Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Civil-Comp Proceedings
ISSN 1759-3433
CCP: 82
Edited by: B.H.V. Topping
Paper 38

Design Optimization of Offshore Platforms using Genetic Algorithms and Wave-Net

M.J. Fadaee and M. Besharat

Civil Engineering Department, Shahid Bahonar University of Kerman, Iran

Full Bibliographic Reference for this paper
M.J. Fadaee, M. Besharat, "Design Optimization of Offshore Platforms using Genetic Algorithms and Wave-Net", in B.H.V. Topping, (Editor), "Proceedings of the Eighth International Conference on the Application of Artificial Intelligence to Civil, Structural and Environmental Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 38, 2005. doi:10.4203/ccp.82.38
Keywords: genetic algorithm, approximation, wavelet theory, neural network, wave-net, optimum design, offshore, platform.

In this paper, the genetic algorithm (GA) method combined with wavelet neural networks has been used for design optimization of offshore jacket platforms. In optimum design of the offshore jacket platforms, the dynamic nature of the applied forces results in a time consuming analysis and so, a lengthy design optimization process. In such cases, approximate analysis is used. For approximate analysis, the neural network method is a very common method and normally gives acceptable results.

In this work, the weight of the platform has been adopted as the objective function. The maximum tension and compression stresses in the elements, and the maximum displacement at the top nodes have been taken as constraints. The maximum stress in each element and the maximum nodal displacements has been calculated in every design iteration with a good approximation using a trained wave-net. The cross sectional areas of the standard pipes have been chosen as discrete variables. The earthquake load is the governing dynamic applied load.

In order to use a neural network as a fast analysis tool in the optimization process, first it is necessary to train the network based upon a set of input-output pairs which are selected from the search space randomly. Obviously, the network application must be controlled after training and before using, to make sure that the output accuracy is within the acceptable range.

After confirming the accuracy of the network and during the structural optimization process, there is no need to re-analyze the structure, and the network is used as a fast analysis tool in the next iterations. Infact, the network has the duty of performing a fast analysis of the structure as a sub-program [1]. Artificial neural networks may have several layers. The multi-layer networks are more powerful than the single-layered ones. The layer whose output is the network output is called the output layer. The other layers are called hidden layers [2].

In this paper, the wave-net method has been established by combining back propagation (BP) neural networks and the wavelet theory. In order to indicate the capability of the method, an offshore jacket platform has been optimized using the genetic algorithm and accurate analysis first; then the same structure has been optimized using genetic algorithm method and, for approximate analysis, the wavelet neural network. It is found that the wave-net method decreases the optimization process time significantly with no serious affect on the results. At the end, the efficiency of the proposed method has been indicated using several numerical examples.

M. Besharat, "Optimum design of offshore platforms using wavelets for approximating the response of dynamic loads", M Sc. Thesis, Civil eng. Dept., Shahid Bahonar University of Kerman, Kerman, Iran, 2005.
M.T. Hagan, H.B. Demuth and M.H. Beale "Neural network design", Boston, MA: PWS Publishing, 1996.

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 £80 +P&P)