ABSTRACT. This paper describes the development and validation of a computational methodology to predict the behavior of steel gravity frames with composite floor systems subjected to fire. Three full-scale compartment fire tests (CF1, CF2, and CF3 Tests) conducted on a two-story steel gravity frame structure with a composite concrete floor slab on steel decking were used in the comprehensive validation process. The experiments were conducted under combined mechanical and fire loading in the National Fire Research Laboratory (NRFL) at the National Institute of Standards and Technology (NIST) [1, 2, 3]. The objective of the computational study was to validate the adequacy of the computational modeling approach to capture the primary structural behaviors observed in the fire tests conducted in the NIST NFRL and to provide additional insights into the structural performance through detailed modeling of the system response. The computational model of the composite floor system included detailed modeling of the system components, including the lightweight reinforced concrete floor slab on profiled steel decking, steel beams, steel girders, supporting columns, shear stud connectors, and bolted steel shear connections. Additionally, temperatures measured during the composite floor fire tests were specified as nodal temperature histories for all components of the composite floor system to model thermal loading resulting from the fire exposure in the experiments.

ABSTRACT. Timber claddings offer a potentially sustainable alternative to conventional fire protection systems for steel columns in hybrid steel–timber structures. However, the available design rules for such systems are limited, and the influence of detailing on the fire performance is not well quantified. This contribution presents an experimental and numerical study on the fire behavior of timber-cladded steel columns and derives corresponding design recommendations and simple design models for practical use. A total of 42 column specimens representing typical rolled sections (HEM, HEA, HEB, IPE) with profile factors Ap/V between 45 m⁻¹ and 301 m⁻¹ were tested in a furnace under ISO 834 standard fire exposure. The columns were thermally but not mechanically loaded. The main parameters were the cladding system (single-layer box cladding with edge screws, double-layer cladding with staggered joints, or single-layer cladding on timber battens), the timber plate type (three-layer panels), cladding thickness, and profile factor. The target fire resistance classes were R30, R60, and R90, with a critical steel temperature of 500 °C as the failure criterion. The tests were supplemented by a calibrated 3D finite-element heat-transfer model, which included a two-stage cladding failure model and an extended parametric study. The results confirm that timber claddings can provide effective fire protection to steel columns; however, their performance is highly sensitive to the detailing of joints and fixings. Simple box claddings with screws only in the narrow faces exhibited large deformations and opening of corner joints, leading to internal hot-gas flow, multi-sided exposure of the cladding, and premature temperature increase in the steel. Double-layer claddings with staggered joints reduced gas ingress but introduced new failure modes: thin lamellae of the three-layer plates locally delaminated, and failure of the outer layer led to accelerated charring and early loss of the inner layer. In contrast, single-layer claddings fixed to timber battens by cross-screws showed robust behavior; all specimens with this configuration achieved the intended fire resistance class. For these systems, minimum cladding thicknesses of 22 mm (R30), 50 mm (R60), and 60 mm (R90) were sufficient for columns with Ap/V ≤ 320 m⁻¹. Double-layer systems require 16+16 mm, 27+22 mm, and 32+32 mm, respectively, over the same Ap/V range. A beneficial influence of low-profile factors (more massive sections) on the required timber thickness was observed whenever failure was not governed by joint opening. The numerical model reproduced the measured steel temperatures conservatively up to cladding failure and was used to derive both tabular design rules and an analytical design equation relating steel temperature to cladding thickness, profile factor, and fire duration. Together with recommended detailing for battens, joints, and fasteners, the results demonstrate that timber-cladded steel columns can be reliably designed up to at least R90, provided that the cladding system is detailed to prevent early joint opening and cladding loss. Overall, the investigations demonstrate that timber claddings can provide a robust and resource-efficient fire protection system for steel columns, if detailing is carefully controlled. Among the tested variants, single-layer claddings mounted on timber battens with cross-screws exhibited the most reliable behavior and consistently achieved the targeted fire resistance classes up to R90. In contrast, simple box claddings and unfavorably laminated double-layer systems were prone to early failure due to joint opening and delamination. The derived minimum cladding thicknesses and the proposed simplified design model offer a practical basis for engineering design and verification of timber-cladded steel columns over a wide range of profile factors. At the same time, the results underline the need for clear detailing rules on joint geometry, edge support, and fixing patterns, and provide a benchmark for future studies on mechanically loaded columns and performance under natural fire exposures.

ABSTRACT. The conference paper will provide an insight to the experimental and numerical results and compare the structural behaviour of 11 steel-reinforced concrete columns tested conventionally and with substructure method. Fire tests under different initial test loads and restraining conditions have been carried out on normal-strength steel-reinforced concrete columns with the same geometry. In addition, a numerical model is used to improve the understanding of the influence of the surrounding structure and the behaviour of the columns in the case of fire.

ABSTRACT. The growing use of mass timber in construction increases the need for improved understanding of its fire comportment under realistic building fire conditions. Current prescriptive design approaches rely on standardized time–temperature curves, such as CAN/ULC-S101, which do not accurately reflect the dynamic of real compartment fires. In practice, incident heat flux, oxygen concentration, and fire dynamics vary significantly during the heating and cooling phases, influencing both the combustion process and the charring behavior of mass timber elements. As charring rate is a key parameter for assessing structural fire resistance, better characterization under realistic fire exposures is essential for the advancement of performance-based fire design. This research investigates the charring behavior of cross-laminated timber (CLT) when subjected to a range of realistic fire conditions. Two experimental approaches were employed. First, small-scale tests using a cone calorimeter and a controlled-atmosphere cone calorimeter were conducted at heat fluxes between 20 and 80 kW/m² and oxygen concentrations between 5 and 20.95%, allowing independent assessment of these parameters. Second, intermediate-scale furnace tests were performed on instrumented 5-ply CLT panels exposed to time–temperature curves representative of real compartment fires, including the cooling phase, and compared with CAN/ULC-S101 reference exposures. The study aims to generate a comprehensive dataset describing how heat flux and oxygen concentration influence charring rate, and to propose an empirical equation for charring predictions under realistic fire scenarios. These findings will support the development of performance-based alternatives to current prescriptive fire design methods and contribute to the use of exposed timber in modern construction.

ABSTRACT. Glued-in Rod (GIR) connections are increasingly used in mid- and high-rise timber structures due to their high stiffness and moment-resisting capacity; however, the direct exposure of the end plate makes them thermally vulnerable under fire. To address the lack of experimental data on the thermal behavior of GIR connections, this study investigates their component-level temperature response under standard fire conditions. Unloaded fire tests were conducted using two variables: an unprotected end plate and a 38-mm fire-protective board. The specimens were heated for 180 minutes following ISO 834-1 (1999), and thermocouples were installed on the end plate, steel rod, threaded rod, and at multiple depths within the glulam column and beam. Test results showed that, in the unprotected specimen, the end plate, steel rod, and threaded rod reached critical temperatures at approximately 45, 120, and 90 minutes, respectively. In the protected specimen, most components remained below critical steel temperatures, though localized temperature spikes occurred when rods became exposed due to charring. A fundamental FEM heat-transfer model was developed from these results for future parametric fire-resistance studies.

ABSTRACT. As one of the sectors with the greatest environmental impact, the construction industry is shifting toward circular practices that promote material reuse and recycling. In this context, the use of demountable structural systems in composite construction enables component recovery and extends the building service life. Such practices also include the use of more sustainable materials in construction, such as engineered timber products. In the framework of a wider research project aiming at characterizing the mechanical performance of demountable steel-timber flooring systems, this paper presents the results of two real scale fire tests on an innovative type of steel-timber composite (STC) beams with demountable shear connectors, incorporating a sustainable passive fire protection system which is proposed as an alternative to other conventional fire protection solutions. The fire tests showed an excellent performance of the demountable STC beams at elevated temperatures, owing to the devised fire protection system. The tested beams outperformed the initially established fire resistance goal of 90 minutes and were capable to eventually withstand 120 minutes standard fire exposure under a demanding load level. The results of this experimental study confirm the suitability of the studied demountable steel-timber flooring systems to be safely used in multi-storey buildings, while also highlighting the compatibility with the circular economy principles and their potential to reduce the carbon footprint in the construction sector.

