基于支持向量机的加筋板健康监测  

Stiffened plate health monitoring based on support vector machines

作　　者：王长坤[1] 袁慎芳[1] 邱雷[1] 黄红梅[1] 

机构地区：[1]南京航空航天大学智能材料与结构航空科技重点实验室,江苏南京210016

出　　处：《传感器与微系统》2008年第5期49-51,57,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(50278029;60772072);航空科学基金资助项目(04A52002;20060952);教育部新世纪优秀人才资助计划资助项目(NCET-04-0513)

摘　　要：故障样本缺乏是制约智能监测发展的重要原因。支持向量机(SVM)基于结构风险最小化原则,对小样本决策具有较好的学习推广性。将SVM的多类分类算法用于航空航天结构中的加强筋结构健康监测,并与BP神经网络进行对比研究,取得了满意的效果。试验表明:在小样本条件下,SVM具有比BP网络更好的分类性能。Fault samples＇ scarcity is a key factor that prevents the progress of the intelligent monitoring research. Support vector machine（SVM）, based on the structural risk minimization criterion, can generalize a conclusion from limited samples and keep robust. Using the SVM-based multi-classification algorithm for the stiffened structure health monitoring in aeronautics and astronautics area and comparing with the BP neural network, the satisfying results are obtained. The experimental results demonstrate that SVM has better classification performance in this research case than the BP neural network.

关 键 词：支持向量机 加筋板 结构健康监测 模式识别 多类分类 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]