Design of double layer grids using backpropagation neural networks

In this paper efficient neural networks are trained for design of double la yer grids. Square diagonal-on-diagonal grids with spans varying between 26. 5 and 75 m are considered. Backpropagation algorithm is employed for traini ng efficient neural networks for evaluation of the maximum deflection, weig ht, and design of double layer grids. A special method is developed for dat a ordering to reduce the nonlinearity of the data and to increase the speed of training. This approach also provides the necessary stability. Addition al neural nets are trained and tested for design of grids using the develop ed data ordering. (C) 2001 Civil-Comp Ltd. and Elsevier Science Ltd. All ri ghts reserved.