High-performance and relatively inexpensive hardware components are capable of supporting advanced Structural Health Monitoring (SHM) paradigms, producing large datasets growing in size and complexity. Machine learning (ML) allows dealing with such large amounts of data and ensures learning from them, even when no prior knowledge of the investigated phenomena is available. However, purely data-driven approaches can mostly learn and predict behavioural patterns that were somehow experienced or simply provide evidence of deviations from baseline conditions. Thus, they require training on data that are representative across a broad range of possible environmental and operational conditions, are prone to overfitting and poorly generalise to out-of-sample scenarios. Furthermore, they are less interpretable than physics-based models, which, despite their limitations, allow for a more transparent understanding of the relationships between variables and the reasoning behind the applied inference. The integration of physics-based modelling, more intuitive for engineers, with ML techniques allows for the combination of the advantages of data-driven methods with the insights delivered by physics-based principles. This ensures an enhanced learning and predictive capability, particularly when training data are limited, by leveraging domain-specific knowledge. Various degrees of combination can be pursued within hybrid architectures, depending on the characteristics and objectives of the application. The scope of this mini-symposium is to present advancements and emerging trends in physics-informed SHM, discussing open challenges and promising solutions to tackle them. It aims to constitute a platform for the exchange of ideas and experiences and foster more coordinated and interdisciplinary research on the wide spectrum of modelling strategies to enforce physical laws or constraints within the learning process. We welcome contributions on the integration of physics-based with data-driven approaches focused on, but not limited to: • Advanced Bayesian approaches for the fusion of SHM and non-destructive evaluation data. • Empowering model-based SHM approaches with surrogate models. • Leveraging digital twins to enhance SHM strategies. • Network-level SHM using transfer learning techniques. • Cutting-edge methods for computation-physical domain adaptation in SHM. • Integrating physics and ML for enhanced SHM.