Predicting the Heat Release Rate of Rack Storage Polyurethane Foam Blocks Using AI-Assisted Symbolic Regression Analysis

This study aimed to quantify the effect of polyurethane foam block (PUFB) arrangement on the heat release rate (HRR) under rack storage conditions and to examine the applicability of AI-assisted symbolic regression for estimating HRR amplification. HRR was measured for 17 configurations of cubic PUFBs, including single, horizontal, and vertical arrangements. From the measured HRR data, the fire growth rate and averaged maximum HRR were determined and normalized using the corresponding single PUFB reference values to derive the amplification factors Fa and FQ. Subsequently, AI-assisted symbolic regression was applied using the arrangement indicator variables as inputs, and the resulting candidate correlations were evaluated based on prediction error, cross-validation, and equation simplicity. As a result, a model incorporating the interaction between the number of horizontally arranged PUFBs and their separation distance was selected, and the mean absolute percentage errors (MAPEs) for Fa and FQ were 31.4% and 17.2%, respectively. The proposed correlation should be interpreted as a preliminary model for estimating HRR amplification under the limited laboratory-scale PUFB arrangements considered in this study, and further validation using real commodities and full-scale rack fire experiments is required for practical applications.