Study on the mechanism and evolution of cross-scale damage in hot-rolled work rolls based on an accelerated proxy model

Predicting the long-term service life of hot-rolled work rolls is of great significance to the steel industry, but it is constrained by their complex cross-scale damage mechanisms. Traditional macro-scale models cannot reveal their physical essence, while high-fidelity multi-scale simulations are computationally too costly. Purely data-driven models, on the other hand, lack physical interpretability and require large amounts of experimental data. This study addresses this challenge by developing a physics-data fusion-driven multi-scale analysis framework that combines the physical rigour of first-principles models with the computational efficiency of data-driven approaches. This framework integrates a macro-scale thermo-mechanical-coupled finite-element model with a micro-scale ‘oxidation-damage’-coupled phase-field-model, elucidating the key mechanism by which oxidation-assisted thermal fatigue cracks nucleate at the oxide/matrix interface. Additionally, a Gaussian Process Regression (GPR) surrogate model is introduced, achieving high predictive accuracy (R 2 > 0.99) and over two orders of magnitude in efficiency improvement, enabling efficient surrogation of computationally intensive micro-damage simulations. This framework enables precise and efficient prediction of damage evolution throughout the entire lifecycle of the work roll, providing both a theoretical basis and practical tools for optimising rolling processes.