<![CDATA[This study aims to determine whether a data-driven-based approach provides more accurate predictions of timber fire-resistance ratings (FRR) compared to conventional empirical methods. To achieve this, a machine learning framework based on the Deep Belief Network (DBN) was employed. A comprehensive database collected from previously published reports was used to train and validate the DBN model. The model’s predictive performance was benchmarked against traditional empirical equations derived from mechanics-based methods. The comparison demonstrated that the DBN model provided superior accuracy in predicting fire-resistance ratings. Model evaluation was further conducted using the Coefficient of Determination (R²) and Root Mean Squared Error (RMSE), confirming the robustness of the proposed approach. In addition, a parametric analysis was performed to assess the influence of input variables on the output. Results indicated that induced load (IDL) and breadth (BRH) were the most influential parameters, whereas ultimate strength (ULS) and elasticity modulus (ELM) had relatively minor effects. This study highlights the potential of advanced machine learning techniques, particularly DBN, to enhance predictive accuracy in structural fire engineering, offering a significant improvement over conventional calculation methods.]]>
Copyrights © 2025
