Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Computational Science, Engineering & Technology Series
ISSN 1759-3158
Edited by: B.H.V. Topping, Y. Tsompanakis
Chapter 7

Challenges and Strategies in using Genetic Algorithms for Structural Identification

C.G. Koh and T.N. Thanh

Department of Civil Engineering, National University of Singapore, Singapore

Full Bibliographic Reference for this chapter
C.G. Koh, T.N. Thanh, "Challenges and Strategies in using Genetic Algorithms for Structural Identification", in B.H.V. Topping, Y. Tsompanakis, (Editors), "Soft Computing in Civil and Structural Engineering", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 7, pp 203-226, 2009. doi:10.4203/csets.23.7
Keywords: dynamics, system identification, damage detection, genetic algorithm.

Significant damage in a structure is often manifested through changes in physical properties, such as decrease in structural stiffness and shift of natural frequencies. If not monitored and rectified early, damage may compromise the performance of structures, increase maintenance cost and, in the unfortunate event, result in structural failure. From the viewpoint of functionality and safety, it is therefore essential to have means of detecting and quantifying structural damage. Structural damage identification has now become a vital component of an emerging engineering discipline known as structural health monitoring. Computationally this typically involves structural identification to determine key parameters and their changes based on the numerical analysis of measured responses. Its feasibility for practical implementation has been enhanced greatly as a result of rapid advances in hardware technology in terms of both instrumentation and computational capabilities. To make this work for real structures, it is essential to have a good numerical strategy to accurately and efficiently quantify system characteristics with limited and noisy measurements.

This presentation focuses on the use of genetic algorithms (GA) to tackle the numerical challenges associated with identification of large structural systems. Three novel strategies working on different principles are presented to improve accuracy and reduce computational time associated with the combinatorial explosion problem. While the first strategy aims to reduce the identification problem size, the other two are designed to improve the search capability of GA. In the first strategy, two divide-and-conquer approaches, including substructural and progressive identification, are developed to deal with large structural systems. This strategy works by dividing a large structural domain with many unknowns into smaller domains with a manageable number of unknowns that are more accurately and efficiently identified by adopting GA. The advantage of this divide-and-conquer strategy is its ability to identify independently some specific parts of the structure without considering the rest of the structure. This is especially helpful for local damage detection in some critical parts of a large structural system.

The next two strategies incorporate GA with some compatible complementary methods to improve the accuracy and efficiency of identification results. A hybrid GA and local search strategy combines the merits of powerful global search of GA and fine-turning capability of local search, thereby accelerating the convergence towards the optimal solutions. The numerical study, performed on a fairly large 50-DOF structural system with incomplete and noisy measurements, indicates the hybrid strategy gives much better results than the GA without local search. In a different way to enhance the search performance, an adaptive strategy of search space reduction, incorporating an improved GA based on migration and artificial selection into the search space reduction method, is developed. This strategy works on multiple populations or "species" allowing for both broad and local search capability to be conducted simultaneously and the search space is reduced adaptively according to the statistics information of identified results. Numerical studies demonstrate the significant improvement of this improved GA strategy over standard GA. The method is also extended to output-only identification with a numerical example of three coupled buildings subjected to seismic excitation.

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 £88 +P&P)