【摘要】 为了避免面板浇筑后堆石体的大量或不均匀变形,最新提出通过对堆石体填筑施工初期的监测资料进行整理和分析,对比邓肯模型、双屈服面弹塑性模型及清华非线性K-G模型在堆石坝应用的优缺点。结合施工初期的实测信息,确定在特定条件下最能反映堆石体真实情况的本构模型,利用BP神经网络结合遗传算法反演模型的最优参数,再正分析计算预测的堆石体施工期的沉降变形情况。将预测得到的沉降变形资料与堆石体填筑完成的实测沉降变形资料进行对比分析,以预测的堆石体沉降变形情况为参考数据,结合工程的实际要求,合理安排预留沉降期,进而确定合理的面板浇筑时机,避免堆石体较大的前期沉降变形对面板带来的不利影响,有效改善面板受力和变形情况。更多还原

【Abstract】 In order to avoid large or non-uniform deformation of rockfill after face slab placement,it is proposed to analyze and sort out the early monitoring data of rockfill filling,and compare the advantages and disadvantages of Duncan model,two-yield surface elastic-plastic model and Tsinghua non-linear K-G model applied in rockfill dams.Based on the measured data at early construction,the constitutive model reflecting actual status of rockfill in specific conditions is determined.In combination with genetic algorithm,the BP neural network is used to back calculate the optimal parameters of the model,then analyze and calculate the predicted rockfill settlement and deformation during construction.By analysis and comparison of the predicted settlement and deformation with the measured settlement and deformation of the completed rockfill,and taking the predicted settlement and deformation of rockfill for reference,the settlement period is rationally prepared in advance and face slab placement time is reasonably determined so as to prevent negative influence of large rockfill settlement and deformation on the face slabs and improve the stress and deformation of face slabs.更多还原