嵌岩桩竖向承载力预测的遗传BP神经网络模型研究  

The study of the model of genetic algorithm and back propagation neural network for predicting the vertical bearing capacity of rock-socketed pile

作　　者：王昆明[1] 蒋洪胜[1] 姜千君 

机构地区：[1]山东建筑大学土木工程学院,山东济南250101 [2]济南同圆建筑设计研究院有限公司,山东济南250001

出　　处：《山东建筑大学学报》2008年第6期529-534,共6页Journal of Shandong Jianzhu University

摘　　要：将遗传算法的全局搜索能力和BP神经网络的高次非线性能力,应用于嵌岩桩竖向承载力研究中。两种算法的结合,避免了BP神经网络收敛于局部极值,使神经网络更快找到全局最优解;并利用济南地区的现场实测资料,建立起了预测嵌岩桩单桩竖向承载力的遗传BP神经网络模型。通过该模型可以建立起高应变动力测试法与静载荷试验之间的内在联系,证明了高应变动力测试法在本地区的可靠性较高。Based on the global searching capability of genetic algorithm and nonlinear capability of the back propagation neural networks, a new prediction model was founded for the prediction of vertical bearing capacity of rock-socketed piles. The model was trained with the in-situ pile testing data in Jinan region. With the model , the interior relationship between high strain dynamic testing method and static loading test was established. It was demonstrated that high strain dynamic testing method has high reliability in this area.

关 键 词：嵌岩桩 高应变动测 静载荷试验 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]