软土地基实测沉降的拟合和预测

为有效利用软土地基，基于实测资料综合介绍了沉降拟合和预测的几种方法，并应用于多个实际工程的实测沉降分析．同时，对拟合预测结果进行后验差检验和关联度分析，使其具有统一的评价标准．结果表明，用人工神经网络法来拟合沉降曲线效果最好，用前期沉降数据训练网络来预测后期沉降效果也非常好；传统曲线拟合法中，双曲线法的预测效果最好；用新陈代谢的灰色预测法来预测后期沉降，预测精度随预测模型的更新优化而很快提高．

Fitting and prediction of observational settlement for soft ground

Several methods are studied for fitting or prediction of observational settlements. The total settlements of two practical projects are analyzed. The back-check and the correlation degree method are used to discuss the results of fitting and prediction so that all these methods can be evaluated by using an unified standard. It is shown that the artificial neural network will be a successful way to fit the whole settlement curve. The results of the artificial neural network forecast are in good coincidence with the observed results. The hyperbolic curves fitting method is considered as the most effective one among traditional methods. When the gray system is used to predict settlements of later period, the precision will be improved with the renovation of the model.