Structural identification of a tied arch bridge using parallel genetic algorithms and ambient vibration monitoring with a wireless sensor network

Abstract

Structural identification has received increased attention as an applied technique for performance-based assessment of large civil structures by providing a means to improve the correlation of simulated responses in numerical models to experimental measurements of actual behavior under in-service conditions. This paper presents the application of structural identification to a large set of modal parameter estimates obtained through ambient vibration monitoring of a tied arch bridge with a wireless sensor network facilitating high-rate, real-time vibration measurement over 48 measurement channels. Model updating of a finite element model of the span is achieved through global optimization of an objective function using an integer-constrained genetic algorithm implemented on a parallel computing cluster to facilitate the use of large population sizes. The influence of the number of modes included in the objective function and the number of uncertain parameters included in the optimization are explored for this real-world application. The results highlight the capability of the genetic algorithm to achieve an exceptionally strong correlation between the calibrated finite element model and the experimentally measured natural frequencies and mode shapes over a large set of modal parameter estimates. However, variations observed in the parameter solutions, identified as the number of uncertain parameters updated and modes included in the objective function are varied, serve to highlight the challenges associated with reliable parameter estimation in structural identification using classical optimization-based approaches.