Optimum design of cold-formed steel space structures using neural dynamics model

Abstract Optimization of space structures made of cold-formed steel is complicated because an effective reduced area must be calculated for members in compression to take into account the non-uniform distribution of stresses in thin cold-formed members due to torsional/flexural buckling. The effective area varies not only with the level of the applied compressive stress but also with its width-to-thickness ratio. For statically indeterminate structures a new effective area has to be calculated for each member in every iteration of the optimization process. As such, the constraints are implicit, non-smooth, and discontinuous functions of design variables. The patented robust neural dynamics model of Adeli And Park (U.S. patent 5,815,394 issued on September 29, 1998) has been adapted for optimum design of space trusses made of commercially available cold-formed shapes in accordance with AISI specification. A CPN network was developed to learn the relationship between the cross-sectional area and dimensions of cold-formed channels. The model has been used to find the minimum weight design for several space trusses commonly used as roof structures in long-span commercial buildings and canopies, including a large structure with 1548 members with excellent convergence results.