OPTIMIZATION OF HIGH-PERFORMANCE CONCRETE STRUCTURES BY VARIABLE NEIGHBORHOOD SEARCH

This paper describes a methodology in designing high-performance concrete for simply supported beams, using a hybrid optimization strategy based on a variable neighborhood search threshold acceptance algorithm. Three strategies have been applied to discrete optimization of reinforced concrete beams: Variable Neighborhood Descent (VND), Reduced Neighborhood Search (RNS) and Basic Variable Neighborhood Search (BVNS). The problem includes 14 variables: two geometrical; one material type; one mix design; and 10 variables for the reinforcement setups. The algorithms are applied to two objective functions: the economic cost and the embedded CO2 emissions. Firstly, this paper presents the application of these three different optimization strategies, which are evaluated by fitting the set of solutions obtained to a three-parameter Weibull distribution function. The Variable Neighborhood Descent with Threshold Accepting acceptance strategy algorithm (VND-TA) results as the most reliable method. Finally, the study presents a parametric study of the span length from 10 to 20 m in which it can be concluded that economic and ecological beams show a good parabolic correlation with the span length.