Physics-Informed Machine Learning for Photopolymerisation: Governing Equations, Data Assimilation and Extrapolation Limits

Physics-informed machine learning can reduce physically implausible predictions and improve data efficiency by combining measurements with governing equations or physics-derived features. Photopolymerisation, however, couples optical attenuation, reaction kinetics, inhibition, species transport, exothermic heat generation, rheology and evolving geometry. No single simplified equation is universally valid across all resin chemistries, filler states and printing regimes. This review therefore treats physics as a set of testable constraints rather than as unquestionable truth. It defines levels of physics integration, identifies state equations that may be useful, and sets validation requirements for extrapolation beyond the calibration domain. The framework distinguishes a model that is numerically consistent with an assumed equation from one that is experimentally validated for a given photopolymer process. ER-405.