Abstract: With the improvement of the cemented sand and gravel (CSG) dam construction technology, demands for the construction of 100-meter CSG dams are increasing; An urgent need is to study their different profile forms and cementation contents. This paper uses the finite element method to formulate a regression equation for the prediction model of a CSG dam response surface. Using this equation as the objective function, the artificial bee colony algorithm is applied to optimize the CSG dam profile form and the gelling content, and a series of optimal profile solution sets are obtained. Finally, a set of dam models so optimized and a real dam project is selected to test experimentally for comparison and validation. The results show that in the loading process, the damage modes of the two groups of dam models are different. The real project model cracks at the contact surface of the dam body and foundation, gradually growing into a through crack, and its dam heel suffers shear crack damage. The optimized model suffers slip instability damage but no obvious structural damage to the dam body is observed, indicating that its structural strength is enhanced significantly. In comparison with real projects under normal working conditions, the optimized model reduces horizontal displacement by 36.9% and vertical displacement by 25.5%, and increases the overload safety factor by 9.1%, respectively. The results lay a basis for the profile forms of CSG dams and would help simplify the design method of 100-meter CSG dams.

Key words: response surface method, artificial bee colony algorithm, cemented sand and gravel dam, model test, profile optimization