Page 87 - 无损检测2024年第六期

缺陷的智能识别与分类专题
缺陷的智能识别与分类专题

DOI：10.11973/wsjc202406009

基于超声检测的主动式钢轨伤损智能识别方法

尹段泉
（国能朔黄铁路发展有限责任公司肃宁分公司，肃宁 062350）

摘 要：常规的主动式钢轨伤损智能识别方法，在钢轨伤损数据采集过程中耗时较多，使得伤
损智能识别时间较长。为解决这一问题，提出一种基于超声检测的主动式钢轨伤损智能识别方法。
首先按照增益范围，获取钢轨伤损回波信号，设置超声换能器的滤波器，对信号进行转换，从而对
数据进行采集；然后根据DBSCAN算法，以伤损识别的最小单元为一个单元，对采集的超声数据
进行组合而构成超声信息群，并对数据进行划分，按照划分数据对钢轨伤损特征进行匹配；最后以
AlexNet网络架构为识别主体结构，以数据匹配结果为基础，建立显图样本数据集，对伤损类型进
行精细搜索，从而得到主动式钢轨伤损智能识别结果。试验结果表明，所提方法对主动式钢轨伤损
识别的时间较短，能实现对主动式钢轨伤损的快速识别，具有较好的应用价值。
关键词：超声；钢轨；伤损智能识别；AlexNet网络架构；DBSCAN算法；滤波器
中图分类号：TP273；TG115.28 文献标志码：A 文章编号：1000-6656(2024)06-0049-05

Active intelligent identification method for steel rail damage based on ultrasonic testing

YIN Duanquan
（Suning Branch of Guoneng Shuozhou-Huanghua Railway Development Co., Ltd., Suning 062350, China）
Abstract: The conventional active intelligent identification method for rail damage takes a long time in the process
of collecting rail damage data, resulting in a longer intelligent identification time for rail damage. Therefore, an active
intelligent identification method for rail damage based on ultrasound was proposed. Based on ultrasonic technology, the
echo signal and rail damage echo signal were obtained according to the gain range, and the filter of ultrasonic transducer was
set to convert the signal to collect data. According to the DBSCAN algorithm, the smallest unit for damage identification
was used as a unit to combine the collected ultrasonic data to form an ultrasonic information group, and the data was divided.
The rail damage features were matched according to the divided data. Finally, Using the AlexNet network architecture as
the main identification structure, based on data matching results, a dataset of displayed samples was established to conduct
a precise search for damage types, thus obtaining active intelligent identification results for rail damage. The experimental
results showed that the method proposed in this paper had a shorter time for identifying active rail damage, and can achieve
rapid identification of active rail damage, which has good application value.
Key words: ultrasonic; steel rail; intelligent identification of damage; AlexNet network architecture; DBSCAN
algorithm; filter

随着铁路技术的快速发展，轨道负载逐渐增加， 伤出现的概率也会同时增加 。对主动式钢轨伤损
[1]
主动式钢轨在长期使用过程中会受到列车产生的长 进行智能识别，有利于对钢轨损伤进行判断并及时
期碰撞等因素影响，裂纹的扩展速度会逐渐增大，核 采取措施，从而避免经济损失和人员伤亡。因此，钢
轨伤损智能识别是保证铁路安全运维的重要保障，
[2]
一直以来都是重要的研究课题 。
收稿日期：2023-08-07
目前主动式钢轨伤损的智能识别主要有以下几
作者简介：尹段泉（1981—），男，本科，工程师，主要研究方向为
[3]
钢轨探伤和钢轨焊接技术 种思路。曾楚琦等 基于光纤光栅，采用可变轨距
通信作者：尹段泉，yinduanquan@163. com 优化了钢轨探伤车的检测速度，并通过影响耦合关
49
2024 年 第 46 卷 第 6 期
无损检测