Evolutionary Developmental System for Structural Design

Contents

This paper discusses the results of extensive computational experiments involving evolutionary developmental representations for structural design. In particular, it describes developmental representations based on one-dimensional cellular automata applied to three complex structural design problems in the domain of steel structural systems in tall buildings. In the experiments conducted with the developmental encodings, several of their key parameters were tested, including the impact of the configuration of the design embryo and the importance of the symmetry constraint on the quality of produced designs. The obtained results were also compared to the results of evolutionary design experiments involving traditionally used direct representations. These comparisons focused primarily on the compactness and evolvability properties of both types of design representations. The results have shown that developmental encodings produce quantitatively better results for the majority of the investigated design problems and also generate structural designs with distinct shaping patterns.