ABSTRACT. Depending on the engineered wood product, the lamination layup can vary in both grade and orientation, which can affect the char layer formation. This is particularly critical for cross-laminated timber (CLT) in the major strength direction where the exterior longitudinal layer provide a significant contribution to flexural resistance and stress distribution. Therefore, if the outermost layer is lost to char, effectively two layers are lost in terms of flexural resistance. Even once flaming combustion has extinguished, residual heat continues to propagate throughout the cross-section and impose a greater risk of damage to the wood. This first stage study evaluates the impact of localized fire damage on the flexural capacity of CLT in one-way flexure. A total of eight CLT specimens measuring 2,950 mm long, 600 mm wide, and 139 mm thick were partially encapsulated in mineral wool rigid insulation to restrict fire damage to a localized area. Fire exposure consisted of a 30-minute methanol pool fire. Four CLT specimens were immediately cooled with water once the methanol fire had extinguished, while the remaining four specimen were cooled with water one hour after the methanol fire had extinguished to examine post-fire thermal propagation. Preliminary results on three specimens indicate that while the peak load and stiffness of the CLT are reduced when fire damaged, the reduction is significantly greater when the shear region (87% reduction) is damaged compared to an undamaged or moment region damaged specimens (60% reduction). Further investigation is required to confirm these trends.

ABSTRACT. Probabilistic models for the temperature-dependent thermal properties of timber can support structural fire analysis in performance-based structural fire engineering, where the variation in material properties has a large effect on the reliability of heat transfer calculations and structural response assessments. However, current design standards rely on deterministic relationships that do not model the inherent uncertainties and variabilities across timber species, which can be quite significant given timber is an orthotropic material. This study compiles experimental data for density, thermal conductivity, and specific heat from the literature at temperatures from 20°C to 435°C, together with existing deterministic relationships from prior studies. An inverse-variance weighted model ensemble of these existing deterministic relationships was used to construct an informed prior, which was then updated using the experimental dataset through an empirical Bayes procedure with temperature-dependent prior weighting. Based on information criterion model selection, the lognormal distribution was adopted for density ratio and specific heat, and the log-logistic distribution for thermal conductivity. This study reports temperature-dependent location and scale parameters in closed-form expressions that are suitable for implementation in reliability analyses and Monte Carlo simulations. For the effective specific heat in the evaporation band, a rectangular spike formulation is proposed that considers ambient moisture content as a lognormal random variable, and thus allows users to propagate moisture content uncertainty analytically into the c_(p,eff) distribution.

ABSTRACT. Urban-scale fires in complex-built environments have emerged as a critical safety issue due to climate-induced extreme weather and dense spatial development. The 2025 Eaton Fire in Los Angeles County burned >50 000 acres and destroyed >16 000 buildings, revealing the need for rapid simulation tools capable of predicting fire-structure interactions at the city scale. Existing models either simplify key mechanisms or lack structural fidelity, leading to limited applicability in wildland-urban interface (WUI) regions. This study develops an integrated Cellular Automata (CA)–Unity framework for three-dimensional simulation of urban fire spread and structural exposure, incorporating four mechanistic pathways: intra-building spread, thermal radiation, direct flame contact, and firebrand spotting. The platform bridges GIS-based data processing and structural fire engineering, offering an interactive and computationally efficient tool for resilience-oriented urban fire analysis.

ABSTRACT. This study presents an integrated performance-based design methodology for assessing the fire resilience of steel structures located within the wildland–urban interface (WUI). The approach couples Computational Fluid Dynamics using Fire Dynamics Simulator (FDS) with finite element modelling (FEM) in ABAQUS to simulate wildfire exposure and resulting thermo-mechanical structural response. Wildfire conditions are represented using the adiabatic surface temperature (AST), enabling realistic, spatially non-uniform thermal loading to be applied to structural components. The methodology is validated against experimental results from the literature and subsequently applied to a steel portal-frame warehouse under multiple fire scenarios, including standardized curves and realistic wildfire exposures. The results demonstrate the capability of the coupled CFD–FEM framework to reproduce temperature evolution, deformation, and structural degradation under wildfire conditions, supporting enhanced performance-based fire engineering for WUI applications.

ABSTRACT. Introduction The increasing complexity of hazardous scenarios that may affect structures during their service life highlights the need of reliable methodologies for multi-risk assessment. Among potential extreme events, earthquakes and fires, although not necessarily related, may occur in sequence over the build-ing life, generating combined actions that must be adequately assessed through integrated analyses. This work focuses on providing a general procedure to perform multi-risk —earthquake and fire—vulnerability analysis of existing buildings. The authors aim to perform both types of analysis—seismic and thermomechanical—using a single calculation programme. More specifically, in this paper, the au-thors illustrate the procedure for deriving the combined earthquake-fire fragility curves of a reinforced concrete building, using OpenSees (the Open System for Earthquake Engineering Simulation [1]). To achieve this, a numerical investigation was carried out on the OpenSeesForFire module [2] by compar-ing the results with SAFIR® [3], a specific calculation programme for analysis in fire conditions. Several aspects emerged from the comparison that led the authors to improve the current OpenSees implemen-tation to optimise the execution of thermomechanical analyses. Methodology The methodological approach proposed for multi-risk analysis is based on an integrated analytical sequence, which was developed with the intention of evaluating the structural response of existing re-inforced concrete (RC) buildings subjected to seismic events and subsequent fire scenarios. It is worth highlighting that, in this study, hazards do not have any causal correlations. As a matter of fact, the structure is assumed to undergo fire action when its capacity is affected by either limited or mod-erate seismic damage. Such conditions are deemed to be representative of a fire occurring while the structure remains in service after the earthquake. It is possible to distinguish different stages of the workflow (see Fig. 1). Step 1 includes: (a) non-linear static analysis, in which the structure, under gravity loads, is subjected to an increasing horizontal load (i.e., pushover analysis); (b) identification of damage limit states on the resulting capacity curve, necessary to correlate the level of earthquake-induced structural degra-dation with the reduction in load-bearing capacity in case of fire; (c) controlled unloading, gradually removing the horizontal load until the structure exhibits only unreversible seismic damage. Step 2, instead, deals with the assessment of the structural behavior in fire: (a) identification of fire performance levels and fire scenarios, relevant for the structure at issue, and execution of thermal analysis; (b) thermomechanical analysis and stress evaluation. Step 2 is repeated for each fire scenar-io of interest. Step 3 concerns data processing, in fact, it consists of two procedures: (a) calculate the D/C ratios for all fire scenarios and for each seismic damage level considered; (b) perform cloud analysis and derive combined fragility curves based on the selected intensity measures (e.g. peak heat release rate HRR, fire load Qf and duration d) [4]. The proposed approach allows the vulnerability of the structure to be investigated under invariant fire conditions with different initial damage conditions, figuring out the effect of the existing damage on the fire capacity reduction. Proof-of-Concept application In order to obtain integrated earthquake-fire fragility curves and highlight the potential of OpenSees in implementing multi-risk analysis, the methodology described was applied to an existing reinforced concrete building, assumed to be representative of RC buildings in urban areas and located in a high-seismic risk area. Furthermore, it is exposed to significant fire hazard, due to the presence of ar-chives containing combustible materials (i.e., paper and documents). The structure was presumably built before 1990, and it is believed that it was not designed with earthquake-resistant criteria. For the purposes of this work, a plane frame consisting of five spans on two levels is considered. The struc-tural elements are made of C25/30 concrete and FeB44k steel. The columns have dimensions of 35x60 cm2, while the beams have dimensions of 40x60 cm2, both with 4ϕ16 above and below, with a concrete cover of 3 cm. In this work, the preventive validation process allowed the consolidation of a modelling environment that can be used with better reliability to simulate the response of reinforced concrete structures sub-jected to extreme events. Furthermore, as a result of the authors' adjustments to the OpenSees-Thermal code, many limitations in the execution of thermomechanical analyses have been solved. The main conclusions arising from this work are outlined below. - OpenSees can be a very powerful tool to perform multi-risk analysis, especially with the re-quired modifications to the OpenSeesForFire module. - The proposed methodology follows an effective and easy-to-apply scheme, which can be replicated to take into account other dangerous phenomena as well. - Pre-existing seismic damage combined with subsequent exposure to fire significantly in-creases the structural vulnerability, as well as in the case of limited initial damage levels. References [1] Pacific Earthquake Engineering Research (PEER) Center, University of California, Berkeley. [2] Jiang, L., Jiang, Y., Zhang, Z., & Usmani, A. (2021). Thermal analysis infrastructure in OpenSees for fire and its smart application interface towards natural fire modelling. Fire Technology, 57(6), 2955-2980. [3] Franssen, J. M. (2005). SAFIR: A thermal and structural program for modeling structures under fire. Engineering Journal, 42(3), 143-158. [4] Miano, A., Mele, A., Del Gaudio, C., Verderame, G. M., & Prota, A. (2024). Updating of the seis-mic fragility curves for RC buildings subjected to slow-moving settlements. Journal of Building Engineering, 86, 108907.

