涡流检测在钢轨裂纹定量化评估中的应用  

Application of eddy current testing in the quantitative evaluation of the rail cracks

作　　者：李国厚[1,2] 黄平捷[1] 陈佩华[1] 侯迪波[1] 张光新[1] 周泽魁[1] 

机构地区：[1]浙江大学控制科学与工程学系,工业控制技术国家重点实验室,浙江杭州310027 [2]河南科技学院信息工程学院,河南新乡453003

出　　处：《浙江大学学报（工学版）》2011年第11期2038-2042,2049,共6页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(50505045,61174005);教育部博士点基金资助项目(200803350058)

摘　　要：针对应用较多的超声检测方法的不足,研究涡流检测技术在钢轨裂纹定量化无损检测中的应用,阐述涡流检测试验系统的组成、原理以及试验的设计,采用减聚类算法对径向基函数(RBF)神经网络进行改进,并基于试验系统检测试件的数据对网络模型进行训练.在试验中采用基于巨磁阻(GMR)传感器的检测探头,有效地提高系统对深层缺陷和表面微小缺陷的检测能力.试验结果表明,采用改进算法建立的模型在对裂纹进行反演时具有较高的精度,同时缩短了反演模型的训练时间,在一定程度上满足钢轨裂纹参数在线检测的要求.In order to overcome the disadvantages of the commonly-used ultrasonic testing method,the application of eddy current testing（ECT） in quantitative nondestructive testing of rail cracks was investigated.The structure and principle of the ECT experimental system and inspection experiment design methods were expounded.The subtractive clustering algorithm was used to improve the RBF neural network,and the experimentd testing data were used to train the network model.A pair of GMR-based probes were used to increase the detecting performance for the deep defects and the minor surface defects.The experimental results show that the constructed model by the improved algorithm possesses higher accuracy in the inversing of the rail cracks,and in the mean time,the training time of the inversing model is decreased,and it is promising for the inline inspection of the rail crack parameters.

关 键 词：钢轨裂纹 涡流无损检测 巨磁阻 反演 RBF神经网络 减聚类算法 

分 类 号：U28[交通运输工程—交通信息工程及控制]