Abstract: To address the discrepancy observed in the CBUF model when predicting the heat release rate (HRR) of large-scale mattresses, a method is proposed in which data from multiple thinner specimens tested using a cone calorimeter are superimposed. This approach approximates the more rapid vertical heat penetration observed in full-scale specimens. Cone calorimeter and furniture calorimeter tests were conducted based on the model’s input parameters. In the full-scale experiments, flame area was quantified using machine vision techniques, and a sensitivity analysis of the model was performed. The model results demonstrate that the predicted HRR values generated by this method show good agreement with the measured values, particularly improving accuracy during the decay phase of the HRR curve. This indicates that for thicker polyurethane (PU) foam materials, the vertical heat penetration rate in full-scale experiments is faster than that implied by the corresponding cone calorimeter data. Consequently, the HRRPUA curve obtained from full-scale experiments tends to resemble a “single-peak” profile rather than the “double-plateau” shape typically seen in cone calorimeter tests.

Key words: combustion of upholstered furniture, heat release rate, machine vision