Neuron-Genetic Optimization of Double Layer Steel Grids

Analysis and design of space structures are normally time consuming since a large number of members is involved. The problem becomes more involved when optimization is performed, where numerous analysis should be carried out. In this paper, neural network and genetic algorithm are applied for optimization of double steel layer grids. 180 models with three topologies, spans between 10 and 75 m and height between 1 and 2.5 m are analyzed and designed for optimum weight. The results have been used for training and testing of proposed neural networks system to predict the weight of the structures. The genetic algorithm based on neural networks is used for design optimization. Thus, for a site plan with arbitrary length and width, the output of the neuron-genetic system is the selection of a double steel grid with optimized shape based on three available topologies with height, distance between columns and length of horizontal members.