Towards a Hybrid Digital Twin: Fusing Sensor Information and Physics in Surrogate Modeling of a Reinforced Concrete Beam

In this study, we investigate the potential of fast-to-evaluate surrogate modeling techniques that fuse the sensor data with non-sensor information, i.e. underlying physics, for developing a hybrid digital twin of a steel-reinforced concrete beam, serving as a representative example of a civil engineering structure such as a bridge. Bridges are critical infrastructures that require continuous monitoring and maintenance with predictive power to ensure their safety and longevity. Therefore, there is a high demand for surrogate models that combine sensor data with physics to construct explainable predictive surrogates. As surrogates, two distinct models are developed utilizing physics-informed neural networks (PINNs), which integrate sensor data with non-sensor context knowledge, i.e. given governing laws of physics by spatio-temporal data integration. The sensor data is obtained from a previously conducted four-point bending test. The first surrogate model focuses on temporal phenomena and predicts strains at fixed locations along the center line of the beam for various time instances. Here, we compare the physics-based approach with a purely data-driven method, revealing the significance of physical laws for the extrapolation capabilities of models in scenarios with limited access to experimental data. Furthermore, we identify the natural frequency of the system by utilizing the physics-based model as an inverse solver. For the second surrogate model, we then focus on spatial phenomena at a fixed instance in time and combine the sensor data with the equations of linear elasticity to predict the strain distribution within the beam. This example shows how the integration of data can improve the insufficiently accurate predictions of a simplified physical model, given suitable loss weights.