Computational & Technology Resources
an online resource for computational,engineering & technology publications |
||||||

Civil-Comp Proceedings
ISSN 1759-3433 CCP: 78
PROCEEDINGS OF THE SEVENTH INTERNATIONAL CONFERENCE ON THE APPLICATION OF ARTIFICIAL INTELLIGENCE TO CIVIL AND STRUCTURAL ENGINEERING Edited by: B.H.V. Topping
Paper 62
BP and RBF Neural Networks for Predicting Displacements and the Design of Schwedler Domes A. Kaveh and M. Raeissi Dehkordi
Department of Civil Engineering, Iran University of Science and Technology, Tehran, Iran A. Kaveh, M. Raeissi Dehkordi, "BP and RBF Neural Networks for Predicting Displacements and the Design of Schwedler Domes", in B.H.V. Topping, (Editor), "Proceedings of the Seventh International Conference on the Application of Artificial Intelligence to Civil and Structural Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 62, 2003. doi:10.4203/ccp.78.62
Keywords: Schwedler domes, design, neural networks, radial basis functions, back propagation.Summary
Space structures are often used for covering large spans. Sport stadiums,
assembly halls, exhibition centers, swimming pools, shopping arcades and
industrial buildings are typical examples of structures where large unobstructed
areas are essential and where minimum interference from internal support is
required. Space structures are often classified as grids, domes and barrel vaults [1].
Configuration processing and data generation for space structures can be simplified using the concepts from formex algebra [2]. For preprocessing and for efficient solutions, theory of graphs can be employed. Though an optimal analysis can be carried out using usual matrix displacement method, however, for large-scale problems, the analysis is usually time consuming due to a large number of equations being involved [3]. Therefore approximate methods are beneficial in the process of nonlinear analysis and optimal design. Neural networks provide a powerful tool for approximate analysis and design of space structures. Such networks are trained using backpropagation [4] and counterpropagation networks [5,6], among many others.
In the present paper, efficient neural networks are trained for prediction of the
displacements and design and of Schwedler domes using The model selected for the study of this paper is a single layer Schwedler dome with ribbed members are considered as beam elements joined together rigidly and diagonal members are taken as bar elements, Figure 62.1. The spans are varied between 42.5m and 65m with the steps of 2.5 meters. The opening angle of the domes is chosen from the angles between 45 and 67.5 degrees. A typical Schwedler dome used in this study is shown in Figure 62.1(a). Simple supports are considered for the exterior ring of the dome. The number of intervals between the ribs is considered as 20 and for rings this number is taken as 8.
In order to simplify the design, and due to symmetry, 8 types of cross sections are considered for the members of a typical rib. The span and deflection at the apex of the dome are illustrated in Figure 62.1(b). References
- 1
- Makowski, Z.S., Analysis, Design and Construction of Braced Schwedler domes, Nichols Pub. Co., New York, 1984.
- 2
- Nooshin H., Disney P. and Yamamoto C., Formian, Multi-Science Publishing Company, UK, 1993.
- 3
- Kaveh A., Optimal Structural Analysis, Research Studies Press (John Wiley), UK, 1997.
- 4
- Kaveh, A. and M. Raeissi Dehkordi, "BP and RBF Neural Networks for the Analysis and Design of ribbed domes", International Journal of Space Structures, accepted for publication, 2003. doi:10.1260/026635103322437463
- 5
- Hajela P. and Berke L., "Neurobiological computational models in structural analysis and design", Computers and Structures, 41(1991)657-667. doi:10.1016/0045-7949(91)90178-O
- 6
- Iranmanesh A. and Kaveh A., "Structural optimisation by gradient base neural networks", International Journal of Numerical Methods in Engineering, 46(1999)297-311. doi:10.1002/(SICI)1097-0207(19990920)46:2<297::AID-NME679>3.3.CO;2-3
- 7
- NeuralWare, Using NeuralWorks, NeuralWare, USA, 1993.
purchase the full-text of this paper (price £20)
go to the previous paper |
||||||