Optimization of concrete shells using genetic algorithms

In structural design, structures are often modeled using the finite elements method (FEM). One of the most common element type is the shell, which is used to model surfaces in a three dimensional space as far as the surface thickness is smaller than the other two dimensions. Many decisions must be taken during the design process, and many physical and loading constraints must be satisfied. Designers are generally interested in providing a solution that respects all the problem constraints, without trying to further improve it. In fact optimization is not trivial, even if it could yield a huge benefit both from the economic and the construction point of view. Additionally, saving materials is one of the fundamental criteria for the sustainable approach to the design. In this paper, we address the Skew Reinforcement Design in Reinforced Concrete Two Dimensional Elements (SRD2D). It consists of determining the minimum reinforcement required to respect all the constraints given by the geometric properties and the internal actions working on it. As this problem is strongly nonlinear and non-convex it cannot be easily solved using exact methods, while heuristics and meta-heuristics are suitable to this purpose. We propose a Genetic Algorithm (GA) and an enhanced version of it, a memetic algorithm, in which we apply an intensification heuristic to the solution obtained by the GA, where each variable is separately optimized by applying a first improvement based on local search strategy. We report computational results showing both the effectiveness of the proposed method, and the benefit of combining GAs with intensification methods.