Bayesian model updating of structural parameters using temperature variation data.

Physical digital twins have now emerged as a technology in structural health monitoring and other domains of structural modeling. A Finite Element Model can be a tool for digital twin technology in conjunction with measured physical data. Developing an accurate digital model requires the integration of various parameters to replicate the real structure; however, temperature is dynamically changing, resulting in difficulty in the identification of structural parameters. Typically, temperature is considered noise or filtered out. Limited research has addressed the physics-based model using temperature variation related to structural modal properties. Finite Element Model Updating (FEMU) is a typical inverse problem-solving technique in structural modeling to match the measured data. In this study, temperature variations can be useful data to update the finite element model for detecting and quantifying structural damage. Numerical simulations are performed on a three-story aluminum shear frame model subjected to both uniform and non-uniform temperature conditions. The typical Bayesian model updating framework has not considered temperature as an important data related to dynamic measured data. A novel Bayesian model updating framework is proposed by accounting for temperature-induced variations in structural stiffness. In particular, this approach creates two conditions for updating both mass and stiffness parameters simultaneously. Under healthy and uniform temperature conditions, stiffness parameters were estimated with high accuracy, with errors below 0.5% and within uncertainty bounds. However, mass parameters exhibited small to moderate errors (up to 13.8%) that exceeded their extremely low standard deviations, suggesting potential model bias- systematic discrepancies between the model and reality. Under non-uniform temperature distributions, accuracy declined, particularly for localized damage cases, where both stiffness and mass parameters deviated significantly from their targets. Overall, this method advances physics-based digital twins and improves vibration-based damage detection under temperature variations.