摘要
利用声发射技术检测水轮机叶片裂纹的产生位置。针对水轮机结构复杂及裂纹位置比较集中等特点,提出利用支持向量机的分类与回归功能对水轮机转轮叶片的裂纹进行定位。结果表明,与BP(误差反向传播)神经网络相比,支持向量机在叶片区域的识别率为100%,高于BP网络;裂纹源到焊缝的距离的预测精度也稍高于BP网络,因而支持向量机是一种适合复杂结构的定位方法,特别是在样本量不大的场所。
Acoustic emission （AE） technique was used to detect cracks and their locations in tur- bine blades. Turbine runner has a complex structure and cracks occurred on some special regions. The source location of cracks in turbine runner blades was researched using classfication and re- gression functions of support vector machines （SVM）. The results show that the SVM technique has 100 percent of recognition rate in crack region, which is higher than back propagation neural network （BPNN）, and that estimating precision of distance from source of cracks to welding seam is somewhat better than BPNN.As a result, SVM algorithm is a suitable source location method for complex structures, especially the situation with small sample.
出处
《广西大学学报（自然科学版）》
CAS
CSCD
北大核心
2009年第3期357-360,共4页
Journal of Guangxi University（Natural Science Edition）
基金
国家自然科学基金资助项目(50465002)
关键词
支持向量机
源定位
声发射
support vector machines
source location
acoustic emission
