Forecasting the influence of the guided flame on the combustibility of timber species using artificial intelligence

Abstract

This paper anticipates the burning rate and optical obscuration characteristics of a 10 mm thick timber species often used in buildings under the influence of a guided flame condition with heat fluxes of 25 kW/m2 and 50 kW/m2. The smoke density chamber was used to test the wood species Pinus strobus, Pinus kesiya, Quercus alba, and Faqus sylvatica. The experimental data: time, specific gravity, mass loss, and heat flow were used as input variables to an artificial neural network (ANN) model. ANN with structure of 4-64-32-2 was built and validated, the results revealed that, the correctness of the established simulation was proven by a high value of R2 (0.99292 for validation) and highest validation performance (MSE = 17.809 at epoch 12). When the heat flow was reduced from 50 kW/m2 to 25 kW/m2, Quercus had the greatest drop in mass optical density (MOD). In the case of 25 kW/m2, the average charring rate was roughly 0.57 mm/min, compared to 0.96 mm/min in the case of 50 kW/m2. The MOD declines asymptotically for all species regardless of heat flux. The findings give statistical support and theoretical reference for fire-related construction norms and standards.

Keywords

• Artificial neural network
• Burning rate
• Heat flux
• Charring rate
• Mass optical density
• Smoke density chamber