Improved optimal design of concrete gravity dams founded on anisotropic soils utilizing simulation-optimization model and hybrid genetic algorithm

Abstract

ABSTRACT Incorporating numerically predicted responses of seepage characteristics based on large numbers of surrogate models to identify the optimum design of Concrete Gravity Dams (CGD) can lead to sub-optimal or local solution, even when the evolutionary Genetic Algorithm (GA) optimization solver is utilized. The optimization task aims to find the safe and minimum cost design for the CGD built on anisotropic permeable soils, considering the effect of seepage under the CGD. In each iteration of the optimization model solution, the GA-based linked Simulation-Optimization (S-O) approach evaluates the objective function and the constraints based on the responses of the Support Vector Machine (SVM) surrogate models. This paper focused on improving the GA performance to find the global or a near global optimum design of a complex optimization model, minimizing the construction cost, and providing a design safety of the CGD. This could be achieved by using the Hybrid Genetic Algorithm (HGA) optimization solver. The HGA merges two optimization search techniques: the direct search methods utilizing GA, and the gradient search method using the Interior Point Algorithm (IPA). The solution results showed that the HGA was more efficient in finding improved or, global optimum design of the CGD. Physically, the low anisotropic ratio of hydraulic conductivity (0.1, 0.3, 0.5) resulted in critical seepage characteristics, which substantially affects the optimum hydraulic design of CGD, and increases the corresponding construction cost.