Design representation and the shape of the pareto front in evolutionary multiobjective structural design

The paper presents results of computational experiments in which the impact of design representations on the performance of an evolutionary multiobjective structural design processes was investigated. Specifically, two classes of design representations (i.e., direct representations and generative representations) were used to minimize the total weight and the maximum horizontal displacement of steel structural systems in tall buildings. In the reported experiments, the Strength-Pareto Evolutionary Algorithm 2 was used to determine the shapes of Pareto fronts for this two-objective design optimization problem and the impact of the design representation on the results produced. The obtained results have shown that the type of the design representation has a significant impact on the shape and location of the Pareto front in this complex problem domain. Generally, direct representations produce Pareto fronts with better coverage (spread) than generative representations. Generative representations, however, produce additional regions of the Pareto front which cannot be found using direct representations.