为预测软土开挖对既有隧道的影响，确保隧道安全正常运营，建立了一种基于人工神经网络（artificial neural network, ANN）模型的隧道水平位移预测模型，用于预测临近开挖引起的隧道变形。该方法使用有限元模型生成的183组数值模拟数据作为数据集，按照70%训练集、15%测试集和15%验证集划分，涵盖7个输入参数（如开挖宽度、开挖深度、挡土墙厚度等），并以隧道水平位移为输出参数，基于多层感知结构，采用反向传播算法训练ANN模型，分析影响隧道水平位移的关键因素。结果表明，所建立的ANN模型能够准确预测开挖引起的隧道水平位移，具有较高的可行性。本文中提出的ANN模型可为软土中临近开挖引起的隧道变形预测提供有效工具，并为实际工程中的风险评估和减灾决策提供科学依据。

In order to predict the impact of excavation in soft soil on existing tunnels and ensure the safe and normal operation of the tunnels, a tunnel horizontal displacement prediction model based on ANN (artificial neural network) was established to forecast the deformation caused by nearby excavations. A total of 183 sets of numerical simulation data generated from a finite element model were used as the dataset, which was randomly divided into 70% training set, 15% testing set, and 15% validation set. The model incorporated seven input parameters (including excavation width, excavation depth, and retaining wall thickness) and considered tunnel horizontal displacement as the output variable. ANN model was trained using a multilayer perceptron structure with a backpropagation algorithm, enabling analysis of the key factors affecting tunnel horizontal displacement. The results indicate that the established ANN model can accurately predict the horizontal displacement of tunnels caused by excavation, demonstrating high feasibility. ANN model proposed provides an effective tool for predicting tunnel deformation due to nearby excavation in soft soil and offers a scientific basis for risk assessment and disaster reduction decision-making in practical engineering.