ABSTRACT. Post-earthquake-fires (PEF) are cascading extreme events in which structures burdened by seismic damage undergo elevated temperatures. One major problem in investigating PEFs is that experimental tests are costly and therefore, are rarely conducted. On the other hand, numerical frameworks often disregard the disruptive effects of earthquake on the structure prior to its exposure to fire. To overcome this limitation, this paper presents a virtual Real-Time Hybrid Fire Simulation (vRTHFS) on a benchmark moment-resisting steel frame to assess the structural behavior during fire with existing earthquake damage. First, the structure is divided into numerical (NS) and virtual-physical substructure (vPS). The frame is subjected to ground excitation. Then, the vPS is exposed to elevated temperatures carrying the seismic damage into to the fire phase. The results indicate that the damage caused by a typical earthquake can reduce the axial load-bearing capacity of members significantly. In addition, the structure may undergo larger deformations under similar forces, resulting in earlier appearance of signs of instability. The vHFT conducted in this study is the first step toward experimental implementation of RTHFS while accounting for the effect of existing earthquake damage. It also indicates that neglecting residual seismic damage at critical regions in a structure prior to its exposure to fire, may result in overestimating structural capacity during fire.

ABSTRACT. Y‑shaped steel columns (Y‑posts) are commonly used to support long‑span roofs in transport hubs, stadia and large atria. Their unique geometry creates fire response mechanisms that differ from prismatic columns: thermally driven joint rotation at the fork, strong sensitivity to temperature differentials between arms, and significant interaction with connected mezzanines. This paper presents a parametric numerical case study based on a Y‑post extracted from the global model of an intermodal station. Non‑uniform thermal actions are derived from a CFD simulation of a train‑fire scenario and are compared with nominal fire curves, including asymmetric exposure of one arm. The study examines (i) time histories of fork rotations and displacements, (ii) redistribution of axial force between arms and stem (load‑path switching), (iii) demand transferred to mezzanine connection, and (iv) local fork deformation assessed with a complementary shell model. Results highlight how small differences in arm heating and rotational restraint at the fork can trigger distinct deformation patterns and stability outcomes, and they motivate practical guidance on mezzanine connection design and local fork detailing.

ABSTRACT. The design for structural fire resistance provisions in the existing Steel Structures Standard, NZS 3404: 1997, were written at the time when design for fire resistance comprised determining that the time to failure in the Standard Fire Test of each structural member, considered in isolation, would equal or exceed the structural fire severity, a committee consensus derived value depending principally on the location and type of construction. Interactions between members and the importance of connections were not expressly considered and barely mentioned. With the standard currently under revision, the opportunity is being taken to incorporate the very large increase in our understanding of steel buildings with composite concrete floors behaviour into the design for fire resistance, including both design and detailing of members and connections to achieve the required performance, which is that the structural adequacy, of the system equals or exceeds the determined structural fire severity, given by FRR or teq with a reserve of strength approx 1.25. There is a two-tier level of design for fire resistance; simple and advanced. The simple design method still predominantly considers isolated member behaviour but covering all structural system members including connections. This approach neglects the interaction effects between columns and adjacent structural components. However, both experimental and numerical studies have demonstrated that such interaction can often trigger localized buckling, as shown in Figure 1.

(a) (b) Figure 1. Internal actions and failure model of the steel column in the simulated sub-assemblage. To improve structural robustness, a new detailing provision and limiting-temperature design requirement for steel columns have been proposed. For interior steel columns with two beams framing in its major axis under four-sided fire exposure, load bearing stiffeners shall be provided for FRR ≥ 60 minutes. A new limiting-temperature design equation, developed specifically for steel columns in this scenario providing a 25% reserve in fire resistance time, corresponding to a Capacity-to-Demand Ratio of 1.25 in the time domain: For the thermal analysis of connections within concrete floor systems, different recommendation factors are adopted based on collective findings from recent large-scale structural fire tests in comparison with EN 1993-1-2: 2024. In the updated design guide, the temperatures of protected connection components attached to the beam web or bottom flange may be taken as 0.8 times the beam temperature at locations away from the connection, while a factor of 0.9 is recommended for unprotected connections. Once the connection temperature is established, the required fire protection can be determined based on the connection’s shear demand and shear capacity. Detailed fire protection requirements for steel connections are provided in the new standard. A notable update concerns the recommended detailing for connections between unprotected secondary beams and protected primary beams, as illustrated in Figure 2. This design recommendation is motivated by the structural behaviour of the primary beam, whose fire performance is primarily governed by its bending moment capacity. When fire protection is omitted from the secondary beam, the resulting thermal influence on the protected primary beam is largely confined to its web. The bottom flange, the critical region governing moment capacity, is only minimally affected both thermally and structurally by the unprotected secondary beam [2].

(a) Web plate connection (b) Flexible end palte connection Figure 2. Recommended fire protection configuration for unprotected secondary beam to protected primary steel beam connected web to web only. One of the key design principles outlined in New Zealand Steel Connection Design Guide of SCNZ 14.1 and SCNZ 14.2 [3] is to provide connections with sufficient rotational ductility for severe earthquake. This helps with fire resistance but is not a complete answer, due to the different rotational demand. The provisions for simple shear connections during both the heating phase and cooling phase have been investigated numerically. The simulation results indicates that the presence of a composite slab significantly influences the overall connection response. A minimum reinforcement ratio of 180 mm²/m is sufficient to prevent bolt shear failure during the cooling phase for FRR <= 60 min [4] due to the connection unzipping. References [1] Standards New Zealand, NZS 3404 1997/2001/2007 Steel structures standard, Wellington, New Zealand, 2007. [2] Meng, F.Q, Clifton, G C. Effect of Coat back on Temperature of Protected Primary Beam Supporting an Unprotected Secondary Beam. 04 Oct 2018. HERA. [3] Steel Construction New Zealand (SCNZ) Steel Connect SCNZ 14.1:2007 & SCNZ 14.2:2007. New Zealand. 2007. [4] Meng, F.Q, Clifton, G C.; Zhu, K.J. Structural fire performance of seismically compatible web-plate connections in a fully composite floor. 5th International Fire Safety Symposium.

