提出了一种检测油气井套管变形段位置与缺陷类型的涡流检测方法，该方法根据电磁检测成像测井仪输出涡流信号，计算其特征信号及波峰，再采用自适应阈值判别变形段。针对每一个变形段进行差分处理，并将连续多组数据合并构成多层数据，对所有数据进行Fisher降维，选择降维后的多维数据作为特征值。 采用神经网络方法进行样本训练和测试数据的识别，试验结果表明：该涡流检测方法能识别出所有变形段的位置及缺陷类型，具有很好的实际工程应用价值。

A eddy current detection method for position and defect type of casing deformation section of oil and gas well is proposed. According to the eddy current signals outputted by electromagnetic flaw detection imaging tool, the peak and features of signal are calculated, and an adaptive threshold is adopted to judge the deformation section. Differencial method is applied to process each deformation section. Successive groups of data are fusioned into multi-layer data. All data are dimensionally reduced by Fisher algorithm, and the dimensionally reduced multidimensional data are selected as the eigenvalue. The neural network algorithm is used for sample training and test data recognition. The experimental results show that the algorithm can identify the location of different deformation sections and the defect types in the deformation sections. The algorithm has certain practical value for applications.