砂土加筋挡墙筋条临界长度的神经网络预测

对于加筋支挡结构的设计,神经网络模型不同于基于凝聚力基础之上的半经验公式,它不需要主观的人为假设,而是模拟人脑思维,通过数据样本的学习来获得预测结果.BP神经网络是对非线性可微分函数进行权值训练的分层网络.文中采用BP网络对给定极限荷载下砂土挡墙筋条的临界长度进行预测,试验点几乎分布在预测曲线附近,说明网络学习是成功的.对不同筋材的极限荷载与临界长度的关系进行了对比分析,表明筋材的弹性模量及筋材与填料之间的摩擦系数对加筋性能有着重要的影响.

Prediction of critical length of reinforcement for reinforced sand retaining wall by neural network

For design of geosythetics-reinforced retaining wall, neural network is different from the semi-empirical formula based on the theory of homo-cohesion. It simulates the function of human brain and obtains predicted result from the learning of experimental data, without any kind of subjective assumption. BP(back-propagation) neural network is a kind of delamination network which trains weights of nonlinear and differential function. Based on MATLAB neural network toolbox and model test data, a BP neural network was designed to predicted the critical length of reinforcement under given ultimate load. Test data spotted near the curve linked with predicted data, so we know that the learning of network is successful. Furthermore, we analyzed the relationship between ultimate load critical length of different reinforcement. The result showed us that the elastic modulus of reinforcement and the frictional coefficient between reinforcement and filling materials are of great importance to the whole performance of geosythetics-reinforced retaining wall.