1] 高山[1] MENG Xian-wen;FENG Chi;GAO Shan(School of Information and Communication,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学信息与通信工程学院

出　　处：《哈尔滨商业大学学报（自然科学版）》2019年第4期451-457,共7页Journal of Harbin University of Commerce：Natural Sciences Edition

基　　金：国家自然科学基金(61805056)

摘　　要：采用特征向量比对的方式进行研究.通过对叶片温度数据时域和频域空间特征提取,获得能够体现叶片温度分布信息的特征向量,并采用主成分分析(PCA)对特征向量进行去相关和降维,进而得到包含绝大部分叶片温度分布信息的低维特征向量.根据叶片温度特征向量与基准特征向量间的对比,评估叶片温度分布情况的一致性.实验结果表明,叶片的特征偏差均在5%以下,偏差较大的叶片数量占总量的5.8%,该研究成果可以用于从温度特征上反映叶片工作状态,为叶片故障诊断及寿命预测提供基础.This paper used the method of eigenvector comparison in order to compare the temperature distribution characteristics of each turbine blade.By extracting features of blade temperature data in time and frequency domains,the eigenvector was obtained to reflect the information of blade temperature distribution.Principal component analysis(PCA)was used to decorrelate and reduce eigenvector dimension,the low-dimensional eigenvector was obtained,which contains most of the information of blade temperature distribution.According to the comparison between the blade temperature eigenvector and the reference eigenvector,the consistency of blade temperature distribution was evaluated.Experimental results showed that blades deviation was less than 5%,and blades number with larger deviation was 5.8%of the total.Research results can be adopted to reflect blades working status from temperature characteristics,which provided a basis for fault diagnosis and life prediction of blades.

关 键 词：特征提取 无量纲指标 小波变换 主成分分析 一致性评估 

分 类 号：TN912[电子电信—通信与信息系统]