Model updating [1] has been developed as a typical topic to calibrate the parameters or the model itself to turn the model prediction towards the experimental measurements. However, it is widely recognised that the unavoidable uncertainties in both experiments and simulations must be understood in model updating. Non-deterministic modelling approaches enable characterisation, propagation, and quantification of uncertainties, providing predictions over a possible range of outcomes rather than a unique solution with maximum fidelity to a single experimental observation.

Structural Health Monitoring (SHM) plays a significant role in providing insight into the structural properties in the life-circle of produces. Model updating has a natural connection with the topic System Identification, implying the inherent properties of the physical system can be identified (or predicted) from the numerical modelling. This brings SHM and model updating together in this project: to develop a reliable numerical model with a precise indicator to support decision-making in SHM.

For uncertainty treatment, a significant aspect is the advanced data technology to extract as much as possible uncertainty information from the available experimental data, to enhance the data-driven decision-making process. From the applicant’s recently organised special issue on advances of model updating , it is surprising to find that Artificial Intelligence (AI) techniques are absent.

This has been caused by two challenges:

1. how to avoid non-unique solutions when the AI-based agent model is employed in model updating
2. how to implement AI in the presence of gross data with multi-source of uncertainties

This project is consequently aimed at filling the gap between the popular AI techniques and their application to SHM and model updating with novel approaches of uncertainty treatment.

[1] Overview of Stochastic Model Updating in Aerospace Application Under Uncertainty Treatment, pages. 115–129.