GPU based optimization of steel frame structure

Frame construction systems are one of the most frequently used structural systems in structural engineering. In recent years, thanks to advances in computer and optimization methods, studies on optimization of these structural systems are increasing rapidly. Since the optimization of frame systems is complex and irregular, meta-heuristic techniques are often preferred to solve such problems. Meta-heuristic methods allow for the solution of such complex problems, but they contain many repetitive processes . Therefore, the solution time can be extended. In addition, although Finite Element Method (FEM), which is frequently used in the analysis of frame structure systems, has a wide range of solution capability and diversity , serious computing times are required especially for the solution of large scale structures. When these two situations are evaluated together, it takes a lot of time for the optimization of large-scale frame systems. Reducing these times to reasonable levels is possible by running GPU (Graphical Processing Unit) processors, which can be found on standard computers, in parallel with programming. In this thesis, an optimum design program which solves steel frame systems in accordance with GPU architecture is developed. As the optimization method, Biogeography Based Optimization (BBO) method was used. GPU parallel programming structure is used in all FEM method. The design and optimization parts of the developed program are written according to the CPU (Central Processing Unit) architecture. In order to compare, a separate version of the program, which does not include parallel processing in CPU architecture, has been prepared. Two different design examples were created according to both versions, which were optimized 48 times for different load combinations and computer characteristics. According to the results obtained, it is demonstrated that GPU processors are more successful in FEM operations than CPU processors and the efficiency of optimization algorithm is revealed.