Stochastic Digital Twins of Bridges for Computing Condition Indicators under Model Form Uncertainty

• Applicants: Dr.-Ing. Jörg F. Unger; Dr. Martin Weiser
• Subject Area: Applied Mechanics, Statics and Dynamics ; Structural Engineering, Building Informatics and Construction Operation
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 562812195

The goal of the proposal is the development of methods to iteratively establish a reliable simulation based digital twin for predicting condition indicators using both continuous monitoring data and laboratory data. The main focus is on quantifying the accuracy of simulation-based predictions to support a reliable decision-making framework. This framework will address not only aleatoric uncertainties but will specifically account for epistemic uncertainties, which are represented by model form uncertainties inherent in any simulation model derived from reality through simplifying modeling assumptions. The basis are parameterized FE-models, with a focus on thermo-mechanical models with parameterizations that depend on space, time and environmental conditions. Due to the computational expense of evaluating these models, we will also explore the development of efficient surrogate models. Subsequently, we will address model calibration and uncertainty quantification in the presence of model form uncertainty, utilizing the embedded bias formulation established during the first funding period and extend this to multiple dimensions with a specific focus on representing correlations and propagating stochastic variables through the forward models using sparse Polynomial Chaos Expansion. Additionally, this will involve explicitly identifying parameter dependencies on spatial, temporal, and environmental conditions to suggest potential model improvements. Next, we will propagate these uncertainties to the computation of Condition Indicators and Prognostics (CIPs) and develop metrics to assess model quality. This will involve defining representative CIPs based on a decomposition of the total deformation field into thermal, traffic induced and permanent mechanical deformations. We will investigate the sensitivity of the CIPs on the inability of the model to represent the data using residual based model updating techniques. Based on this, methods and metrics are developed to automatically detect anomalies. Finally, these methodologies will be implemented in a software demonstrator. Our goal is to develop a comprehensive software package and a query-based server component for two demonstrator projects (Nibelungenbrücke in Worms and IDA-KI in Bautzen) that can be utilized by partner projects within the SPP with the aim to compensate the influence of environmental conditions on the mechanical response and subsequently identify permanent changes in the structural behavior.

• DFG Programme: Priority Programmes
• Subproject of: SPP 2388: Hundred plus - Extending the Lifetime of Complex Engineering Structures through Intelligent Digitalization
• Co-Investigator: Professor Phaedon-Stelios Koutsourelakis, Ph.D.