钢丝绳内外部断丝损伤识别

Damage identification of broken wires inside and outside the wire rope

  • 摘要: 针对钢丝绳断丝损伤检测以外部断丝损伤检测为主,对内部断丝损伤检测的研究较少且内外部断丝识别精度不高的问题,提出了一种钢丝绳内外部断丝损伤识别方法。通过钢丝绳损伤径向漏磁检测器采集钢丝绳断丝损伤产生的漏磁信号;采用双密度双树复小波变换对漏磁信号进行降噪处理;通过设置自适应阈值提取降噪信号的时域特征,同时提取原始漏磁信号的频域特征;采用基于类间距离和互信息的方法进行特征选择,即先对所有特征进行归一化处理,剔除标准差较大及类间距离较小的特征,然后计算特征之间的互信息,排除包含损伤信息较为相似的特征,最后计算特征中区分度最差的2种损伤类型,并从剔除的特征中收回这2种类型类间距离最大的特征;将保留的特征融合作为最优特征子集并输入BP神经网络进行分类识别。测试结果表明,该方法能识别钢丝绳内外部断丝损伤且识别准确率达97.8%。

     

    Abstract: The wire rope broken wire damage detection is mainly focused on the outside broken wire damage detection, not on the inside broken wire damage detection. Moreover, the inside and outside broken wire identification accuracy is not high. In order to solve the above problems, this paper proposes a wire rope inside and outside broken wire damage identification method. The magnetic flux leakage signal generated by the wire rope broken wire damage is collected by the wire rope damage radial magnetic flux leakage detector. The double-density dual-tree complex wavelet transform is used to reduce the noise of the magnetic flux leakage signal. By setting an adaptive threshold, it is able to extract the time domain characteristics of the noise reduction signal and extract the frequency domain characteristics of the original magnetic flux leakage signal at the same time. A method based on the distance between classes and mutual information is used for characteristics selection. Firstly, all characteristics are normalized to eliminate characteristics with large standard deviation and small distance between classes. Secondly, the mutual information between characteristics is calculated to exclude characteristics with similar damage information. Thirdly, the two damage types with the worst discrimination among the characteristics are calculated, and the characteristics with the largest distance between these two types are retrieved from the eliminated characteristics. The retained characteristics are fused as the optimal characteristic subset and input into the BP neural network for classification and recognition. The test results show that the method can identify broken wire damage inside and outside the wire rope with a recognition accuracy of 97.8%.

     

/

返回文章
返回