摘要
针对叶片在服役过程中缺陷特征提取困难,提出一种基于变分模态能量熵结合BP神经网络的叶片缺陷诊断方法。首先对声发射信号进行变分模态分解,通过方差贡献率筛选不同缺陷的主要模态分量,之后求取不同缺陷主要模态分量的能量熵构造不同缺陷的特征向量。为验证特征向量选取的准确性,将不同缺陷能量熵向量输入BP神经网络进行缺陷模式识别。结果表明:缺陷识别正确率高达90%,表明变分模态能量熵结合BP神经网络的叶片缺陷诊断方法能够实现叶片早期缺陷识别,具有一定的应用价值。
Given the difficulty to extract the defect features of blades during service, a diagnosis method of blade defects based on variational mode decomposition（VMD） energy entropy and BP neural network is proposed in this paper. Firstly, the acoustic emission signal originated from blade was decomposed by VMD,and the intrinsic mode functions（IMF） containing main feature information were selected through the variance contribution rate. Then, the energy entropy of IMF of different defects is obtained to construct the eigenvector of different defects. Finally, in order to verify the accuracy of the eigenvector selected, the energy entropy vector of different defects was input to BP neural network to achieve defect mode recognition. The results show that the accuracy of defect recognition is higher than 90%, and the diagnosis method of blade defect with a combination of VMD energy entropy and BP neural network can realize the blade defect recognition in early stage, with certain application value.
作者
张鹏林
徐旭
杨超
董拴涛
ZHANG Penglin1, XU Xu2, YANG Chao1, DONG Shuantao1(1. School of Materials Science and Engineering, Lanzhou University of Technology, Lanzhou 730050, China; 2. Lanzhou LS Testing Technology Co., Ltd., Lanzhou 730314, China)
出处
《中国测试》
CAS
北大核心
2018年第9期115-120,130,共7页
China Measurement & Test
