神经网络在涡流检测信号处理中的应用  

APPLICATION OF NEURAL NETWORK TO SIGNAL PROCESSING IN EDDY CURRENT TESTING

作　　者：梁玉红[1] 吴步宁[1] 

机构地区：[1]湖北汽车工业学院电气工程系,十堰市442002

出　　处：《无损检测》2000年第12期545-547,共3页Nondestructive Testing

摘　　要：研究了在涡流检测中用神经网络对缺陷进行分类的方法 ,针对 LVQ神经网络学习算法的不足 ,提出了改进算法 ,提高了网络分类的可靠性。利用 Fourier变换解决了特征抽取问题 ,做到了所需特征参数少 ,且彼此间不相关。分类正确率达 1 0 0 %。The classification of defects by neural network in eddy current testing was studied. In light of the shortage of LVQ neural network study algorithm, an improved algorithm was put forward to heighten the reliability of classification. Characteristic was extracted by Fourier conversion, so that less characteristic parameters were needed and they were not interrelated. The correctness of classification was 100%.

关 键 词：傅里叶变换 信息处理 涡流检验 

分 类 号：TB302.5[一般工业技术—材料科学与工程] TP183[自动化与计算机技术—控制理论与控制工程]