ABSTRACT. 1. Introduction Modular construction has rapidly expanded as a sustainable and smart building solution, accelerating schedules, improving factory-controlled quality, and reducing labor. However, fire safety verification for volumetric modular systems remains immature. Conventional standard fire resistance tests typically focus on isolated structural members, making them insufficient for predicting system-level composite thermal response, inter-module heat transfer, and, most critically, installation-sensitive failure mechanisms of non-load-bearing finishes such as modular ceilings. In many modular units, the ceiling provides the only thermal barrier to the steel beams above, commonly consisting of gypsum board layers without active fireproofing. The performance of this assembly in delaying the temperature rise of unprotected RHS (SS275) steel beams, and the sensitivity of its failure time to construction detailing, remain largely unverified at building scale. This study presents rare full-scale experimental evidence on the installation-dependent integrity loss and failure-critical time of gypsum-board-only ceiling assemblies in volumetric modular steel units, providing insight essential for predicting non-load-bearing finish failure time in performance-based design (PBD) frameworks. 2. Full-Scale Experiment Program The two fire experiments were conducted on the full-scale modular unit (3270 mm width × 7950 mm length × 3050 mm height), built following the LPS 1501 guidelines [1], as shown in Figure 1, utilizing an equivalent fire curve to simulate the 1-hour and 2-hour standard fire exposures, respectively. The structural frame consisted of beams, columns, and joists made of rectangular hollow sections (RHS) using structural steel SS275, with no fire protection applied. The columns had cross sections of 200 mm × 100 mm × 6.0 mm matching module heights, and the upper beams had uniform sections of 150 mm × 100 mm × 4.5 mm. To control deflection during lifting and placement, the long-side lower beam thickness was increased to 6.0 mm. The joists (50 mm × 100 mm × 3.2 mm) connected the upper and lower beams, topped by a 100 mm concrete slab at the bottom and a fire-resistant panel roof at the top. The side walls were constructed with 100 mm glass wool insulation and were finished with fire-resistant gypsum board layers. Specifically, two layers of 12.5 mm board were adopted to meet the Korean fire resistance requirements for the 1-hour fire resistance rating. The gypsum board layers served as the primary thermal barrier, providing initial protection against flame and heat penetration, and delaying heat transfer to the structural steel members behind. An additional 150 mm of glass wool insulation was applied to the external walls exposed to outdoor conditions. The ceiling included three layers of a 15 mm fire-resistant gypsum board with no internal partitions or floor finishes installed. The fire originated within the center module and lasted for a maximum of 3 hours including cooling phase. Unlike individual component fire tests, which use furnaces to precisely control temperatures, full-scale fire tests cannot exactly replicate the conditions of a standard fire curve. Thus, an alternative method that approximates standard fire conditions is necessary. The equivalent fire load method outlined in Annex E of EN 1991-1-2 [2] was used in this study, which informed the construction of the fire load (wood cribs) to achieve the equivalent fire curve simulating the 1-hour and 2-hour standard exposures. The fire load densities applied in the experiments were 493 MJ/m2 and 940 MJ/m2, which were used to simulate the target fire resistance periods of 1 hour and 2 hours, respectively. The fire load was implemented by evenly distributing six wood cribs within the center module. These cribs, composed of 50 mm × 50 mm × 1000 mm timber, were configured with 10.5 layers for the 1-hour target fire resistance and 20 layers for the 2-hour target, respectively. Additionally, sandbags weighing 1.6 tons were uniformly placed on the floor of the upper module to apply a uniformly distributed load (UDL) of 0.75 kN/m, in accordance with LPS 1501, which recommends experimental loading for residential compartments. Key instrumentation locations included: (1) Fire temperature within the compartment (thermocouples), (2) Surface temperatures of the gypsum board (inner and outer face), (3) Temperatures of the main steel beam's lower flange and web directly above the ceiling, (4) Five copper disk thermocouples on the walls and floors of the adjacent modules including side and upper module and (5) Vertical displacement at the mid-span of the floor (LVDT). Data was continuously recorded for maximum 180 minutes from ignition. In accordance with LPS 1501 Clause 4.2, the test compartment is considered to have met the fire endurance target if (a) no flame or fire spread occurs into adjoining modules or concealed cavities (integrity), (b) unexposed surfaces of adjacent modules remain below 140 °C on average and 180 °C at any local peak (insulation), and (c) the loaded upper floor sustains the applied load without deflection exceeding span/20 (stability). 3. Principal Experimental Observations and Practical Implications 1) In Test 1 (1-h target; 493 MJ/m²), the ceiling collapsed at 27 minutes, followed by rapid temperature escalation in the unprotected beam. 2) In Test 2 (2-h target; 940 MJ/m²), no ceiling collapse occurred, even under greater thermal exposure, indicating that gypsum-board-only ceilings may exhibit significant installation sensitivity, which governs premature integrity loss. 3) These results do not claim universal generalization, but strongly indicate that construction configuration—particularly fastening layout and spacing—can act as a dominant parameter in integrity-critical time delay, influencing the thermal moderation of the unprotected steel above. The experiments suggest that fire endurance of modular gypsum-only ceiling assemblies cannot be predicted by material selection alone, as installation-level detailing—including fastening spacing, board restraint, and inter-module layout—can govern the integrity-critical time; importantly, the observations indicate that premature integrity collapse of the ceiling finish may occur significantly earlier than the structural frame reaching conventional critical temperature limits (e.g., an average of 538°C and a maximum of 649°C), implying that non-load-bearing ceiling failure, rather than the structural stability limit, becomes the system-governing critical time and must be explicitly assessed in PBD frameworks. References [1] Building Research Establishment (2014) Fire Test and Performance Requirements for Innovative Methods of Building Construction, LPS 1501-1.1, UK. [2] European Committee for Standardization (2002) Eurocode 1: Actions on Structures, Part 1-2: General Actions - Actions on Structures Exposed to Fire, EN 1991-1-2, Belgium. [3] Hopkin, D. J. et al. (2011). Full-scale natural fire tests on gypsum lined structural insulated panel (SIP) and engineered floor joist assemblies, 46(8), 528-542.

ABSTRACT. Unrestrained beams and columns exposed to fire on three sides, due to the presence of slabs or walls, respectively, are common in steel structures. In such cases, the thermal gradient across the cross-section can influence the determination of fire resistance under lateral-torsional buckling (LTB). Eurocode 3 (EN 1993-1-2:2024) allows the use of the maximum steel temperature in the compression flange (θa,com) for beams with Class 1 and 2 cross-sections to determine the reduction factors for steel mechanical properties (yield strength and Young’s modulus) in LTB calculations. However, this temperature may be lower than the temperatures in other parts of the cross-section, which can lead to unconservative results, considering that the formulae in EN 1993-1-2:2024 were developed with uniform temperature distribution in the section [1]. Therefore, this study presents a numerical investigation using GMINA with SAFIR software [2] to evaluate the fire resistance of unrestrained steel I-beams exposed to three-side fire, explicitly considering non-uniform temperature distributions along their cross-sections. The main objective is to understand the accuracy and safety of EN1993-1-2:2024 proposal for those cases, comparing with numerical results.

ABSTRACT. Modular steel buildings are increasingly adopted in residential, educational, and commercial projects because they offer rapid construction, high quality control, and efficient off-site fabrication. However, the rapid degradation of steel strength and stiffness at elevated temperatures makes securing adequate fire resistance a critical design requirement. In modular systems, beams and columns are surrounded by layered fireproofing, insulation, cavities, and concrete slabs, so their temperature histories can differ from those of members tested individually. To design modular buildings both safely and economically, the fire performance of these members must therefore be evaluated in a realistic way that reflects module-specific configurations and heat transfer mechanisms. For this reason, performance-based structural fire design should be implemented at the module level, rather than relying solely on member-level fire resistance ratings that were originally developed for conventional steel frames. In current practice, however, many modular projects still adopt such member-based ratings without modification. This can lead to excessive fire protection thickness, reduced spatial efficiency due to thicker walls and ceilings, and limitations in modular planning and layout flexibility. A more rational design framework requires temperature prediction methods that are tailored to modular assemblies yet simple enough for routine engineering use. Previous work addressed this need at a research level by conducting a full-scale fire test on a representative modular steel unit in accordance with LPS 1501-1.1 and developing a coupled CFD–FEM model. The full-scale test provided gas temperatures and steel temperature histories for key members, while the coupled model reproduced the observed fire-thermal behavior. Although this approach offers detailed insight, setting up and analyzing a coupled CFD–FEM model requires intensive time and cost, and demands specialized expertise in both fire dynamics and finite element analysis. These requirements limit their application in design practice, especially in early design and adjustment stages when many alternatives must be evaluated quickly. In addition, existing simplified formulas for predicting steel temperatures in conventional members are not directly applicable to modular systems. Those formulas typically assume relatively simple cross-sections with uniform protection and idealized heating conditions, whereas modular beams and columns interact with composite slabs, discontinuous linings, and ventilated or closed cavities. As a result, conventional equations may underestimate or overestimate steel temperatures when applied to modular assemblies, undermining both safety and economy. To address this limitation, the present study develops two-dimensional heat transfer models for representative modular steel columns, ceiling beams, and floor beams, and uses them as the basis for deriving simplified temperature-prediction relationships. The 2D models include cross-sections of steel, gypsum boards, insulation, and, for floor beams, concrete slabs. Temperature-dependent thermal properties are assigned to each material, and thermal boundary conditions are specified to be consistent with gas temperatures derived from the full-scale test. Using these models, a parametric study is conducted on key variables such as steel thickness, fire protection thickness, cavity height, and slab thickness. From the resulting numerical database, characteristic heat transfer mechanisms are identified for each member type, and the influence of the governing parameters on maximum steel temperature and heating rate is quantified. Based on these findings, simplified temperature-prediction relationships are proposed that are compatible with the critical temperature method used in structural fire design. The proposed relationships provide a practical tool for estimating steel temperatures in steel beams and columns for modular buildings and, when combined with load ratio, can be used to determine required fire protection thickness at the module level. In this way, the study supports more rational, performance-based fire design of modular steel buildings while maintaining a level of simplicity suitable for routine engineering practice.

