Optimization of Large Scale Three Dimensional Reticulated Structures Using Cellular Genetics and Neural Networks

Abstract

In this work, the input for large space structures is created using Formian [1]. In this paper, a new optimization technique called Cellular Automata (CA) has been combined with Genetic algorithm (GA) to develop a different search and optimization algorithm known as Cellular Genetic Algorithm (CGA), which considers the areas of the members of the space structures as discrete variables. The objective function for each cell in Cellular automata is obtained either by using structural analysis package like FEAST (ISRO 1995), ANSYS or SAP or by using neural networks. Initially to obtain the data for training the neural network structural analysis package is used. The use of Neural Network is motivated by time consuming repeated analysis required by Cellular Genetics during Optimization process. In addition, a multilevel optimization approach is implemented by reducing the size of the search space for individual design variables in each successive level of optimization process. The non- linear load deflection behavior of the optimized structure is also studied. The numerical tests presented demonstrate the computational advantages of the proposed approach “Cellular Genetic Algorithm” combined with Neural Networks (NN) become pronounced for large scale Optimization problems.