Bayesian structural identification of a suspension bridge using temperature and traffic loading

Large civil infrastructure such as long suspension bridges are critical connecting elements, whose behaviour is difficult to discern and should be constantly monitored for safety. Critical factors relevant for this purpose usually are non-directly measurable, having therefore to be inferred, based on monitored data. This structural identification is usually highly susceptible to uncertainties, having numerous methodologies been developed for this purpose. This paper presents an enhanced modular Bayesian approach for structural identification, able to comprehensively quantify uncertainties in data interpreted by a physics-based model. Gaussian processes are used as emulators of the model response surface and existent modelling discrepancies. The Metropolis Hastings algorithm was used to expand the original methodology for multiple parameters inference. This method was applied to calibrate a finite element model of Tamar suspension bridge using long-term monitoring data. Measurements of temperature, traffic, mid-span displacement and natural frequencies in a one year span were used for identification of the prestress in the main/side cables of the bridge and stiffness of its thermal expansion gap. Results indicate a relative increase in the force of the suspension cables, while the expansion gap stiffness remains considerably low. This work is believed to pave a new way for identification of critical properties in large-scale infrastructures.