THE APPLICATION OF ARTIFICIAL NEURAL NETWORKS IN THE INITIAL DESIGN PROCESS OF STRUCTURES

ABSTRACT

This paper presents the application of a backpropagation neural network model in the preliminary design of reinforced concrete beams with the aim of introducing neural networks application in structural design. Artificial neural networks are algorithms used for cognitive tasks, such as learning and optimization. They possess the ability to learn and generalize from examples without being explicitly programmed. In this paper, the mathematical model for the optimum design of a simply supported reinforced concrete beams presented by (Chakrabarty, 1992) is used to calculate the optimum design of beams under different configurations. These design examples are then used to train a neural network, which is designed specifically for the optimum design of beams. Then, a set of new design values generated with the same mathematical model are presented to the network in order to validate the network’s performance and to demonstrate its generalization properties. The neural network showed good generalization properties as it was able to predict the various design parameters to a fair degree of accuracy. The parallel processing of many input parameters at once was a major advantage of the neural-network approach. It means that the network was able to predict the output much more quickly than a mathematical optimizer, which uses procedural programming, and this saves a lot of time. It is the writer’s hope that with growing awareness, it would be possible to incorporate ANN in the design process of structures