Population-Based Structural Health Monitoring (PBSHM) is an emerging research field aimed at transferring knowledge from well-monitored structures to similar assets within a fleet. By leveraging data across entire populations, PBSHM holds the potential to change how we monitor, assess, and maintain critical infrastructure. However, advancing these methods relies on access to diverse datasets, a significant challenge because collecting comprehensive real-world data across structural populations is difficult, costly, and time-consuming.
To bridge this data gap, researchers are turning to realistic synthetic data generation. Synthetic datasets offer a scalable, controlled environment to simulate complex structural behaviours, environmental variations, and damage scenarios that are otherwise rare or dangerous to capture in the field.
In this talk, we will explore the methodologies behind generating realistic synthetic data tailored for PBSHM. We will discuss how physics-informed and data-driven generative models can produce high-fidelity datasets, the key challenges in ensuring synthetic accuracy, and how these data foundations are being used in the development, validation of PBSHM methods