基于人工神经网络的土石坝沉降组合模型

ARTIFICIAL NEURAL NETWORKS-BASED COMPOSITE MODEL FOR EARTH-ROCKFILL DAM SETTLEMENT

影响土石坝沉降的主要因素是时效变形,目前常用的统计模型中时效分量主要采用指数模型和对数模型,但在某些工程实际中两子模型的拟合值与实测值偏差较大。文中在统计子模型基础上研究了一种基于人工神经网络的土石坝沉降组合模型,用于对土石坝沉降的拟合和预测,并给出实例。结果表明,组合模型较好地发挥了神经网络的非线性映射能力,并比统计子模型的拟合效果好。

Time-dependent deformation is the main factor influencing earth-rockfill dam settlement. For the time-dependent component, usually, the exponential form or logarithmic form is adopted in the statistical model. But there is a large difference between the fitted value and the measured value if either of the two forms is used. So a composite model is established based on artificial neural networks. An example shows that the composite model demonstrates high fitting and prediction accuracy.