关键词: 拱坝位移, 黏弹性滞后变形性态, HHST模型, SVM, 组合监控模型, 锦屏一级拱坝

Abstract: A displacement monitoring model should well interpret and predict the deformation behavior of arch dam. HHST model could explain the viscoelastic hysteretic deformation behavior of Jinping-I arch dam. To further improve the prediction accuracy of the HHST model, the nonlinear relationship between the finite element method (FEM)-calculated viscoelastic hysteretic displacement of arch dam and its causal factors is modeled by the support vector machine (SVM) and is used as a whole variable in the HHST model. In subsequence, a combinatorial monitoring model is established for the displacement of arch dam based on multiple linear regression (MLR). Case study of the Jinping-I arch dam shows that the prediction accuracy of the combinatorial monitoring model, which has a reduced number of input factors, is significantly higher than that of simple models directly established with all the 18 causal factors of the HHST model. SVM has a better prediction accuracy for the hysteretic hydraulic displacement than that of constrained least square method-based linear regression model. The two combined monitoring models, respectively using the SVM and linear regression-based hysteretic hydraulic displacement component, have similar interpretation ability for the measured deformation behavior of arch dams, while the former can effectively improve the prediction accuracy of dam displacement, with the average mean square error(MSE) of multiple monitoring points dropping by 21.67% and the average determination coefficient R² rising by 0.07%.

Key words: displacement of arch dams, viscoelastic hysteretic deformation behaviour, HHST model, support vector machine, combinatorial monitoring model, Jinping-I arch dam