通过构建磁记忆信号采集系统，检测高速铁路钢轨用U71Mn钢试件法向磁记忆信号，对信号进行小波包变换与Hilbert变换，得到法向磁记忆信号包络，求取梯度特征值；提出磁记忆信号的小波包能量谱分析方法，通过对有中心小孔人工缺陷的U71Mn钢试件进行拉伸试验，采集不同载荷下试件表面的法向磁记忆信号，以小波包分解频带的能量作为试件应力集中的判定特征值，根据小波包不同空间能量谱中能量的大小及分布，判断试件的应力集中部位及应力集中程度。结果表明:该方法对钢轨用U71Mn钢的检测具有较好的可行性,可望在磁记忆信号的计算机自动识别方面具有工程应用价值。

Magnetic memory testing system is developed to acquire normal direction magnetic signals from the U71Mn specimen. Wavelet packet transform and Hilbert transform is used to get the envelop curve of the signals so that gradient eigenvalues are picked effectively; By different loads stretching experiments on specimens U71Mn steel specimens with artificial hole defect, normal direction magnetic memory signals can be acquired by the testing system，the magnetic signals energy of the frequency domain based on wavelet packet decomposition is taken as the feature information of stress concentration, and the amount and the distribution of the energy in different wavelet package energy spectrum spaces indicate the stress concentration level. The results show that this method has a good feasibility for the detection of U71Mn steel rail, and it is expected to have certain engineering application value in the computer automatic recognition of magnetic memory signals.