This study investigates flame intensification caused by foam discharge in transformer oil-pool fires. Extinguishment experiments were performed using AFFF generated with CO2, N2, and air as foaming gases under different nozzle heights, pressures, and expansion ratios. Under identical conditions, CO2 foam achieved the shortest extinguishment time, approximately 7% and 39% shorter than those of N2 and air foams, respectively. Its dimensionless HRR increment was consistently lower and decreased with increasing expansion ratio. Under optimal conditions, the dimensionless HRR increment of CO2 foam reached only 0.20, indicating nearly complete suppression of flame intensification. A predictive model was also established for CO2 foam, showing that it reduces flame intensification by about 10% compared with N2 foam and 37% compared with air foam. Overall, CO2 foam achieved an average total suppression of 37.3%, including 27.2% from inerting/dilution effects and 10.1% from thermo-chemical effects. The heat-absorption effect, which lowers flame temperature, was identified as the main mechanism responsible for the enhanced suppression. These results demonstrate that CO2 foam, due to its distinct thermo-chemical properties, offers superior suppression of flame intensification, particularly at high expansion ratios.
The cabin environment of next generation inhabited spacecraft is characterized by sub-atmospheric pressures and elevated oxygen concentrations, creating unique combustion conditions for onboard solid materials. The influence of inert-gas dilution on flame spread and extinction is of critical importance for spacecraft fire safety. Experiments were conducted to investigate the opposed flame spread and extinction behaviors over a thick PMMA in low-velocity flow, focusing on the distinct influences of N2, Ar and CO2. The flamelets emerge under sub-atmospheric environment, and the critical flow-velocity decreases as oxygen concentration increases. The flame spread rate decreases with the reduced ambient pressure as predicted, but becomes independent of the gas species when the pressure is lower to a critical value. The decreased ambient pressure reduces the flame spread rate through thermal effect, although different dilutions influence through various effects. An effective Lewis number and Péclet number are proposed as the governing parameters for near-limit flames, since they represent the thermophysical properties of materials and environmental conditions, respectively. A normalized map is then established to compromise the effect of flow velocity, ambient pressure and gas dilution. These findings provide fundamental insights into flame spread and extinction under sub-atmospheric, diluted conditions relevant to microgravity spacecraft environment.
This study investigates how flame retardants influence the thermal decomposition pathways of wood fibre insulation, with particular emphasis on distinguishing condensed-phase reactions from gas-phase combustibility and their implications for fire behaviour. Two loose-fill wood fibre insulation materials were studied, with an ammonium salt-treated sample compared to an untreated reference. An integrated approach combining thermogravimetric analysis (TGA), differential scanning calorimetry (DSC), Fourier-transform infrared spectroscopy (FTIR), and microscale combustion calorimetry (MCC) was applied under inert and oxidative conditions at multiple heating rates. In addition, model-free kinetic analysis using the Friedman method was performed to determine activation energy as a function of conversion based on both TGA and MCC data.The flame-retardant-treated material (WL-h-10) exhibits a distinct decomposition pathway compared to the untreated reference (WL-h-Ref). TGA and MCC show an earlier onset of decomposition, shifted to lower temperatures by ∼100 K, together with a clear reduction in peak mass-loss rate by ∼43%, peak heat release rate by ∼75%, and effective heat of combustion of the evolved volatiles by ∼46%. In contrast, DSC indicates only minor changes in the total heat of reaction, distributed over a broader temperature range of 120–150 K. FTIR analysis reveals a shift toward dehydration and carbonisation, leading to reduced emission of combustible volatiles and increased char formation (+65%).These results indicate a predominantly condensed-phase flameretardant mechanism that reduces flammability by limiting the release of combustible volatiles. However, increased char yield and extended oxidation may favour smouldering, highlighting the need to distinguish between flaming and smouldering behaviour.
Residential fires pose a threat to public safety, yet limited research has examined the factors associated with their occurrence. This study analyzes household, dwelling, and neighborhood characteristics linked to residential fire incidents in the Netherlands. Using 2022 national fire incident data from the Integrated Dispatch Center System (GMS), linked with Statistics Netherlands microdata, over 7 million households were included, of which 5622 experienced a reported fire.Multivariable logistic regression showed that fires were more common in older dwellings and detached or semi-detached houses. Households with children were more likely to report a fire. An interaction between age and care dependency indicated that care dependency increased fire occurrence among younger households (<65 years), but not among older households (≥65 years). Lower-income households (1st–50th percentile) had higher odds of fire compared to the highest income group. Migration background showed only a weak association with fire risk, suggesting that behavioral or cultural factors may offer better explanatory value.The findings underscore the importance of targeted fire prevention strategies, particularly for vulnerable groups and high-risk dwelling types. Limitations include the absence of behavioral data and potential non-reported (minor) fires. Future research should incorporate behavioral and home renovation data to better understand fire risk mechanisms.
