基于Gumbel Copula函数的堆石坝沉降模型因子优选研究

为克服常规沉降模型因考虑众多影响因子造成欠拟合及预测精度不高的缺点,首先由因子优选准则得到沉降模型的预选因子集,运用了Copula熵和PMI(偏互信息)两种方法对预选因子集进行优选,再将优选因子集引入沉降模型并计算典型测点的沉降值,验证了该优选方法的可行性。实例表明,基于Gumbel函数的沉降模型拟合预测精度优于常规模型,具有较高的工程指导意义及较好的推广价值,可运用于堆石坝的变形预测。

Optimization of the Factors for Rockfill Dam Settlement Model Based on Gumbel Copula Function

In order to overcome the shortcomings of under fitting and low prediction accuracy caused by considering many influence factors in the conventional settlement model,the preselected factor set of settlement model is firstly obtained by the factor optimization criteria and two methods of Copula entropy and partial mutual information(PMI)are used to optimize the preselected factor set,and then the optimized influence factor set is introduced into the settlement model to calculate the settlement values of typical measurement points,which verifies the feasibility of the method.The example shows that the accuracy of settlement model fitting prediction based on Gumbel function is better than that of the conventional model,which has higher engineering guiding significance and better popularization value,and can be applied to the deformation prediction of CFRD.