Machine-Learning-Assisted Photopolymer Formulation: Inverse Design, Active Learning and Multi-Objective Optimisation
Segurola, Juan
2026-08-21 · Report · Version 1.2
Photopolymer formulation is a constrained multi-objective problem in which liquid rheology, cure kinetics, optical attenuation, conversion, dimensional fidelity and final mechanical or functional properties interact. Machine learning can reduce experimental burden, but small datasets, correlated formulation variables, measurement noise and process-dependent labels create substantial risk of overfitting and false optimisation. Recent work has directly demonstrated ML-guided DLP-elastomer formulation and broader materials studies have established uncertainty-aware active learning and Bayesian optimisation as practical strategies for navigating expensive experimental spaces. This review defines an engineering architecture for ML-assisted photoresin development: encode chemically meaningful variables; standardise process and test metadata; use uncertainty to guide experiments; treat feasibility constraints explicitly; optimise Pareto trade-offs rather than a single score; and close the loop with printed validation. 3Dresyns customisation services are included only as implementation context, without disclosure or inference of internal algorithms.