Fuel leakage during storage and transportation processes poses challenges to the fire safety of industrial systems. Especially when fuel continuously leaks from a vertical surface, the increased flow velocity caused by gravity leads to stronger convection. This paper presents an experimental investigation of the fire behavior of spill fires on a vertical plate. Two fuels (methanol and n-heptane) were used at different leakage rates. The experimental results indicate that methanol forms a stable fuel film structure in an arc shape in the quasi-stable state, while the liquid film structure of n-heptane is disrupted and breaks from the flame front to form many split flames. This phenomenon can be attributed to the reverse heating effect caused by the plate temperature exceeding the fuel temperature. A refined Sherwood number (Sh) is proposed to estimate the fuel burning rate from the perspective of evaporative mass transfer. Additionally, the heat transfer analysis during the quasi-stable stage indicates that the burning process is mainly dominated by the flame convection. Specifically, convection heat accounts for 98% of the total heat transfer in methanol and 95% in n-heptane. Furthermore, approximately 1.5% of the additional convective heat transfer in n-heptane is derived from the plate.
The small-world network (SWN) has been adapted for wildland fire spread prediction by incorporating heat transfer, ignition, and combustion dynamics. However, previous studies have not considered the effects of time step and cell resolution on simulation results. This work proposes an adaptive small-world network (ASWN) framework for fire spread modeling. The adaptive cell resolution method dynamically refines the local preheating cells around burning cells. This helps the model identify fuel ignition at the sub-cell scale. The adaptive time step method dynamically adjusts the time step. This aligns the transition time of preheating cells to the burning state with the actual physical ignition time. The ASWN framework is validated using laboratory-scale upslope fire spread experiments. The results show that the ASWN accurately reproduces the fire front shape. Compared to the SWN, the ASWN performs better in terms of simulation accuracy and efficiency. The average relative error of the ASWN for the upslope fire spread rate is less than 15%. Moreover, sensitivity analysis indicates that uncertainties in flame temperature and combustion time of the burning cell significantly affect the predicted rate of fire spread, especially under slow spread conditions.
This study investigates the bulk density of char formed from engineered wood products (EWPs) under different char formation conditions. Char residues from medium-density fibreboard (MDF), particleboard (PB), and oriented strand board (OSB) were produced in a cone calorimeter at various applied heat fluxes under oxidative conditions and compared with chars formed under inert conditions in nitrogen. A photogrammetry-based method was successfully applied to determine the bulk density.The results show that under inert nitrogen conditions, char bulk density generally decreases with increasing external heat flux, whereas under oxidative conditions this relationship is not preserved, indicating a significant influence of oxidation on the resulting char structure. Oxidation effects were also reflected in the elemental composition of the resulting char, although the observed trends differed from those observed for bulk density.The bulk density of char remains strongly correlated with that of the corresponding virgin material and typically corresponds to approximately 50 % of its initial value.These findings highlight the influence of both the virgin material structure and the char formation environment on char properties. The proposed bulk density values may improve the representation of char layers in pyrolysis models and contribute to more reliable predictions of temperature profiles during thermal decomposition.
This study experimentally investigates flame behavior and wall thermal exposure in vertical parallel-wall cavities under three ventilation configurations: side inflow, frontal bottom-slot inflow, and bilateral bottom inflow. Intermediate-scale experiments were conducted using a modified FM 4411–based apparatus with cavity widths of 50.8, 101.6, and 152.4 mm. Flame height, gas-temperature distribution, and total incident wall heat flux were measured using image analysis, thermocouple arrays, and water-cooled heat-flux gauges, respectively. A semi-empirical ventilation factor was introduced to account for frontal bottom-slot-induced air inflow, using the cavity-to-ambient temperature difference as an indicator of buoyancy-driven ventilation. The factor improved the collapse of flame-height data within the tested range. Frontal bottom-slot inflow induced near-base gas-temperature cooling and upward displacement of the high-temperature region, with this effect diminishing as cavity width increased. Total incident wall heat flux generally decreased with gauge height, but its magnitude and vertical distribution depended on ventilation pathway, cavity width, burner heat release rate, and frontal slot height. The measured heat flux exceeded estimates based only on direct flame radiation, indicating that wall thermal exposure cannot be attributed to flame radiation alone. Future studies should quantify radiative and convective components using wall surface temperature, near-wall velocity, and independent radiative heat-flux measurements.
