Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Computational Science, Engineering & Technology Series
ISSN 1759-3158
Edited by: Y. Tsompanakis and B.H.V. Topping
Chapter 1

Mastering Computationally Demanding Problems in Mechanics

M. Papadrakakis, Ch.Ch. Mitropoulou and N.D. Lagaros

Institute of Structural Analysis and Seismic Research, School of Civil Engineering, National Technical University of Athens, Greece

Full Bibliographic Reference for this chapter
M. Papadrakakis, Ch.Ch. Mitropoulou, N.D. Lagaros, "Mastering Computationally Demanding Problems in Mechanics", in Y. Tsompanakis and B.H.V. Topping, (Editor), "Soft Computing Methods for Civil and Structural Engineering", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 1, pp 1-16, 2011. doi:10.4203/csets.29.1
Keywords: metaheuristics, neural networks, incremental dynamic analysis, fragility analysis, vertical statistics.

Structural analysis methods were traditionally based on rigorous scientific procedures that are formed on mathematical methods and the principles of theoretical mechanics and are solved with the implementation of numerical simulation methods based on discretized continua. However, three decades ago new families of computational methods, denoted as soft computing (SC) methods, were proposed. These methods are based on heuristic approaches rather than on rigorous mathematics. Despite the fact that these methods were initially received with suspicion, their use in various areas of engineering science is continuously growing. Neural networks (NN), genetic algorithms and fuzzy logic are the most popular approaches of SC. An NN can store experimental knowledge and make it available for later use. It features adaptive learning, self-organizing capability during training and fault imprecision during applications. The main advantage of using NNs is that it can deal with problems that do not have an algorithmic solution or for which an algorithmic solution is too complex to be found. Over the last ten years NN have emerged as a powerful alternate tool for replacing time consuming numerical procedures in many engineering applications. Among others, NNs have been used for the identification of nonlinear dynamic systems and damage assessment. Predictions of the structural behaviour by neural networks have been employed in the context of design optimization and structural reliability analysis. Furthermore, NNs have been applied to the prediction of the structural response under static or dynamic loading conditions.

A soft computing based framework for the fragility assessment of three-dimensional buildings is proposed in this work. The computational effort required for a fragility analysis of structural systems can become excessive, far beyond the capability of modern computing systems, especially when dealing with real-world structures. For the purpose of making attainable fragility analyses, a NN implementation is presented here, which can provide accurate predictions of the structural response at a fraction of the computational time required by a conventional analysis. The main advantage of using NN predictions is that they can deal with problems, without having an algorithmic solution or with an algorithmic solution that is too complex to be found. The proposed methodology is applied to three-dimensional reinforced concrete buildings where it was found that with the proposed implementation of a NN a reduction of one order of magnitude r is achieved in the computational effort for performing a full fragility analysis.

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

go to the previous chapter
go to the next chapter
return to the table of contents
return to the book description
purchase this book (price £85 +P&P)