摘要
针对神经网络用于基坑变形预测存在结构难确定、训练易陷入局部最优及易过学习等问题,以已有的周边地表沉降为样本,利用最小二乘支持向量机(LS-SVM)建立基坑地表沉降预测模型,应用网格搜索算法优化模型参数,对基坑周边地表沉降进行连续滚动的多步预测。实例结果表明,LS-SVM用于基坑周边地表沉降预测效果较好,具有所需数据少、推广能力强等优点。
In light of the problems in forecasting of the deformation of foundation pit in neural network, which include difficult structure determination, practice＇s easily trapping into local optimum, and overfitting phenomenon, with the existing ground settlement as a sample, use is made of LS-SVM to establish forecasting models for ground settlement of foundation pit. The Gridding search algorithm is used to optimize the model parameters and make rolling multistep predictions on ground settlement. The results of example show that LS- SVM-based forecasting for ground settlement of foundation pit is of better effect.
出处
《南宁职业技术学院学报》
2013年第4期89-92,共4页
Journal of Nanning College for Vocational Technology
基金
国家自然科学基金项目<砂土中静压群桩压入和加载过程中桩与桩之间相互作用机理研究>研究成果
(50978086
51178165)
