基于遗传—BP神经网络预测单桩竖向承载力  

Prediction of vertical bearing capacity of single pile with genetic algorithm and back propagation neural networks

作　　者：蒋洪胜[1] 戚靖骅[1] 万立华[2] 

机构地区：[1]山东建筑大学土木工程学院,山东济南250101 [2]济南市建筑工程质量监督站,山东济南250013

出　　处：《山东建筑工程学院学报》2006年第3期197-200,共4页Journal of Shandong Institute of Architecture and Engineering

摘　　要：在遗传算法改进BP神经网络的基础上,利用遗传算法的全局优化能力和BP神经网络的较强的高次非线性能力和自学习能力,建立预测单桩竖向承载力的遗传—BP神经网络模型。通过利用济南地区现场试桩资料对该模型进行训练和检验,证明本文所建立的遗传—BP神经网络预测模型预测精度高,适用性强,可以作为桩基工程设计和理论研究的参考计算工具。Based on the back propagation neural networks improved by genetic algorithm, a new prediction model is founded for prediction of the vertical bearing capacity of single pile. The model is trained with in-situ pile testing data in Jinan region. It is shown that the predicted value of vertical bearing capacity of single pile has a good agreement with the site tested value. The founded new model could be used for both practical design and theoretical study.

关 键 词：遗传算法 BP神经网络 单桩 竖向承载力 

分 类 号：TU473.1[建筑科学—结构工程]