ABSTRACT. 1. Introduction Modern buildings require comprehensive safety measures against disasters such as earthquakes and fires. Consequently, there is a growing demand for construction materials that not only possess seismic performance but also ensure structural stability during a fire. To meet these dual requirements, the fire-resistant and seismic steel grade SHN355FR has been developed (KS D 3866, with a nominal yield strength of 355 MPa). SHN355FR steel is characterized by its ability to satisfy both seismic and fire-resistant performance standards. Unlike conventional structural steels, its yield strength does not degrade significantly at elevated temperatures. Furthermore, it is designed to maintain a low yield ratio while preserving fire resistance, thereby preventing brittle fracture through plastic deformation during seismic events. As SHN355FR maintains its room-temperature strength and seismic resistance even when exposed to high temperatures, it is evaluated as a key material for enhancing the safety of structural members. This study evaluates the material properties of SHN355FR at elevated temperatures and compares the fire resistance of unprotected SHN355FR columns with that of conventional H-section steel columns (SM355, KS D 3515). Additionally, a comparative study was conducted to establish a critical limiting temperature for this fire-resistant steel, referencing the average limiting temperature of 538°C (1,000°F) for non-loaded steel columns defined in ASTM E 119.

2. Material Properties at Elevated Temperatures To evaluate the mechanical properties of SHN355FR at high temperatures, tensile tests were conducted at 100°C intervals from room temperature up to 900°C. Since the fire-resistant steel did not exhibit a distinct yield plateau at high temperatures compared to conventional steel, the 0.2% offset method was applied to determine the yield strength. The test results confirmed that SHN355FR exhibits a superior strength reduction factor (higher strength retention) in the range of 200°C to 700°C compared to conventional steel. Notably, at 600°C, the material retained more than two-thirds of its nominal yield strength at room temperature, which satisfies the requirements for fire-resistant steel specified in KS D 3866. Figure 1 illustrates the stress-strain relationship and strength reduction factors at various temperatures. Thermal property tests, including specific heat, thermal conductivity, and coefficient of thermal expansion, were also conducted up to 900°C (specific heat up to 500°C). The results showed that although SHN355FR contains alloying elements such as Chromium (0.1% ~ 0.7%) and Molybdenum (0.1% ~ 0.9%), its basic thermal properties did not differ significantly from those of conventional carbon steel. The thermal properties at elevated temperatures were found to be consistent with the values for carbon steel presented in Eurocode 3.

3. Loaded Fire Resistance Tests To compare the fire resistance of conventional and fire-resistant steel members, loaded fire tests were performed on unprotected H-300x300x10x15 column members with a length of 3 meters. The tests were conducted under load ratios of 0.4, 0.5, and 0.6 based on the nominal yield strength at room temperature. Figure 2 shows the deformation of the SM-A-P4 specimen after testing. The critical temperatures at failure were compared for fire-resistant steel (SHN355FR) and conventional steel (SM355) under identical load ratios. Failure was defined as the point when the column (length h) exceeded a post-buckling deformation of h/100 or a deformation rate of 3h/1000 per minute. The results indicated that the fire-resistant steel maintained structural integrity up to temperatures 50°C to 133°C higher than conventional steel under the same load conditions. Precise numerical analysis using ABAQUS was performed to verify the experimental results. An analysis model capable of effectively simulating the loaded fire tests was developed by incorporating variables such as residual stress, initial imperfection, temperature deviation, and temperature distribution. Using the verified model, the critical temperature of the fire-resistant steel SHN355FR was re-evaluated for various slenderness ratios and temperature deviations. The limiting temperature was determined based on the load ratio of the fire-resistant column corresponding to the critical load ratio of a conventional steel column at an average temperature of 538°C (1,000°F). The analysis indicated an average critical temperature of 626°C for the SHN355FR steel. Consequently, based on the test and simulation data, it is concluded that the limiting temperature for SHN355FR steel can be conservatively set at 600°C. Further details on the experimental and analytical results will be presented in the full paper.

ABSTRACT. Recent advancements in engineering practices and manufacturing methods have enabled the construction of higher consequence mass timber buildings; however, legitimate concerns relating to the fire safety of these buildings persist. Where traditional fire-resistance-based design assumptions may not be appropriate, and given the demonstrated susceptibility of loadbearing timber elements to failure under comparatively mild heating, understanding temperature-dependent variations in the mechanical properties of timber is essential for the fire safe design of higher consequence buildings. This study investigates the influence of structural utilisation (nominally 50% and 20%) and pre-heating moisture content (nominally 0% and 10%) on the compressive response of timber under transient heating. Small-scale specimens were tested in direct axial compression parallel to the grain and subjected to non-charring temperatures until failure. At 50% utilisation, moisture considerably influenced failure temperatures when compared with dry specimens, whereas at 20% utilisation, the presence of moisture appeared to result in differing failure mechanisms. Distinct failure modes were identified between specimen types, attributed to transient thermal and moisture-induced effects. This study highlights and confirms the significance of hydrothermal effects on mechanical response, and the need for further research into the interaction of these factors on the mechanical behaviour of timber structural elements.

ABSTRACT. Encapsulated Mass Timber Construction (EMTC) is a new construction technique that enables the use of mass timber in mid- and high-rise buildings by providing a non-combustible, fire-rated protective layer to delay ignition. Type X gypsum board is a commonly used encapsulation material; however, gypsum-based encapsulation is prone to shrinkage, fall-off, and joint separation under non-standard fire exposure conditions. The current study presents a novel geopolymer-based fabric-reinforced cementitious matrix (FRCM) system as a robust, low-carbon alternative to conventional gypsum-board encapsulation for mass timber elements. A full-size glulam column encapsulated with a 40 mm-thick geopolymer FRCM layer reinforced with a bidirectional carbon grid was fabricated and tested under standard fire exposure in accordance with CAN/ULC-S101 standard. The encapsulation performance of the new FRCM-based system was compared with a reference glulam column encapsulated with three layers of 15.9 mm Type X gypsum board. The FRCM system achieved an encapsulation rating of approximately 90 minutes while maintaining structural integrity, adhesion, and spalling resistance throughout the test. In contrast, the gypsum board system exhibited shrinkage, corner exposure, and board fall-off, which led to flaming combustion and deeper charring of the glulam core section. The new experimental results demonstrate that geopolymer-based FRCM provides superior fire protection compared to conventional gypsum board, underscoring its potential as a high-performance encapsulation solution for mass timber structures.

ABSTRACT. The paper presents the findings of a large-scale natural fire test on a timber frame compartment designed to replicate realistic boundary conditions. The study addresses the limitations of standard fire tests by examining the structural and thermal performance of timber frame systems under natural fire exposure. The compartment, constructed using load-bearing timber studs and engineered floor joists, was subjected to a fire load equivalent to 60 minutes of the standard fire curve, followed by extended monitoring for 24 hours post-burnout. Results indicate that the type F plasterboard protection effectively prevented significant charring of structural members, maintaining integrity and stability throughout the complete burnout of the movable fire load. The measurements and findings from the large-scale fire test can provide essential data for performance-based design and advanced numerical modelling, supporting the application of timber frame construction in modern building practice.

