A Bayesian Damage Identification Technique Using Evolutionary Algorithms - a Comparative Study





Structural Health Monitoring, Power Spectral Density, Severity, Bayesian, Evolutionary Algorithm


In this paper, a one-stage model-based damage identification technique based on the response power spectral density of a structure is investigated. The technique uses a finite element updating method with a Bayesian probabilistic framework that considers the uncertainty caused by measurement noise and modelling errors. The efficacy of two different evolutionary algorithms – a genetic algorithm and a covariance matrix adaptation evolution strategy – is examined via numerical simulation of time-history response data for a beam structure. A range of different damage scenarios have been considered including: both single and multiple damage locations; varying damage severity; the introduction of noise and modelling errors and incompleteness in the number of captured modes and measurement response data. The results clearly show that both evolutionary algorithms implemented are effective and their overall performance, measured in terms of accuracy, is very similar. However, the covariance matrix strategy is found to be significantly superior in terms of its convergence rate and the number of function evaluations required to find the solution for both noisy and noise-free response data.


Download data is not yet available.




How to Cite

M. Varmazyar, Haritos, N. and M. Kirley (2015) “A Bayesian Damage Identification Technique Using Evolutionary Algorithms - a Comparative Study”, Electronic Journal of Structural Engineering, 14(1), pp. 1–19. doi: 10.56748/ejse.141851.