Virtual sensors incorporate physics-based simulations with data-driven inference, aiming towards providing predictive sensor data for quantities that are difficult to measure using conventional methods. This approach can provide predictive capabilities for condition monitoring and trend identification of phenomena embedded within the structure [1]. This work explores virtual sensing in steel-reinforced concrete structures using physical knowledge and in-situ data. To apply the physical knowledge, a mixed-dimensional finite element model of the structure is developed based on a so-called beam-to-solid volume coupling approach [2]; The model is created with 1D geometrically exact beam finite elements representing the steel reinforcements within a 3D solid representation of concrete. This approach allows for capturing the intricate interactions between the reinforcement and the concrete compound. The results of the finite element model are combined with the data provided from the structure via sensors and measurements to create a dataset. The dataset is used to train a Gaussian process, which learns the relation between the provided information and the values of interest. The trained Gaussian process can provide probabilistic predictions of values normally provided by physical sensors, inferred from already available information. To summarize, the virtual sensor approach can prove to be an essential tool to portrait the behaviour of reinforced concrete structures. This method can then predict key values in existing structures that have no embedded sensors available, or infer data not measured normally, hence assisting in protecting critical infrastructure, along with the structural health monitoring (SHM) concept. References: [1] K. Maes, G. Lombaert, Validation of virtual sensing for the reconstruction of stresses in a railway bridge using field data of the KW51 bridge. Mechanical Systems and Signal Processing, Volume 190, 2023, 110142, ISSN 0888-3270, [2] I. Steinbrecher, M. Mayr, M. J. Grill, J. Kremheller, C. Meier, A. Popp, A mortar-type finite element approach for embedding 1D beams into 3D solid volumes. Computational Mechanics, 66:1377-1398, 2020.