基于遗传算法改进BP神经网络的桩基横向承载力预测  

作　　者：李云峰 闫思行 冉斌斌 陈涛 秦玮 张小龙 

机构地区：[1]重庆交通大学数学与统计学院,重庆400074 [2]重庆交通大学机电与车辆工程学院,重庆400074 [3]重庆交通大学河海学院,重庆400074 [4]重庆交通大学经济与管理学院,重庆400074

出　　处：《科技创新与应用》2024年第33期30-33,共4页Technology Innovation and Application

摘　　要：针对传统BP神经网络在桩基横向承载力预测中存在的局限性,如易陷入局部最优和收敛速度慢等问题,该文提出一种基于遗传算法(GA)改进的BP神经网络模型。该模型利用遗传算法优化初始权重和偏置,以提高预测精度和模型泛化能力。选取影响桩基横向承载力的关键因素作为输入参数:桩径、荷载的偏心距、桩入土深度及土的不排水抗剪强度。通过训练与测试,对比分析传统BP神经网络模型和基于遗传算法改进的BP神经网络模型的预测效果。结果表明,GA-BP模型在测试集上的相对误差平均值降低至2.53%,明显优于BP模型的6.44%。此外,GA-BP模型未出现过度拟合现象,表明其在捕捉数据潜在模式和泛化新样本方面表现出色。综上所述,基于遗传算法优化的BP神经网络为横向受荷桩承载力的准确预测提供一种有效途径,对于工程实践具有一定的指导意义和应用价值。Aiming at the limitations of traditional BP neural network in predicting the lateral bearing capacity of pile foundations,such as easy to fall into local optimization and slow convergence speed,this paper proposes an improved BP neural network model based on genetic algorithm(GA).The model uses genetic algorithm to optimize initial weights and offsets to improve prediction accuracy and model generalization ability.Select key factors affecting the lateral bearing capacity of pile foundation as input parameters:pile diameter,load eccentricity,pile penetration depth and undrained shear strength of soil.Through training and testing,the prediction effects of traditional BP neural network model and improved BP neural network model based on genetic algorithm are compared and analyzed.The results show that the average relative error of the GA-BP model on the test set is reduced to 2.53%,which is significantly better than the 6.44%of the BP model.In addition,the GA-BP model did not show overfitting,indicating that it performed well in capturing potential patterns in data and generalizing new samples.To sum up,the BP neural network optimized based on genetic algorithm provides an effective way to accurately predict the bearing capacity of laterally loaded piles,and has certain guiding significance and application value for engineering practice.

关 键 词：桩基横向承载力 遗传算法 BP神经网络 承载力预测 桥梁工程 

分 类 号：TU473[建筑科学—结构工程]