基于BP神经网络的静压桩承载力时间效应预测

Prediction of time effect of static pile bearing capacity by BP neural network

基于珠海软土地区3根PHC管桩隔时复压试验数据,采用BP神经网络建立了静压桩承载力时间效应的BP神经网络模型来预测静压桩的长期承载力。在建模过程中将桩长、桩截面积、土体摩擦角、土体变形模量、渗透系数、最终压桩力及休止期等与静压桩承载力密切相关的7个参数引入到输入层,用Visual Basic语言编制了以最终压桩力和休止期为主要输入因素的计算程序,程序可以对比显示计算和实测曲线。在样本训练和学习过程中,任意选取2根桩的试验数据来预测第3根桩的长期承载力。通过对施工现场工程桩的试算,预测结果与实测值较为吻合,表明提出的BP神经网络模型用于预测静压桩长期承载力是切实可行的。

Based on the cycle loading test data of 3 PHC piles in soft ground in Zhuhai, a BP neural network model of time effect of jacked pile bearing capacity is established to predict the long-term bearing capacity of piles. In the process of modeling, the parameters closely related to the jacked pile bearing capacity including pile length, pile section area, soil friction, soil permeability coefficients, deformation modulus, the final jacking pressure and intermission time are introduced into the input layer. A calculation program is compiled with Visual Basic language taking the final jacking pressure and intermission time as the main input factors. Both calculation and measured curve could be displayed by the program. Two pile testing data are chosen at random to predict long-term bearing capacity of the 3 rd pile in the training and learning. The results of prediction coincide with measured values of a trial pile. The BP neural network model is proved feasible to predict long-term bearing capacity of jacked pile.