神经网络在软基沉降预测中的应用

软土本身具有很多特性,沉降一直是很重要的土工问题,通过沉降预测可为软基工程的设计与施工提供方法上的支持。以人工神经网络理论为基础,利用神经网络具有的自组织、自适应、容错性和较强的学习、联想能力,通过对数据样本的训练学习和测试,反演软土地基力学参数,并结合有限元程序建立模型预测沉降。结果表明,预测数据与实测数据误差小于10%。说明该方法预测精度较高,通过BP网络反演地基参数结合有限元计算预测沉降的方法是合理可行的。

Application of neural network to settlement prediction of soft soil foundation

Soft soil itself has many characters,and the geotechnical settlement is always one of the very important problems.This paper provides a method for soft soil foundation design and construction through the settlement prediction.Based on the artificial neural networks(ANN) theory,by taking advantage of ANN's self-organizing,adaptive identifying,self-studying and being tolerant towards errors characteristics,through studies,training and tests of the sample data,and back analysis parameters of soft soil foundation,the settlement prediction model is thus established by finite element method.The comparison of prediction data and measured data indicates that the model can make a precise forecast,and it is proved reasonable and feasible.