A residual-based Gaussian process model framework for finite element model updating

• The residual rather than structural response is used as the output of GPM.
• The analytical variance-based GSA is implemented within residual-based GPM framework.
• The analytical first and second derivatives are utilized to accelerate the optimization process.
• A real arch bridge is provided to verify the feasibility of the proposed method.

A residual-based Gaussian process model (GPM) framework is proposed for finite element model updating (FEMU). The core idea of the proposed method is that GPM is adopted to characterize the relationship between the residual and the selected parameters. Within the residual-based GPM framework, the powerful variance-based global sensitivity analysis can be analytically implemented for parameter selection, and the rate of convergence of the optimization process is accelerated substantially by providing the analytical gradient and Hessian information. A real-world arch bridge is presented to illustrate the proposed residual-based GPM framework and verify its feasibility in FEMU.