基于改进完备集成经验模态分解的钢丝绳缺陷漏磁检测方法

Steel wire rope defect magnetic flux leakage detection method based on improved complementary ensemble empirical mode decomposition

钢丝绳小缺陷信号往往被淹没在股波噪声中,存在钢丝绳小缺陷检测困难、易漏检等问题。针对该问题,提出了一种基于改进完备集成经验模态分解(ICEEMD)的钢丝绳缺陷漏磁检测方法。为了避免钢丝绳表面润滑剂或尘埃对检测信号造成影响,采用电磁检测法。将ICEEMD、小波阈值滤波(WTF)、维纳滤波(WF)相结合,得到ICEEMD-WTF-WF多级降噪方法:通过ICEEMD分解钢丝绳漏磁信号,得到本征模态函数(IMF)分量;计算IMF分量的能量比、排列熵、互相关系数,取出IMF趋势分量和IMF股波噪声分量,并对股波噪声分量进行WTF,筛选有用的IMF分量重构信号;对重构后的信号进行WF,去除随机噪声。提取降噪后的缺陷特征值,输入BP神经网络并进行训练,识别钢丝绳缺陷漏磁信号。实验结果表明:ICEEMD-WTF-WF多级降噪方法对钢丝绳漏磁信号具有良好的降噪效果,信噪比、峭度指标优于WTF、移动平均滤波和WF;基于ICEEMD-WTF-WF的BP神经网络模型检测耗时短,对小缺陷的平均准判率达到98.13%,能较好地满足钢丝绳缺陷检测要求。

The signal of small defects in steel wire rope is often submerged in wave noise.Therefore,it is difficult to detect small defects in wire rope and easy to miss detection.In order to solve this problem,a steel wire rope defect magnetic flux leakage detection method based on improved complementary ensemble empirical mode decomposition(ICEEMD)is proposed.To avoid the influence of the lubricant or dust on the surface of the wire rope on the detection signal,the electromagnetic detection method is adopted.CEEMD-WTF-WF multi-stage noise reduction method is obtained by combining ICEEMD,wavelet threshold filtering(WTF)and Wiener filtering(WF).The intrinsic mode function(IMF)component is obtained by decomposing the magnetic flux leakage signal of steel wire rope through ICEEMD.The energy ratio,permutation entropy and cross-correlation coefficient of IMF components are calculated.The IMF trend component and IMF stock noise component are extracted.WTF is conducted on the stock noise component to filter the useful IMF component reconstruction signal.WF is applied to the reconstructed signal to remove random noise.The eigenvalues of the de-noised defects are extracted,input and trained by BP neural network.The magnetic flux leakage signals of the steel wire rope defects are identified.The experimental results show that ICEEMD-WTF-WF multi-stage noise reduction method has good noise reduction effect on the magnetic flux leakage signal of steel wire rope.The SNR and kurtosis indexes are better than those of WTF,moving average filter and WF.The BP neural network model based on ICEEMD-WTF-WF takes a short time to detect.The average accuracy rate of small defects reaches 98.13%,which can better meet the requirements of wire rope defect detection.