Two-phase genetic algorithm for size optimization of free-form steel space-frame roof structures

• Two-phase genetic algorithm for minimum weight design of free-form roof structures.
• Structure is subjected to nonlinear implicit and constraints of AISC LRFD code.
• Two real-life roof structures are presented to be built as train stations in Canada.

A two-phase GA approach is presented for minimum weight design of free-form steel space-frame roof structures consisting of discrete commercially available rectangular hollow structural sections (HSS). The new methodology is applied to two roof structures subjected to the AISC LRFD code and ASCE-10 snow, wind, and seismic loadings. They are two of the thirteen train stations making up the Ottawa Light Rail Transit (OLRT) system to be completed in Ottawa, Canada, in 2018. Both examples have a diamond grid pattern and their members are subjected to torsion in addition to bending and axial forces. The initial design in both cases is an actual design performed in a design office by the first author iteratively using general-purpose structural analysis software over a period of days. The optimum solutions obtained using the new methodology resulted in savings of 12% and 7% for the two examples. The advantages of the two-phase GA optimization algorithm presented in this paper are three fold: a) automation of the design process of a complicated and one-of-a-kind structure; b) relieving the designer of days of iterative design process; and c) achieving a lighter and therefore more economical design.