Vision-based fire detection is crucial for minimizing catastrophic damage, yet traditional deep learning approaches are fundamentally limited by their reliance on large-scale labeled datasets. While Vision-Language Models (VLMs) offer promising solutions through rich cross-modal prior knowledge, adapting them to domain-specific fire scenarios under extreme data scarcity remains a significant challenge. In this paper, we propose a novel dual-pathway few-shot fire classification framework that synergistically integrates visual cache adaptation and test-time prompt tuning. First, we design a cross-attention cache adapter that constructs a category-partitioned feature repository from limited support samples, dynamically generating test-specific visual classifiers. Second, we introduce an unsupervised test-time prompt tuning strategy with a confidence-based selection mechanism, which optimizes fire-specific text prompts on-the-fly to ensure robust adaptation across diverse environmental conditions. Extensive experiments on four benchmark datasets (AIDER, DeepQuest-AI, FASDD-RS, and SmokeRS) demonstrate that our method establishes a new state-of-the-art in few-shot regimes. Specifically, our approach consistently outperforms existing advanced VLM fine-tuning paradigms, including state-of-the-art prompt tuning and feature adaptation methods. Furthermore, our framework surpasses traditional CNN and Vision Transformer architectures trained on complete datasets, while utilizing significantly fewer training samples. This study highlights the immense potential of our dual-pathway VLM adaptation for real-world, data-constrained fire monitoring applications.
To improve our understanding on the usage of tenability criteria in Performance Based Fire Safety Engineering, support future research and code development, an online survey was conducted, gathering 254 responses from 43 countries. Visibility in smoke is confirmed as both the most frequently measured (90.1%) and most influential (91.4%) tenability criterion. Temperature shows the largest measurement-versus-influence gap: tracked by 76.5% but governing decisions for only 45.8% of respondents. Toxicity-related criteria remain underused — FED is applied by 22.2% and FEC by fewer than 10% of practitioners, with FED adoption declining among the most experienced engineers (8.4% vs 19.4% among juniors). The C-factor in Jin's visibility model ranges from 3 to 8 in practice, with 30.5% of respondents choosing values on a project-by-project basis and 21% unable to specify their choice. Visibility thresholds span 3 m to 30 m across regions, with 10 m being dominant (from 97% in the UK to 64% in the Americas). Measurement heights range from 1.5 m to 2.5 m, with 2.0 m most common (52.1%) followed by 1.8 m (29.5%). Five-region analysis demonstrates that local regulatory traditions shape criterion selection, threshold values, and measurement methodology as strongly as independent engineering judgement.
This study investigated the impact of external winds blowing at downward and upward angles on smoke flow and critical ventilation velocity within tunnels. Two series of experiments were conducted by using a 1:20 scale model of a typical tunnel. The first series investigated the flow field without a fire under the influence of external wind, while the other explored the effect of external wind on the fire and smoke behaviors. Different external wind angles and velocities were applied. Our data show that the airflow inside the tunnel could be divided into turbulent and unidirectional flow zones when the external winds entered the tunnel. Additionally, the fire was tilt when it was located in the unidirectional flow zone, the mass loss rate decreased as the wind velocity increased. Conversely, the mass loss rate increased with the wind velocity when the fire source was in the turbulent flow zone. Furthermore, the length of smoke spread toward downstream was decreased with the addition of external wind, and smoke flowed back to the uptream. Therefore, a higher ucr to prevent smoke back-layering was needed when both the wind velocity and angle increased.
The paper focuses on large eddy simulations of reduced scale, well-ventilated compartment fire scenarios and aims to analyse and compare CFD results as obtained with the current modelling capabilities of FireFOAM and FDS. To this purpose, a qualitative and quantitative comparison of the results with the two CFD codes as a function of grid size is reported, for a range of fuel flow rates and ventilation flow rates. Both codes correctly predict flame tilting and a relatively uniform smoke layer. Yet, there are obvious differences between the predictions with the model settings as applied in this study. More specifically, FireFOAM predicts better the fire plume dynamics, in terms of maximum flame temperatures, with significantly less grid dependency. On the other hand, FDS predicts slightly better the smoke layer temperatures for all the grid sizes considered. An analysis of the flow field inside the enclosure is presented. The observed deviations between the codes are discussed and qualitatively analysed. Overall, there were no significant differences in the predicted air supply flow rates between the codes, which remained within 15% of the experimental measurements.
