Optimization of Large-Scale 3-D Trusses Using Evolution Strategies and Neural Networks

Abstract

The objective of this paper is to investigate the efficiency of optimization algorithms, based on evolution strategies, for the solution of large-scale structural optimization problems. Furthermore, the structural analysis phase is replaced by a neural network prediction for the computation of the necessary data for the evolution strategies (ES) optimization procedure. The use of neural networks (NN) was motivated by the time-consuming repeated analyses required by ES during the optimization process. A back propagation algorithm is implemented for training the NN using data derived from selected analyses. The trained NN is then used to predict, within an acceptable accuracy, the values of the objective and constraint functions. The proposed methodology has been applied in sizing structural optimization problems of large-scale three dimensional roof trusses. The numerical tests presented demonstrate the computational advantages of the proposed approach which become more pronounced for large-scale optimization problems.