基于Copula-RF的混凝土坝变形监测模型

Concrete Dam Deformation Monitoring Model Based on Copula-RF

混凝土坝的变形与各种环境量之间存在复杂的非线性映射关系,传统统计模型中预报因子的选择有一定的主观性,且易出现过拟合问题.针对上述问题,本文采用Copula函数对变形影响因子进行非线性相关检验,以确定最优因子集.在此基础上,结合泛化能力强的随机森林(RF)理论,建立了基于Copula-RF的混凝土坝变形监测模型.随后,以某混凝土重力坝为例,根据大坝实测数据验证Copula-RF模型的准确性,通过与基于经典最小二乘回归的混凝土坝变形预测模型分析结果比较,根据均方根误差、平均绝对误差及平均绝对百分比误差等指标评价模型的拟合、预测精度.计算结果表明,相比于采用最小二乘回归的混凝土坝变形预测模型,Copula-RF模型预测精度更高,性能更加稳定,为混凝土坝变形监测提供了一种新方法.

There is a complex nonlinear mapping relationship between the deformation of concrete dam and various environmental quantities.The choice of predictors in traditional statistical models is subjective and prone to over-fitting problems.To solve the problems,a concrete dam deformation monitoring model based on Copula-RF is proposed to determine the optimal set of factors.Firstly,the Copula theory and RF model are introduced.Then,the Copula function is used to perform nonlinear correlation test on the predictive factors of the deformation monitoring model to objectively select the forecast factor set.Finally,the random forest(RF)model is used to realize the concrete dam.Taking the monitoring data of a gravity dam as an example,the above model is validated in contrast to traditional least square regression method,and the accuracy of the model is evaluated by means of root mean square error(RMSE),mean absolute error(MAE)and mean absolute percentage error(MAPE).The results show that the Copula-RF model has a much higher prediction accuracy and also a better stability.It provides a brand-new method for concrete dam deformation monitoring.