Physics-aware machine learning for predicting the durability of textile-reinforced mortars under environmental and time-dependent mechanical loading

This project aims to develop a physics-aware machine learning framework for predicting the longterm
durability of textile-reinforced mortars (TRMs) under combined environmental and mechanical
loads. TRMs are increasingly used for strengthening and preserving masonry structures, yet their
degradation under realistic service conditions, such as acid attack, salt exposure, creep, and fatigue,
remains poorly understood and difficult to predict. By combining legacy datasets (from over 3,000
specimens), multi-sensor experimental data (AE, DIC, UPV, hygrothermal probes), and
microstructural analysis (SEM/EDS, XRD, FTIR, TGA), the project will build interpretable,
generalizable models that link degradation mechanisms at material scale to performance loss at
structural level. A core novelty lies in embedding physical constraints into the machine learning
process using Physics-Informed Neural Networks (PINNs), enabling the models to remain realistic
and transparent. The resulting predictive tools will support risk assessment, service-life design, and
maintenance planning for sustainable strengthening applications, and contribute to European
standardization and open data initiatives.