摘要
桥梁群桩基础承载性能数值计算结果与岩土参数的选取密切相关,为探讨更为符合实际工程地质情况的群桩基础受力计算模式,采用MATLAB编制的改进BP神经网络与遗传算法相结合的遗传神经网络计算程序,建立参数反演算法。该方法首先利用改进的BP神经网络对正交设计的样本数据进行训练,然后采用遗传算法进行最优求解。基于戛洒江特大桥深厚嵌岩超长桩试桩静载试验的数据,以嵌岩段中风化页岩和中风化砂岩的体积模量Gn和剪切模量Gs为反演参数,以数值计算为正分析,获得合理的参数值并应用于深厚嵌岩群桩基础受力分析。结果表明:本文提供的获取深厚嵌岩群桩基础受力分析参数的反演算法,可以获取合理的岩土参数,较好地用于该桥深厚嵌岩群桩基础受力分析,对群桩基础分析具有参考意义。
The calculated load bearing capacity of group pile foundation is closely related to the geotechnical parameters selected.To work out the calculation mode of load bearing capacity for group pile foundation that is applicable for specific geotechnical conditions,a parameter inversion method is proposed on the basis of a genetic neural network that combines the upgraded BP neural network compiled by MATLAB and the genetic algorithm.In this method,the sampling data of orthogonal design are first trained by the upgraded BP neural network,and then treated by the genetic algorithm to obtain the optimal solutions.The data of static load tests of the very long rock-socketed piles of Jiasajiang River Bridge are used for analysis.The volume modulus G n and shearing modulus G s of moderately decomposed shale and sandstones in the rock-socket section are taken as inversion parameters,and using the numerical calculation as positive analysis to obtain the rational parameters.The obtained parameters are in turn used to analyze the load bearing behavior of the deep rock-socketed group pile foundation.The results show that the parameter inversion algorithm method proposed to analyze the load bearing behavior of the deep rock-socketed group pile foundation can generate rational geotechnical parameters.And the engineering practice of Jiasajiang River Bridge is a good case for load bearing analysis of group pile foundation.
作者
张良翰
杨荣双
杨劲屾
ZHANG Liang-han;YANG Rong-shuang;YANG Jin-shen(Yunnan Infrastructure Co.,Ltd.,Kunming 650501,China;Yuxi Dajia Expressway Investment Construction Development Co.,Ltd.,Yuxi 653100,China)
出处
《世界桥梁》
北大核心
2021年第2期96-100,共5页
World Bridges
关键词
深厚嵌岩桩
群桩基础
参数反演
BP神经网络
遗传算法
仿真计算
静载试验
受力性能
deep rock-socketed pile
group pile foundation
parameter inversion
BP neural network
genetic algorithm
simulation calculation
static load test
load bearing behavior
