Prediction of the optical properties and optimization of the manufacturing conditions of glass-ceramic products using machine learning

Glass-ceramics have excellent aesthetic properties, and their optical properties depend on the crystallization temperature. Traditional manufacturing has relied on experience and trial and error. The objective of this study is to model the relationship between the manufacturing conditions and optical properties (CIE color coordinates L* , a* and b* ) using machine learning and propose the optimal manufacturing conditions for achieving desired properties. The models were evaluated by double cross-validation and interpreted by the cross-validated permutation feature importance, which accounts for the correlations among the variables. TP1, TP2, and TP3—variables related to the crystallization temperature—were found to significantly affect the optical properties. Inverse analysis was performed to adjust these variables and identify the optimal manufacturing conditions to achieve the target optical properties. Glass-ceramics fabricated under these conditions showed optical properties closely matching the predicted values, confirming the model’s effectiveness and accuracy. Production-scale validation under the identified conditions showed that 10 of the 11 samples met the optical property standards. These results support the practical usefulness of the proposed ML-based workflow for designing manufacturing conditions.