Ventilation configuration governs fire-induced thermal behaviour in enclosed building com- partments, yet the effects of non-aligned openings remain poorly understood. In this experimental study, sixteen full-scale fire tests were conducted in a 12.2 × 2.4 × 2.3 m compartment with three non-aligned openings. Fuel load (10–50 kg wood cribs) and ventilation configuration were varied to examine their influence on temperature evolution, flashover onset, and fully developed fire stage condition within the compartment, complemented by machine-learning and statistical an- alyses to enhance understanding of ventilation-driven fire behavior. Empirical results showed that the configuration and location of openings significantly affected thermal development in the compartment. Under side-opening conditions, the temperature rise preceding flashover onset occurred about 50s earlier as the fuel load increased from 30 to 40 kg, resulting in higher peak temperatures, whereas front openings yielded peaks that were 10–15 % lower and occurred about 60s later. The fully-developed stage under side-opening ventilation persisted longer, maintaining high-temperature conditions around 800–870 ◦C. Machine-learning models predicted peak tem- peratures with ~8 % error and classified the thermal transition linked to flashover with ~93 % accuracy, identifying fuel load as the key predictor. Statistical analysis confirmed strong corre- lations among thermal variables (r > 0.9) and an inverse relationship between fuel load and time to flashover (r ≈ 0.72). Experimental findings from this study, complemented by data-driven analyses, contribute to a better understanding of ventilation effects on compartment fire behavior and provide practical insights for ventilation-based fire-safety design, building venti- lation design, and thermal management in enclosed building environments.

Keywords