机翼结构小损伤程度识别研究  

Research of Small Damage Severity Identification for Wing Structure

作　　者：李春锦[1] 闫云聚[1] 张宇[1] 

机构地区：[1]西北工业大学力学与土木建筑学院,陕西西安710129

出　　处：《计算机仿真》2014年第9期75-79,共5页Computer Simulation

基　　金：国家自然科学基金资助项目(50875212);教育部博士点基金资助项目(200806990019);2012年博士点基金资助课题(20126102130004)

摘　　要：在飞机损伤准确识别问题的研究中,飞机机翼结构的局部小损伤识别对于机翼的在线监测和及时检修有重要的意义。针对仅能获得有限试验样本而造成识别精度不高的问题,提出了一种将模态频率和支持向量机(SVM)相结合的新方法,对机翼结构进行局部小损伤程度识别研究。利用ANSYS软件构建无损和有损机翼仿真模型,于不同损伤工况下对其进行模态分析,在SVM良好小样本泛化能力基础上建立了准确的损伤识别学习机,并通过网格寻优对参数进行优化处理,最后得出识别结果。为了验证识别结果的可信性,同时取用另两组核函数和BP神经网络方法进行对比试验。结果表明,小样本条件下,用模态频率和SVM结合来识别机翼的小损伤程度的方法是可靠有效的。The local small damage identification of aircraft wing structure has important significance for the online monitoring and timely maintenance. For the problem of limited samples and low accuracy, a new approach combining modal frequency and support vector machine （SVM） is proposed to identify the local small damage severity of the wing structure. An intact and damaged simulation model is constructed by the ANSYS. After modal analysis under different damage conditions, a damage identification model is established based on the perfect generalization ability for small sample of SVM, and the parameters are optimized through the grid optimization. At last the identification results are given. The comparisons with other two groups of kernel functions and BP neural network method are made in order to validate the credibility of the results. The simulation results show that the approach combining modal frequency and SVM to identify the local small damage severity of the wing structure is reliable and effective under the condition of limited samples.

关 键 词：机翼结构 损伤识别 模态频率 支持向量机 核函数 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术] O302[自动化与计算机技术—计算机科学与技术]