Parametric Analysis of the “Optimal” Seismic Retrofit of RC Frames

Seismic retrofitting of existing structures can be pursued by adopting either member-level (e.g. RC member confinement) or structural-level (e.g. the introduction of steel bracings) techniques. Sometimes, they can be combined with the aim to exploit the main advantages of each of them. However, no well-established and consistent design rule exists to drive engineers in selecting the “best” combination of those techniques. In fact, this task is a typical optimization problem, although it is not generally regarded in this light. Therefore, recently developed Artificial Intelligence (AI) techniques may help to find the “fittest” retrofit solution, with regard to one predefined optimization criterion (e.g. the initial cost of intervention). Specifically, a Soft-Computing approach (SC), formulating a Genetic Algorithm (GA) procedure for an “optimal” retrofit of existing Reinforced Concrete (RC) structures, has been recently proposed. At each iteration of the procedure, a “population” of possible solutions is defined; each solution is defined by an “individual phenotype”, which includes both its member-level (namely, the confinement of single columns using FRP layers) and its structural-level (mainly, the realization of a concentric steel bracing system) interventions. Then, the three main genetic operators (selection, crossover and mutation) are handled to simulate the driving mechanisms of the evolution of species, resulting in the “survival of the fittest” rule. As a consequence, at each iteration the cost-effectiveness of the technically-admissible solutions is increased. In the present paper, a parametric study on a RC structure based on a similar GA procedure is reported, varying some relevant “genetic algorithm parameters”, which concern the application of the above-cited genetic operators, leading to quite different results characterized by relevant differences.