Automated topology design of high-rise diagrid buildings by genetic algorithm optimization

摘要

In this study, in order to determine the outer shell geometry of high-rise diagrid structures by means of minimum structural material utilization, simplified strength- and stiffness-based design methods are encoded in MATLAB with the help of SAP2000 OAPI (Open Application Programming Interface) codes and genetic algorithm (GA). The computer code is designed to get most of the inputs from the building model in SAP2000, to have minimal manual inputs that define diagrid shell geometry. The code is also built so that the engineer evaluates the optimum diagrid geometry for any rectangular plan building, which is modeled in SAP2000 software. Three different buildings, which consist of 30, 60, and 90 stories, are modeled in SAP2000 without diagrid shells on facades. Based on these models, the diagrid shell geometries are analyzed for both uniform angle (UA) and varying angle (VA) geometries. A different geometry approach to VA models is proposed, consisting of symmetry and continuum for the load distribution and asymmetry, achieving minimal weight. In order to evaluate the optimum VA topology, GA, is used. Predetermined dead, live, and wind loads are applied to the models based on minimum design loads according to American Society of Civil Engineers (ASCE) 7-10, with corresponding load factors to be able to get more realistic results. Based on the resultant module diagonal cross-sectional areas, an algorithm is developed and encoded in order to determine minimum required pipe section sizes according to buckling design of compression members as stated in American Institute of Steel Construction (AISC) 360-10, to give the engineer an approximate idea of what the required sections should be. Additionally, the feasibility of the optimum models is compared by complexity index parameters. The results of the analyses show that optimum VA models are more efficient than UA models concerning the diagrid shell's total weight. However, UA models are more feasible, according to complexity index values.