ABSTRACT. Post-earthquake fires (or PEFs) are a low frequency but high consequence event where structural damage, electrical circuitry malfunctions, and gas leaks resulting from earthquake shaking can trigger ignitions, which, in the absence of functional active fire suppression, can potentially lead to severe structural damage and even collapse. Historically, in terms of disaster response and risk mitigation, PEF events have proven to be more difficult to manage as compared to singular, non-sequential hazards. In fact, post-event assessments in some cases have attributed around 80 to 85% of the total damage to the cascading fire event, rather than to the preceding seismic ground motion. For example, the 1906 San Francisco earthquake resulted in extensive urban fires that accounted for most of the structural damage and displaced more than 200,000 of the city’s approximately 400,000 residents. Similarly, the Great Kantō earthquake in Tokyo caused 52 initial ignitions that gradually developed into large-scale conflagrations burning for nearly three days. More recently, over 290 fires were ignited during the 1995 Kobe earthquake. In urban environments with a growing density of mass timber construction, including seismically active regions in Canada, Japan, and the United States, PEF scenarios are an under-researched but entirely possible risk to structural integrity and occupant safety. Yet, despite the risks, very limited information is available on the fire performance of such structures under cascading hazards such as PEF events. A limited amount of research does exist within the behavior steel or concrete frames when subjected to fire or elevated temperatures following an earthquake. However, at present, most full-scale fire studies on timber structures assume that compartment boundaries remain intact. This, of course, may not be the case during or immediately after a design level earthquake. Therefore, a systematic investigation is now urgently needed into the failure modes and residual capacity of a mass timber structures under fire exposure following seismic damage. This research presents a first step towards evaluating the effects of combined fire and gravity demands from a seismically damaged Glulam moment frame.

ABSTRACT. This study uses 3D FE modeling to analyze the thermal-structural response of wood truss assemblies under fire and calibrate a simplified method for predicting the failure time. The FE models are validated against 8 full-scale fire resistance tests. The models reveal that failure of the bottom member, which experiences extensive charring, governed the assembly collapse. Based on a parametric numerical analyses, a simplified method is developed to predict the assembly failure time based on thermal profile in the bottom member for different fires. The simplified method allows prediction of the failure time based on 2D FE thermal analysis of the bottom member combined with an analytical evaluation of its temperature-dependent capacity over time, avoiding the need for 3D FE models of the whole assembly.

ABSTRACT. History records accidents in which the collapse of a structure, occurring after the maximum temperature of a burning gas has been reached, has resulted in considerable loss of life or property. This study assessed different methods used in the fire design of cross-laminated timber (CLT) structures and evaluated their accuracy in predicting the actual behaviour of the structure. The fire resistance of the structure during the cooling phase has been assessed using the Duration of Heating Phase (DHP) -concept. A fire test was conducted in which a loaded CLT slab was exposed to a temperature-time curve based on the DHP-concept, with a heating phase of 35 minutes followed by controlled cooling phase. Consequently, the Burnout-resistance rating (B) according to DHP-concept of the CLT specimen was determined to be at most B 34, as it did not withstand a 35-minute heating phase without failure during the cooling phase. Temperature propagation and structural deflection continued significantly after the heating phase ended. The structure ultimately lost its load-bearing capacity at 263 minutes, meaning the structural collapse occurred 228 minutes after the fire gas reached its peak temperature. The fire test was numerically simulated using SAFIR. With parameters matching the experimental setup, the numerically modelled structure was observed to fail during the cooling phase at 91 minutes. The Burnout-resistance rating (B) according to DHP-concept determined numerically was B25. Based on the results of the fire test and numerical modelling, the B-rating of the assessed structure was estimated to be 53–73 % of its R-rating, while the load ratio was approximately 28 % when calculated according to the new generation of the EN 1995-1-2. The findings of this study indicate that load-bearing failure during the cooling phase is a notable risk for CLT horizontal structures even with relatively short duration of heating phases.

ABSTRACT. The contribution presents a numerical simulation approach to determine the actual temperature-dependent thermal material properties for intumescent coatings using the experimental data from in situ fire tests. Moreover, a literature review on existing numerical simulation approaches and the thermal material properties of intumescent coatings will be included. Several examples are used to demonstrate the developed simulation approach and the analytical functions used to describe the temperature-dependent thermal behaviour of the intumescent coating.

ABSTRACT. The bond between reinforcing steel and concrete governs anchorage and force transfer in reinforced concrete structures and deteriorates significantly when exposed to elevated temperatures. This study develops a prediction approach for the bond performance of reinforced concrete at elevated temperatures using experimental data from multiple studies. Machine-learning (ML) models, including Gradient Boosting (GB), eXtreme Gradient Boosting (XGBoost), K-Nearest Neighbour (KNN) and Decision Tree (DT) regressors, were developed and benchmarked against existing analytical equations and soft-computing approaches. The results show substantial improvements in predictive accuracy, with the GB model achieving R² = 0.99 and 0.97 for training and testing data, respectively. SHapley Additive exPlanations (SHAP) analysis was used to quantify the influence of input variables on the predicted bond strength (Tb). The results confirm the influence of temperature and geometric parameters on bond degradation and demonstrate the potential of explainable Artificial Intelligence (AI) to support structural fire assessment.

ABSTRACT. Unintended ruptures of high-pressure gas transmission pipelines can produce large-scale jet fires caused by immediate ignition of the released gas and impinging radiant heat fields that can cause damage to nearby buildings. There have been several incidents of pipeline ruptures and fires that have impacted buildings or posed a significant risk. For instance, Putra Heights gas pipeline fire in Malaysia in 2025 damaged more than 190 houses [1], the R.M. Palmer Company gas-fueled explosion and fire in Pennsylvania, USA in 2023 severely damaged an adjacent building of the factory and a nearby apartment block [2], and the Lagos pipeline rupture and fire in Nigeria in 2020 destroyed over 50 buildings [3].

While structural fire engineering has largely focused on compartment fires, external exposure scenarios, especially those arising from pipeline ruptures, are under-addressed in practice and regulation. This gap matters because growing urban density and the close siting of critical infrastructure raise the chances that a pipeline rupture will generate thermal loads (i.e., jet-fire impingement and extreme heat fluxes) far exceeding those assumed in standard design fire curves. Therefore, it is necessary to estimate the consequence range of jet fire impingement and thermal radiation. One existing method for estimating consequence zones is the Pipeline and Hazardous Materials Safety Administration (PHMSA) Potential Impact Radius (PIR) equation, which uses a 15.8 kW/m2 radiation intensity to define high-consequence areas (HCA) where people and structures are vulnerable [4]. However, the PIR equation has technical limitations rooted in its simplified physics.

The PIR equation assumes a hemispherical, uniform thermal radiation field around the rupture point. This makes the equation non conservative when dealing with high momentum releases that form directional jet flames. Real rupture fires produce long, angled jets driven by pressure, turbulence, terrain, and wind, as observed in actual fire incidents such as the 2019 Danville gas pipeline rupture and fire [5]. The observed flame development, heat-flux footprint, and damage pattern extended far beyond a circular PIR radius. In addition, The PIR equation produces a fixed-radius consequence zone based on a single intensity threshold, even though real events change over time. The thermal intensity received by a target also varies sharply with its distance and orientation relative to the jet flame, which the PIR equation cannot capture. Moreover, jet fires rotate, collapse, re-ignite, and impinge on structures in changing ways. Wind and slope further distort the thermal footprint. These asymmetric hazards cannot be captured by a single static radius.

Computational fluid dynamics (CFD) model offers a way to replace these simplifying assumptions with physics-based modeling. It captures flame shape, turbulence, radiative heat transfer, and wind effects directly. The result is asymmetric thermal fields and realistic flame impingement on structures. This produces realistic estimate of consequence areas and supports engineering design, emergency planning, and performance-based safety assessments [6]. This study employs CFD-based fire model tool, the fire dynamics simulator (FDS), to model jet fire release, ignition, and heat flux received by buildings. The jet fire scenario is defined based on the Danville incident report [5], as most parameters and damage data are available. Few studies have demonstrated the application of FDS to simulate jet fires, highlighting its effectiveness in capturing complex fire behavior [7], [8].

