基于人工神经网络的软土地基沉降预测技术研究

A study on settlement prediction of soft soil foundation based on artificial neural network

针对软土地基大变形沉降预测,引入人工神经网络预测技术,通过增加权值调整时的动量修正量对传统BP模型的算法进行改进,使其更加适应对于软土地基沉降的预测。以天津滨海地区围海造陆工程为例,选择吹填过程中吹填土的原始标高、含水量、塑性指数以及吹填土地基加固后的土体含水量和强度为输入量对吹填土地基的沉降量进行预测。预测结果显示预测值与实测值误差在15%以内。

According to the basic feature of soft ground settlement, a new artificial neural network-based（ANN）approach was presented to predict settlement of soft foundation in this paper. The actual prediction ability of BP network model was improved, and training velocity was increased by adding momentum factor and using the weight control calculation in the BP network model. Taking Tianjin Binhai area reclamation engineering as an example, withinitial elevation, water content, plasticity index of dredger fill, soil strength and water content after reinforcement asinputs, the settlement of filled soil was predicted. Forecast results indicate that predicted values are closer to measured values.