Keywords: artificial neural networks, back propagation error algorithm, laminated plates, finite element.

Research in artificial neural networks has recently active as a new means of information processing. ANN try to mimic the biological brain neural networks into mathematical model [1]. From 1946 to 1960, a movement attempt to carry out interdisciplinary research on the brain and computers to solve the basic principles of intelligent information processing. Until 1970, there is no significant progress in neural networks. In the 1970's, Werbos originally developed a back-propagation algorithm. In the mid 1980's, the back-propagation algorithm as the learning algorithm of the feed-forward neural network was also rediscovered by Parker and Rumelhart et al [1]. Until now, researches was focused on the application of neural networks in solving of the problems in engineering sciences. In civil engineering, the methodology has been successfully applied to a number of area. Some typical application in civil engineering include structural damage detection [2], active control of structures [3], reinforced concrete columns capacity predication [4], cost analysis of structures [5], proportioning of concrete mixes [6], prediction of damping in structures [7].

This study explores the learning capacity of neural networks to map nonlinear relationship between a laminated plate dimension and its deflection. In this paper, the back propagation error algorithm is introduced and used for calculating the laminated plate deflection. To do this, 11 artificial neural network models has been used to prediction of the deflection in the middle of plate. The network used could be trained to any desire level (the lowest tried being RMS=0.0079). Use of the ANN model has been found to be very effective for correctly prediction of laminated plate deflection less than a second with no requirement of complicated finite element analysis. However, as the lowest testing Root RMS was 0.0206, it is suggested that a better prediction might be made by varying factors such as number of hidden layers, number of PEs in each layer, input and output scaling, normalization techniques, initial learning rate, momentum and weights, training algorithm and data representation. These possibilities might be most effectively searched by a genetic algorithm. Furthermore, the generation of more data might allow the net to generalize better and thus predict more accurately over a large application domain.

1

Fukuda, Toshio, and Shibata, Takanori, "Theory and Application of Neural Networks for Industrial Control Systems", IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, 39(6), 472-489, 1992. doi:10.1109/41.170966

2

Wu, X., Gaboussi, J., and Garrett, J. H., "Use of Neural Networks in Detection of Structural Damage", Computers & Structures, 42(4), 649-639,1992. doi:10.1016/0045-7949(92)90132-J

3

Tang, Yu, "Active Control of SDF System Using Artificial Neural Network", Computers & Structures, 60(5), 695-703, 1996. doi:10.1016/0045-7949(95)00438-6

4

Chuang, P.H., Anthony, Goh, T.C , and Wu, X., "Modeling the Capacity of Pin Ended Slender Reinforced Concrete Columns Using Neural Networks", ASCE Journal of Structural Engineering, 124(7), 830-838, 1998. doi:10.1061/(ASCE)0733-9445(1998)124:7(830)

5

Tam, C.M., and Fang, Clara, "Comparative Cost Analysis of Using High-Performance Concrete in Tall Building Construction by Artificial Neural Networks", ACI Structural Journal, 96(6), 927-936, 1999.

6

Oh, Ju-Won, Lee, In-Won, Kim, Ju Tae, and Lee, Gyu-Won, "Application of Neural Networks for Proportioning of Concrete Mixes", ACI Material Journal, 96(1), 61-67, 1999.

7

Li, Q.S., Liu, D.K., Fang, J.Q., Jeary, A.P., and Wong, C.K., "Damping in Buildings: its Neural Network Model and AR Model", Engineering Structures, 22, 1216-1223, 2000. doi:10.1016/S0141-0296(99)00050-4

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £82 +P&P)