This study presents a comprehensive risk assessment of battery vent gas (BVG) from lithium-ion cells, focusing on its dynamic composition and laminar burning velocity (Su). It highlights lithium titanate (LTO) cells, which use Li4Ti5O12 anode, and compares them with NMC, NCA, and LFP chemistries.LTO cells improved safety by mitigating lithium plating and dendrite growth, key causes of internal short circuits and thermal runaway (TR). BVG composition evolves significantly during TR and depends on cell chemistry and state of charge (SoC), affecting combustion behaviour.BVG composition and Su were quantified during different stages of thermal abuse for cells with various chemistries and SoC (50% and 100%). A lab-scale setup with FT-IR and Micro-GC identified gases such as H2, CO, CO2, CH4, HF, and organic carbonates. Additional analyses (FT-IR) identified unburned species, while CHEMKIN-Premix calculated Su and sensitivity factors. Results show that LTO and LFP cells are safer under thermal abuse. At 100% SoC, NMC produced the most hazardous BVG, with higher hydrogen and the highest Su. At 50% SoC, TR was suppressed for LTO and LFP cells.These findings provide key insights into combustion dynamics and support improved mitigation of fire and explosion hazards in lithium-ion batteries.
Accidental explosions involving combustible dust-air or dust-gas-air mixtures remain a major safety concern in the food, pharmaceutical, and chemical industries. Standardized 20 L sphere and 1 m3 vessels are used to determine explosion parameters, including the maximum explosion pressure (Pmax), deflagration index (KSt), and minimum explosible concentration (MEC). However, these systems often exhibit non-uniform dust dispersion and poorly controlled pre-ignition turbulence, reducing the reproducibility of experimental results. This study investigates a fluidized-bed configuration as an alternative method for generating reproducible dust clouds under controlled flow conditions. Three representative combustible dusts, lycopodium, anthraquinone, and niacin, were analyzed using CFD simulations to characterize turbulent kinetic energy and solid volume fraction in the riser during bubbling fluidization. The results showed low, spatially homogeneous turbulence (u = 10−2 to 10−1 m/s), corresponding to quasi-laminar conditions favorable for reproducible explosion testing. Dust concentration was effectively controlled by fluidization time, which increased from lycopodium to anthraquinone and niacin because of differences in particle density. A theoretical evaluation for lycopodium indicated only a weak dependence of KSt on concentration, decreasing from 26 to 18 bar m/s. Overall, the proposed fluidized-bed system provides a controllable, low-turbulence environment for reliable pre-ignition characterization and improved repeatability in dust explosion testing.
Risk assessment of thermal runaway (TR) propagation in lithium-ion battery (LIB) systems requires quantitative characterization of heat generation, sensible energy carried by ejected gaseous and condensed materials, and their temporal evolution. In this study, a Double-Slug Battery Calorimeter (DSBC) is developed to quantify these values during safety venting (SV) and TR stages for 18650 cells. The DSBC consists of two copper slugs: one housing the LIB cell and inducing TR, and the other capturing ejected materials and their associated energy. The DSBC setup is used to examine two cell types with cathodes of LiCoO2 and LiFePO4, both with graphite anodes. A detailed methodology is provided for determining internal heat generation and its fractions. Results show that the total internal heat generation and energy carried by ejected materials can be quantified with controlled uncertainty using direct measurements of the initial and final temperature and mass of different components, together with calibrated heat-transfer and heat-loss terms. Furthermore, temporal heat release characteristics, such as heat generation during SV and TR, are estimated semi-quantitatively using additional assumptions, offering valuable insights into transient behavior and informing models for TR propagation in large-scale systems. The DSBC setup distinguishes different battery chemistries and provides key parameters for hazard assessment, establishing a foundation for developing standardized test methods for LIB safety evaluation.
In this study, a series of experimental tests was performed using an open controlled-atmosphere cone calorimeter (CACC) to evaluate the burning behavior and flame extinction characteristics of pure bitumen, a material widely used in applications such as pavement construction and in the encapsulation of nuclear wastes. Despite its broad usage, the fire behavior of bitumen remains poorly understood, particularly under oxygen-limited conditions. The work aimed to investigate the influence of oxygen-depleted environments on forced combustion behavior under varying external heat fluxes. Tests were conducted under both well-ventilated and nitrogen-diluted atmospheres at ambient pressure. The effects of different external heat fluxes and oxygen mass fractions on the thermal response of the condensed and gas phases were thoroughly analyzed. Key combustion parameters were determined along with other thermal properties of the bitumen samples. Temperature profiles within the condensed phase revealed a uniform temperature distribution across the sample in the decomposition process, consistent with a liquid state during the combustion process. Consistent repeatability was observed until the onset of overflow from the sample container and subsequent carbonization. At this stage, a sudden increase in viscosity was noted, accompanied by the formation of a solid char layer on top of the sample. This transition was associated with significant swelling, resulting in complex flame dynamics and a sharp peak in heat release rate. Analysis of the effect of oxygen concentration demonstrated a direct influence on the gas phase, leading to a proportional trend in the flaming tests.
