Modeling compartment fire Heat Release Rate and temperature development: An alternative to the Eurocode parametric fire

This study presents a simple yet accurate data-driven model for generating complete compartment fire heat release rate (HRR)–time curves and, from these, predicting temperature–time histories through numerical simulation. The approach is based on 72 full-scale compartment fire tests without combustible linings. Using symbolic regression, a transparent machine learning technique, a concise equation was developed to estimate the maximum HRR (Q max ) from fuel load and ventilation characteristics. Q max , along with other key fire development points, was used to construct a three-phase growth–plateau–decay HRR curve, which was then applied in CFAST to obtain temperature–time profiles. Model validation against experimental data yielded an R 2 of 84.9 % with an RMSE of 0.99 MW for Q max prediction. Validation against 61 full-scale tests with temperature measurements showed that the proposed model predicted peak temperatures with an RMSE of 271 °C. In comparison, the Eurocode parametric fire (EcPF) model showed the same accuracy, but it applied to a narrower range of compartment conditions. All components of this study were implemented in a simple software tool for performance-based and risk-informed fire design.