Answers are grounded in this paper and cite their sources.
Intumescent flame-retardant coatings (IFRCs) can modify the thermal degradation and mass-loss behavior of wood. This study experimentally evaluated the mass-loss characteristics of Fagus sylvatica and Pinus rigida under coated and uncoated conditions, using data from four independent cone calorimeter experiments conducted at 50 kW m −2 in accordance with ISO 5660-1:2015. The experiments generated 845 observations, and we assessed key mass-loss characteristics, including total mass loss, t 50 , peak instantaneous mass-loss rate, and time to burnout. For these individual experiments, the data indicate that the coating reduced total mass loss in both wood species. Total mass loss declined from 86.87% to 80.93% for F. sylvatica and from 89.07% to 81.92% for P. rigida . For F. sylvatica , the coating also delayed the onset of rapid mass loss from 25 to 90 s and reduced the fraction of rapid mass loss from 36.6% to 14.9%. Four machine-learning models, CatBoost, NGBoost, Random Forest, and HistGB, were subsequently evaluated for predicting specimen mass using default and Bayesian-optimized hyperparameters. NGBoost showed favorable relative performance under 5-fold pointwise cross-validation, reflecting within-trajectory predictive performance, yielding an MAE of 0.167 g, MSE of 0.074 g 2 , and R 2 of 0.999769. Under a leave-one-run-out validation framework, the HistGB showed notable run-level generalization within the evaluated set, achieving an overall R 2 of 0.8941 and MAE of 4.5554 g. These results suggest that the experimental–ML framework can yield consistent predictions within the specific experimental domain
No takes yet. Share an insight, caveat, or question.
Olimat et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: