Zoned embankment dams are widely used because of their economic and hydraulic advantages. Yet, the stiffness contrast between the clay core and surrounding rockfill may induce transverse arching, reducing core stresses and increasing hydraulic fracturing susceptibility. This study aimed to develop explicit predictive equations for the minimum transverse arching ratio at the upstream side of the core and at the core centerline under end-of-construction conditions. A structured numerical database consisting of 288 two-dimensional plane-strain finite element analyses was generated by varying dam height, section geometry, clay-core elastic modulus, and Poisson’s ratio. Based on this database, Gene Expression Programming (GEP) and Multiple Linear Regression Analysis (MLRA) were used to derive predictive formulations. The GEP equations showed high predictive accuracy, with overall performance metrics of R2 = 0.9486, RMSE = 0.0200, MAE = 0.0159, and MAPE = 2.9797% for Rmin,ups, and R2 = 0.9644, RMSE = 0.0163, MAE = 0.0132, and MAPE = 1.7993% for Rmin,core. In both responses, normalized filter thickness emerged as the dominant variable. Compared with MLRA, the GEP formulations performed better, particularly for the upstream-side response. The proposed equations provide a practical alternative to repeated, time-consuming numerical analyses in preliminary designs and safety assessments.
Topçu et al. (Tue,) studied this question.