Keywords: identification, bridge, genetic algorithms, nonlinear model, parallel selection, hybrid strategy.

This work deals with parameter identification in dynamic response. The principle of this process is to determine different structural parameters of a dynamic system based on numerical analysis of measurement of input and the corresponding output. The structure we try to identify is a bridge that corresponds to its real model of 1:50. This structure is subject to a seismic transversal load as an input. The objective of this study is placed in the continuity of a previous work of parameter identification of a linear model that had been performed by the first author at al. [1].

Among the difficulties to surmount in parameter identification, we have the nature of the objective functions (error function) to minimize which is highly multimodal. In our case, this function represents the distance of the vector contained the experimental time displacements from the vector containing the corresponding analytical time displacements. Another aspect of the identification is also the high number of parameters, the scale and the unknown ranges of variation of each parameter to identify.

The above points make the common search methods based on the steepest gradients not efficient to locate the global solution. Those methods need in general one started solution that must be close to the global search landscape. To overcome those difficulties, we have developed a hybrid algorithm that uses a heuristic search method based on the genetic algorithm [2,3] with parallel selection GAPS coupled with a local search method. The parallel selection consists in introducing in the operator of selection one or several individuals coming from other selection processes. This allows the different processes to exchange information related to the area of search for a better exploration of the landscape. This idea, presents the following advantages: first, it decreases the number of evaluations of the objective function and second, all the processes converge in the same area. The efficiency of the algorithm developed is demonstrated through an example of a structural model of a bridge. In this example twelve parameters of the mathematical model will be identified.

1

J.L. Zapico, M.P. González, M.I. Friswell, C.A. Taylor, A.J. Crewe, "Finite Element Model Updating on a Small Scale Bridge", Journal of Sound and Vibration, 268 (5), 993-1012, 2003. doi:10.1016/S0022-460X(03)00409-7

2

J.H. Holland, "Adaptation in natural and artificial systems". Ann Arbor: University of Michigan Press, 1975.

3

D.E. Goldberg, "Genetic algorithms in search, optimisation, and machine learning". New York: Addison-Wesley, 1989.

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 £135 +P&P)