摘要
Abstract
The application of steel support axial force servo systems in deep foundation pit engineering has not yet been widely adopted,and field measurement data is particularly scarce.Based on 2 212 measured data points,5 machine learning methods(radial basis function neural networks,backpropagation neural networks,k-nearest neighbour algorithms,support vector machines,and random forests)are employed for analysis.The dataset includes one output parameter(support axial force)and seven input parameters(support vertical position,support planar position,time,temperature,soil density,cohesion,and internal friction angle).The model performance was eval-uated by 3 metrics of root mean square error,correlation coefficient,and mean absolute error.The results indicate that the backpropagation neural network performed best on the test dataset,with the root mean square error,corre-lation coefficient,and mean absolute error values of 71.8 kN,0.991 8,and 56.2 kN,respectively.This confirms the potential of machine learning methods for the precise prediction of support axial forces,providing effective sup-port for safety control and construction planning,which is crucial for the safety and stability of deep foundation pit engineering.关键词
深基坑开挖/钢支撑结构/支撑轴力/机器学习/时间/温度Key words
deep foundation pit excavation/steel support structure/support axial force/machine learning/time/temperature分类
建筑与水利
