Boilover is a destructive phenomenon that occurs during a fire in hydrocarbon tanks with a bottom layer of water; it is characterised by an explosive eruption of burning fuel and a sudden increase in thermal radiation. Despite research conducted since the 1950s, our understanding of this phenomenon remains fragmentary, with a limited body of scientific literature (118 publications indexed in Scopus between 1959 and 2025) and a lack of validation at an industrial scale. To address these limitations, this study combines bibliometric analysis with a systematic literature review to analyse the evolution of boilover research and synthesise current knowledge of the phenomenon's mechanisms, experimental investigations, analytical and numerical modelling approaches, and the main factors influencing its development. The bibliometric analysis examines the dynamics of scientific output, collaboration networks and the thematic structure of the field, whilst the systematic review identifies the main scientific findings, the limitations of current knowledge and the key research gaps. The results highlight a high concentration of scientific impact centred on pioneering work, a division of the field into several distinct areas (combustion, process safety and risk management), and a persistent fragmentation of collaborations. The review further shows that, despite significant progress in understanding boilover mechanisms and developing analytical and numerical models, large-scale experimental validation remains limited, predictive capabilities remain insufficient, and several scientific challenges persist, particularly concerning emulsified crude oils, wall effects and scaling phenomena. This review therefore highlights the main scientific gaps and proposes several research priorities to improve understanding, prediction, and prevention of the boilover phenomenon. The review further shows that, despite significant progress in understanding boilover mechanisms and in developing analytical and numerical models, large-scale experimental validation remains limited, predictive capabilities remain insufficient, and several scientific challenges persist, particularly regarding emulsified crude oils, wall effects, and scaling phenomena. These findings highlight the main scientific gaps and establish research priorities to improve the understanding, prediction, and prevention of the boilover phenomenon.
Predicting the strength of geopolymer concrete (GPC) after fire incidents is crucial for ensuring structural safety. From this lens, the present study investigates the prediction capabilities of machine learning (ML) algorithms to examine the residual strength of GPC after being subjected to the ISO 834 fire curve. More specifically, a comprehensive dataset from fire experiments comprising of ground granulated blast furnace slag (GGBS), fly ash (FA), sodium hydroxide (SH), sodium silicate (SS), alkaline activator to binder (AAS/B) ratio, molarity (M), extra water (EW), curing temperature (T), exposure to elevated temperature (ET) and exposure time (t) was collected and analyzed. This dataset is utilized to train and test six ML models, namely, Extreme Gradient Boosting (XGBoost), Random Forest (RF), Light Gradient Boosting Machine (LightGBM), Categorial Boosting (CatBoost), Adaptive Boosting (AdaBoost), and Gradient Boosting (GB). The models were trained via a cross-validation process and then examined using statistical indicators. In comparison with all the ML models, CatBoost shows the highest accuracy with an R2 of 0.997 and the lowest RMSE value. To further enhance the developed models, their interpretability was examined using SHAP, which shows that t, T, and ET were the most influential factors that govern the strength of the GPC mix. This finding highlights the potential of ML in developing light and cost-efficient prediction frameworks for post-fire assessment of binary blended GPC to help augment the specialized and expensive nature of fire testing.
Building façade fires resulting from the ignition of combustible cladding materials have been reported extensively. However, research specifically addressing the risk posed by external fire sources, such as vehicles parked near cladding, is still limited. This study applies probabilistic analysis to assess fire risks associated with car fires in proximity to combustible cladding, considering Aluminium Composite Panels (ACP) containing a polymer-based core and Expanded Polystyrene (EPS). The influence of wind conditions is incorporated into the analysis. The objective is to create a database supporting decision-making on rectification measures, reducing façade fire risks, and ensuring occupant safety. This study is developed within Cladding Safety Victoria’s framework for mitigating fire risks.
Burning velocity is one of the fundamental flammability parameters, and the maximum burning velocity is a crucial criterion of flammability classification of refrigerants. In this study, an experimental system for measuring the burning velocity of low-carbon flammable gases was established based on ISO standard 817. The combustion reactions of mixture refrigerants consisting of R1243zf, RE170 and flame retardants (R13I1, R134a, or R125) were experimentally investigated. In addition, a combustion reaction kinetic model applicable to R1243zf and its mixtures was established, consisting of 178 species and 1442 elementary reactions. The modified kinetic mechanism reproduced the ISO 817 vertical-tube-derived burning-velocity dataset for R1243zf and its mixtures with a mean deviation of 14.5
This article studied the temperature evolution and flame ejection of enclosure fires with the upper and lower openings. Experiments were conducted inside a reduced-scale enclosure with the side length of 0.4 m, which has two openings of the same size, vertically distributed and separated at a distance. Temperature inside the enclosure and flame behavior were recorded for different window sizes, heat release rates and opening separation distances. Result shows that: (1) Temperature inside enclosure rises as total HRR increases, and falls as the opening separation distance increases. This is because air flow rate entering the enclosure increases with the increase of opening separation distance. (2) Critical HRR of ejection rises as the window size increases, and decreases as the increase of opening separation distance. This is because that when the upper opening position is higher, flames are more likely to eject out from the opening after reaching the enclosure top. (3) A normalized separation distance was proposed, which well characterized the effect of separation distance on inside temperature and flame ejection, and the temperature and critical flame ejection were correlated well by the normalized distance. This study for the first time revealed the temperature evolution and flame ejection inside a double-opening enclosure, analyzed the effect of opening separation distance, and proposed the models to quantify the inside temperature and critical HRR for flame ejection, which would make a significant contribution in the theory of building fire dynamics.
