基于信号谱熵的模糊Hamming网络缺陷识别  

Application of Fuzzy Hamming NN Based on Information Entropy of Signal in Recognizing Defect

作　　者：秦旭达[1] 刘兴荣[2] 王太勇[1] 商同[1] 

机构地区：[1]天津大学机械工程学院,天津300072 [2]北京市计量科学研究所,北京100029

出　　处：《天津大学学报（自然科学与工程技术版）》2004年第2期140-143,151,共5页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目(50175081);天津市重点基金资助项目(993802411).

摘　　要：无缝钢管损伤监测和诊断的关键在于缺陷信号特征的提取和识别.作者研究了无缝钢管内伤、外伤及孔洞等缺陷信号的频域特征,提取信号二维谱熵指标作为特征指标,运用改进的模糊Hamming神经网络识别方法,对无缝钢管的缺陷进行识别.结果表明,该方法具有很高的识别精度,实例中识别准确率达100%.由于该识别方法不需要示教学习过程,因此能够应用于实时在线缺陷识别.Abstracting and identifing the signal features are the keys to diagnose the seamless tube flaw.The seamless tube flaw signals of internal flaws,external flaws and holes were investigated in frequency domain.With two-dimension spectrum entropy indexes being selected as character indexes in feature abstraction of defect, the seamless tube flaws were recognized by the improved fuzzy adaptive Hamming NN(Neural Networks) algorithm.The experiments demonstrate that the recognition algorithm possesses high precision, with an accuracy of 100%.The algorithm,which does not require to be taught, is suitable for real time monitor.

关 键 词：信号谱熵 模糊Hamming神经网络 无损检测 无缝钢管 二维谱熵 缺陷识别 漏磁检测 

分 类 号：TG115.28[金属学及工艺—物理冶金] TG142.15[金属学及工艺—金属学]