Fault identification is an important aspect of seismic data interpretation and a key step in structural interpretation. Traditional fault identification involves manual marking by geological interpreters, which is not only time-consuming and inefficient but also prone to human error. A deep learning–based fault identification method, which uses an attention mechanism to focus on target features, is proposed to address the aforementioned issues and increase the accuracy of fault identification. A convolutional block attention module (CBAM) is used in the decoding layer of the U-Net network, and a ResNet-50 residual block is utilized in the encoding layer. Thus, a fault identification method based on convolutional neural networks, which is referred to as Res–CBAM–UNet, is established. Data augmentation on synthetic seismic data and their corresponding fault labels was performed, and the model was trained using the newly generated training dataset as the input to enhance the generalization capability of the network model. Subsequently, the model was compared and analyzed using CBAM–UNet, ResNet34–UNet, and ResNet50–UNet networks and tested using seismic data from actual working areas. Results indicate that the designed Res–CBAM–UNet network has good fault identification performance, with high continuity in fault identification and high computational efficiency.

断层识别是地震资料解释的重要内容，也是构造解释的关键步骤。传统的断层识别需要地质解译人员手动标记，不仅费时、低效，而且容易出现人为错误。为了解决上述问题，提出一种基于深度学习的故障识别方法，利用注意力机制关注目标特征，提高故障识别的准确性。U-Net 网络的解码层使用卷积块注意模块（CBAM），编码层使用 ResNet-50 残差块。由此，建立了一种基于卷积神经网络的故障识别方法，简称Res-CBAM-UNet。对合成地震数据及其相应的断层标签进行数据增强，并使用新生成的训练数据集作为输入来训练模型，以增强网络模型的泛化能力。随后，利用CBAM-UNet、ResNet34-UNet和ResNet50-UNet网络对模型进行了比较分析，并利用实际工作区的地震数据进行了测试。结果表明，设计的Res-CBAM-UNet网络具有良好的故障识别性能，故障识别连续性高，计算效率高。和 ResNet50-UNet 网络，并使用实际工作区域的地震数据进行测试。结果表明，设计的Res-CBAM-UNet网络具有良好的故障识别性能，故障识别连续性高，计算效率高。和 ResNet50-UNet 网络，并使用实际工作区域的地震数据进行测试。结果表明，设计的Res-CBAM-UNet网络具有良好的故障识别性能，故障识别连续性高，计算效率高。