【摘要】 支撑轴力监测是深基坑工程施工中一级基坑的必选项，支撑轴力的动态变化实时反映着基坑整体的安全状态。传统人工监测方法存在监测频率低、无法及时预警的问题，无人化、自动化监测的趋势在所难免。利用长短期记忆网络（long short-term memory,LSTM）深度学习模型，可预测基坑施工全过程的钢筋混凝土支撑轴力的变化。支撑轴力受多种因素共同影响，通过分析轴力全过程数据，结合施工日志信息，分析施工工况、挡墙测斜位移、立柱回弹、环境温度对轴力的影响，并将上述因素作为变量输入LSTM预测模型。试验结果表明：多变量LSTM能够有效预测混凝土支撑轴力变化；测斜及立柱数据对轴力预测效果提升显著，随着输入变量种类的增多，预测精度、稳定性随之增长。研究结果为基坑工程的安全检测与风险预警提供了科学依据。更多还原

【Abstract】 Monitoring support axial force is a mandatory requirement for first-class foundation pits during deep excavation construction, and dynamic changes in axial force reflect the overall safety status of the foundation pit in real time. Traditional manual monitoring methods suffer from low frequency and inability to provide timely warnings, making unmanned and automated monitoring inevitable. Utilizing the long short-term memory(LSTM) deep learning model, changes in reinforced concrete support axial force throughout foundation pit construction are predicted. Axial force is influenced by multiple factors. Entire process data of axial force, combined with construction log information, are analyzed to assess the effects of construction conditions, retaining wall inclinometer displacement, column rebound, and environmental temperature on axial force, with these factors subsequently input as variables into the LSTM prediction model. Experimental results show that the multivariable LSTM effectively predicts changes in reinforced concrete support axial force; inclinometer and column data significantly enhance prediction accuracy, with prediction accuracy and stability increasing as the variety of input variables grows. The results provide a scientific basis for safety monitoring and risk warning of foundation pit engineering.更多还原