A three-dimensional computational domain of 200 × 120 × 50 m is modeled, including nearby structures. Boundaries are set to open, except for the ground being an inert surface, as shown in Fig.1a. A finer mesh (grid size of 0.02 m) in the near-field region and a coarser mesh (grid size of 0.1 m) in the far-field domain are used. The jet flame is assumed to be horizontal because the gas release is primarily driven by high-momentum flow in the horizontal direction, as per the incident report [5]. For the numerical simulation, the nozzle is positioned 0.5 m above ground level to represent the observed release geometry and to account for a slight upward inclination of the jet axis relative to the ground plane. The initial mass flow rate of 2968 kg/s was estimated from the operating pressure (936 psi) and diameter (30 inches) using choked flow conditions (under-expanded jet) for full-bore rupture. The thermal radiative factor of 0.15 (i.e., 15% of energy released as radiation) is used for natural gas. From the simulation, flame impingement behavior and the resulting thermal fields are explicitly captured and used in the consequence assessment. These outputs allow evaluation of ignition potential, thermal failure of external wall assemblies and cladding, and the likelihood of heat penetrating into the building core.

The maximum gas temperature in the domain is around 1600 °C, and the maximum temperature measured on one of the exposed structures located 80 m from the jet fire origin is around 1200 °C, as shown in Fig.1b. The measured heat flux ranged from 24 to 58 kW/m2. Although this was recorded over a few seconds, longer exposures yield higher fluxes, leading to internal ignition. These values substantially exceed typical flashover thresholds, 20 kW/m2. This strongly correlates with the damage observed in the Danville pipeline incident at the corresponding structural locations, supporting the hypothesis of severe thermal exposure consistent with jet-fire behavior. This study contributes to establishing a direct link between pipeline rupture fire dynamics (jet fire and radiation) and structural damage. This can enable scenario-based consequence assessment of buildings (existing and new) under these extreme fires. The findings enable more rigorous risk quantification and mitigation for pipeline-adjacent communities, reducing the likelihood that future ruptures translate into catastrophic building damage.

References

[1] A. Press. “A fireball from a burst gas pipeline in Malaysia injures 145 people. ”[Online]. Available: https://apnews.com/article/malaysia-putra-heights-selangor-fire-9e11b47ba0bb2b87cf9271a4de43d4f5. 18/11/2025. [2] NSTB, “UGI Corporation Natural Gas-Fueled Explosion and Fire,” National Transportation Safety Board, West Reading, Pennsylvania, USA, Tech. Rep., 2023. [3] Reuters. “Lagos pipeline blast kills 15, destroys several buildings. ”[Online]. Available: https://www.reuters.com/article/business/energy/lagos-pipeline-blast-kills-15-destroys-several-buildings-nigerian-officials-idUSL8N2B80TT/. 18/11/2025. [4] M. J. Stephens, “A model for sizing high consequence areas associated with natural gas pipelines,” C-FER Technologies, Edmonton, Alberta, Canada, Tech. Rep., 2000. [5] NSTB, “Enbridge Inc. Natural Gas Transmission Pipeline Rupture and Fire,” National Transportation Safety Board, Washington DC, USA, Tech. Rep., 2022. [6] K. McGrattan, R. McDermott, J. Floyd, S. Hostikka, G. Forney, and H. Baum, “Computational fluid dynamics modelling of fire,” International journal of computational fluid dynamics, vol. 26, no. 6-8, pp. 349–361, 2012. [7] A. Rajendram, F. Khan, and V. Garaniya, “Modelling of fire risks in an offshore facility,” Fire Safety Journal, vol. 71, pp. 79–85, 2015. [8] A. Henriksson, “CFD-simulation of hydrogen jet fires using Fire Dynamics Simulator,” Master’s Thesis, Aalto University, 2024. [Online]. Available: https://urn.fi/URN:NBN:fi:aalto-202412117736.

ABSTRACT. Abstract: This paper introduces the recent achievement of an on-going research on the development of AI-driven natural fire testing method to assess the fire performance of structural systems. The concept of the new method and recent experiment investigations will be introduced.

1. The concept

Figure 1 The concept of natural fire testing Figure 1 illustrates the proposed natural fire testing concept, which has the following innovations compared with the conventional furnace testing method,  Furnace tests can only be conducted in laboratory settings. Test specimens usually need to be scaled down, and only simplified boundary conditions and heating conditions (uniform heating) can be simulated. In contrast, the proposed natural fire testing method can be conducted outdoors or even within actual buildings, without scaling—using full-scale specimens with realistic mechanical boundary conditions—and can accommodate diverse and complex fire scenarios (e.g., non-uniform fire exposure).  Furnace tests use the principle of premixed combustion: controlling the gas and air flow into the furnace to create a prescribed fire environment (e.g., a standard fire curve). In contrast, real building fires involve diffusion combustion, where the oxygen necessary for combustion must be supplied from the surroundings. The natural fire testing method adopts this diffusion combustion principle, generating the desired fire environment by designing factors such as room openings and burner heat release rates.  Furnace tests evaluate only fire temperature and cannot evaluate heat release rate (HRR), even though HRR is the most important parameter for quantifying fire hazard. In contrast, natural fire testing evaluates both temperature and HRR, providing a more comprehensive assessment of fire hazards and a more reliable evaluation of the fire performance of new structures and materials.

2. Experiment investigation Figure 2 shows the natural fire tests conducted in a compartment (2.8 m × 2.8 m × 2.6 m) constructed within a steel frame and placed in an open outdoor area. Using an in-house–designed pool-fire burner, two tests (Test 1 and Test 2) were carried out, both producing a fire environment that closely approximated the ISO 834 standard fire curve within the compartment. . (a) test structure (b) pool fire burner (c) fire test photo

(d) calibration of burner (e) test 1 result (f) test 2 result Figure 2 Compartment for the on-site fire test 3. AI compartment fire model Figure shows the AI compartment fire model developed for real-time prediction and regulation of fire parameters of compartment fire test based on deep learning. The model is developed for controlling the natural fire testing procedure.

Figure 3 Architecture of the AI compartment fire model

ABSTRACT. Failure modes of cellular steel beams subjected to combined axial force and bending moments at elevated temperatures are investigated, focusing on the applicability of the second generation of Eurocode. Provisions for steel beams with large web openings at elevated temperatures have restrictions, including limits on axial force and validation only for simply supported beams without axial force interaction, hindering design for moment-resisting frames. A shell finite element model using SAFIR software was developed and validated against full-scale tests from literature. Numerical results are compared to the analytical ones provided by Eurocode, including combined axial force and bending, which falls outside the scope of application. The study revealed that local failure modes are well predicted by design methods incorporating axial force effects, while global failure modes like lateral torsional buckling show discrepancies, with analytical methods overestimating capacity by up to 20% under high axial loads at elevated temperatures. Based on these findings, an experimental campaign will be conducted within the RFCS project TaperFrame to validate numerical models further and enable a parametric study aimed at enhancing analytical approaches for cellular beam design.

