基于多输出支持向量回归的声发射源平面定位  

Planar location of acoustic emission source based on multi-output support vector regression

作　　者：于金涛[1,2] 丁明理[1] 王祁[1] 

机构地区：[1]哈尔滨工业大学自动化测试与控制系,哈尔滨150001 [2]哈尔滨商业大学计算机与信息工程学院,哈尔滨150028

出　　处：《仪器仪表学报》2011年第9期2139-2145,共7页Chinese Journal of Scientific Instrument

基　　金：黑龙江省自然科学基金(No.F201018);国家自然科学基金(No.60901042)资助项目

摘　　要：为了解决直升机动部件疲劳损伤定位问题,提出了基于多输出支持向量回归算法的声发射源平面定位方法。以声发射信号的多个时域参数作为输入,破损点的平面坐标(x,y)作为输出,用支持向量回归机逼近输入输出之间的非线性映射关系,然后利用支持向量回归机的泛化推广能力,实现声发射源的平面定位。通过碳纤维材料试件断铅定位试验结果表明:该方法有效的实现了声发射源的平面定位,并且在收敛速度和定位精度上优于RBF神经网络。To solve the fatigue damage location problem of helicopter moving component, a new approach for planar location of acoustic emission （AE） source based on multi-output support vector regression （M-SVR） was proposed. Several time domain parameters of the AE signal are taken as the inputs, and the planar coordinates （x,y） of the breakpoints as the outputs. Support vector regression machine is applied to approximate the mapping relationship between the inputs and outputs, and then the generalization ability of the support vector regression machine is utilized to implement the planar location of the AE source. The results of pencil lead break location experiment on a specimen of carbon fiber material indicate that the proposed approach can implement the planar location of AE source effectively, and has better performance in convergence rate and location accuracy than RBF neural network.

关 键 词：多输出支持向量回归机 RBF神经网络 平面定位 断铅试验 声发射 

分 类 号：TG115.28[金属学及工艺—物理冶金]