ABSTRACT. Machine Learning–Based Prediction of Fire Resistance and Failure Modes of Steel Connections: A Multi-Study Experimental Synthesis with Code Comparison Fire-induced degradation of steel connections governs the structural stability of steel frames subjected to extreme thermal exposures. Existing design provisions, including Eurocode 3 Part 1-2, AISC fire design guidelines, ASCE structural fire calculation methods, and ASTM/NFPA fire resistance test standards, provide important material degradation rules and simplified analytical procedures. However, these approaches treat connection components largely in isolation, rely on reduction factors rather than full connection behavior, and cannot capture the complex, interacting mechanisms observed in real elevated-temperature experiments. Decades of experimental fire tests have generated a rich but highly fragmented empirical record of steel connection failure modes and fire resistance across various geometries, load levels, and heating regimes. This body of evidence has not yet been systematically synthesized or used to evaluate the predictive adequacy of existing fire design standards. This study introduces a machine-learning (ML) framework that aggregates published experimental fire-test data from global research programs and develops predictive models for (a) fire-induced failure mode and (b) fire resistance (time or temperature at failure) of steel connections. The workflow, summarized in Fig. 1, integrates data collection, dataset construction, ML prediction, and direct benchmarking against code-based fire design methods. The objective is to provide a multi-study, data-driven surrogate model for steel connection performance in fire and to assess how prescriptive design rules compare with experimentally validated ML predictions. A unified dataset is created by manually extracting quantitative parameters from experimental programs reported in journals (e.g., Fire Safety Journal, Engineering Structures, Journal of Structural Engineering, Steel and Composite Structures), Eurocode background documents, and NIST experimental fire studies. The dataset includes diverse connection types, such as bolted end-plate connections, fin-plate connections, shear tabs, gusset plates, and welded joints, along with key variables, such as bolt diameter and grade, plate thickness, end-plate configuration, load ratio, initial rotational stiffness, heating curve type, heating rate, measured temperature, time histories, connection rotations, and observed failure modes. Failure modes include bolt shear, bolt fracture, plate yielding, local buckling, weld fracture, prying-action-dominated failure, and excessive rotation leading to connection disengagement or catenary action. Multiple ML models, including gradient-boosted trees (XGBoost), random forests, support vector machines, and neural networks, are trained to predict both failure mode and fire resistance. Cross-validation ensures generalization across studies, and feature-importance analysis using SHAP values identifies the most influential predictors. Results indicate that ML models achieve high accuracy in classifying failure modes and predicting failure temperature or failure time, outperforming simplified analytical expressions when evaluated on nonstandard geometries or high-load scenarios. A key contribution of this study is the development of an ML-based framework for predicting the fire performance of steel connections, along with a quantitative comparison between ML-derived predictions and fire design provisions. While standards such as Eurocode 3 provide reduction factors for material strength and stiffness, they do not explicitly account for bolt relaxation, plate deformation pathways, or multi-mode failures that are commonly observed in experiments. Similarly, AISC and ASCE guidelines often treat connection components independently, without capturing the coupled bolt-plate-weld interactions that govern performance at elevated temperatures. The benchmarking results show that existing codes generally align with ML predictions for simple, conventional connection configurations but can under-predict or over-predict failure temperatures for connections with thin plates, large bolt groups, high load ratios, or nonuniform heating. ML models, grounded in empirical evidence, naturally identify parameter interactions, such as the combined influence of bolt diameter-to-plate thickness ratios and heating rate transitions, that are not explicitly represented in prescriptive rules. The workflow illustrated in Fig. 1 strengthens these contributions by clarifying the link between empirical data, ML prediction, and code comparison. Panel 1 shows the integration of experimental evidence across multiple sources. Panel 2 illustrates the structured extraction of connection variables needed for prediction. Panel 3 depicts the ML models used to classify failure mode and fire resistance. Panel 4 presents the comparison against Eurocode and AISC provisions, enabling a direct evaluation of code performance. Panel 5 highlights the resulting engineering insights, including identification of under-represented parameters in current standards and opportunities for improving performance-based fire design. This paper aims to provide a cross-study, data-driven surrogate for steel connection behavior in fire; it objectively evaluates the predictive accuracy of existing fire design provisions; and it offers evidence-based insights that can inform future revisions of fire design codes. More broadly, this framework indicates that ML, when grounded in experimentally validated structural data, can serve as a powerful complement to traditional fire engineering, supporting rapid assessment, risk-informed decision-making, and improved connection detailing for enhanced fire resilience.

ABSTRACT. There has recently been an increase in the industry need for performance-based design approaches. However, within Canada, the steel standard applying to alternative structural fire design (Annex K) is underutilised. A primary reason is the lack of experimental datasets for validation of modelling to further develop the standard and provide confidence to stakeholders. Therefore, the study herein aims to provide important and novel experimental data for beam-to-biaxial column connections exposed to fire conditions. The experimental setup was designed inspired by an existing high-rise building in Europe which features a biaxial steel frame with the objective of capturing temperature, deflection, and force data. The experimental program consisted of a modified biaxial frame, with shear tab connections between one vertical and one inclined HSS column. Following a previous study in which both columns were vertical, the frame was tested over a methanol pool fire for a total of 18 tests. Thermocouples and Digital Image Correlation with Narrow Spectrum Illumination technology were used to track temperatures, deflections, and forces. All temperatures followed the same patterns with maximum temperatures reaching approximately 550 °C. Uneven thermal expansion of the beams resulted in varying amounts and rates of deflections. The inclined column configuration experienced higher deflection compared to the vertical column, likely due to increased moments generated by the inclined column. Lastly, within the inclined column configuration, the inclined column experienced less horizontal restraining forces than the vertical column.

ABSTRACT. Connections in mass timber buildings degrade during fire due to charring and temperature-dependent material reduction, leading to significant fastener deformation. This study develops a temperature-dependent Beam-on-Foundation (BoF) formulation for CLT balloon-frame connections exposed to the WOODWISE large-scale compartment fire. The ambient-temperature BoF model was first validated against embedment and single-shear tests from the literature, followed by development and validation of a heat transfer model using thermocouple data from WOODWISE Test #3. The validated temperatures were sequentially coupled to the mechanical model, incorporating temperature-dependent degradation and progressive removal of timber springs to represent charring. Results show that the predicted global force–displacement response follows the increase in char depth and spring removal. The framework also provides local force–deformation and stress–strain histories, enabling investigation of thermo-mechanical degradation mechanisms in CLT connections under fire.

ABSTRACT. INTRODUCTION The performance of steel structures in fire is governed by the degradation of steel properties and by thermal expansion, which can generate high axial forces and large deformations, leading to premature failure. Past collapses, such as the World Trade Center towers [1] and the Plasco Building [2] highlight the need to understand the global behaviour of steel frames under elevated temperatures. Although member and sub-frame-level behaviour in fire has been widely studied [3], the influence of beam-to-column connection typology on the fire performance of whole frames remains insufficiently explored. This is a significant issue because joint stiffness and ductility govern axial restraint, force redistribution, and the development of catenary action. The FREEDAM dissipative joint provides controlled rotational capacity through a friction-based slip mechanism [4], offering a promising strategy to limit axial restraint. While its behaviour has been demonstrated under ambient and seismic loading, its system-level performance in fire has not been fully assessed. This study addresses this gap by analysing the fire response of a real steel frame equipped with FREEDAM joints and comparing it with an equivalent frame using rigid joints.

STUDY CASE AND RESULTS The study focuses on the external frames of the DREAMERS building, designed and constructed as part of the RFCS DREAMERS project (RFCS-2020-101034015). The frames include four 6 m spans with HEB 400 fixed-base columns and IPE 450/IPE 400 beams (Figure 1a). Internal joints employ FREEDAM dissipative connections (Figure 1b), while external joints are pinned. The FREEDAM joint incorporates a friction damper at the beam lower flange, allowing slip under bending and providing controlled rotational capacity without damaging structural components [4]. Three fire scenarios (FS1, FS2 and FS3) were analysed, and for each scenario, both the ISO 834 time–temperature curve and a parametric fire curve defined by EN 1991-1-2 were applied. Finite element analyses were conducted using ABAQUS with B31 beam elements and temperature-dependent steel properties. The FREEDAM joints were modelled using a component-based spring approach calibrated to account for the degradation of stiffness and strength with temperature variation observed experimentally [5]. Gravity loads corresponding to 50 % of the joint frictional moment resistance were applied before the fire. Fire performance was assessed based on ISO deformation criteria and on connection failure. For comparison, all analyses were repeated using equivalent frames with rigid beam-to-column joints. The analyses show that the slip mechanism in the FREEDAM joints activates within the first 5–10 minutes of heating, which substantially reduces axial compression forces in the heated beams compared with those observed in the rigid frames (Figure 2a). Across all three fire scenarios, the maximum compressive forces in FREEDAM beams were roughly 1.5 to 2 times lower, demonstrating the reduced restraint provided by the friction device during thermal expansion. This reduction delays vertical deformation and the degradation of flexural stiffness, as depicted in Figure 2b. Under the ISO 834 curve, only the beams reached the deformation limits, and no column failures were observed. FREEDAM joints consistently delayed failure by several minutes: about 6 minutes in FS1 (Figure 2a), 2 minutes in FS2, and 8 minutes in FS3. They also altered the failure mechanism by shifting the first failing member, due to the lower axial compression and resulting redistribution of internal forces. Under parametric fires, none of the scenarios reached ISO deformation limits because peak temperatures were lower, and stiffness partially recovered during cooling. Nevertheless, the joint typology still influenced the response: in the FREEDAM frames, the early reduction of compression promoted a smoother transition to tension and catenary action, whereas rigid frames retained higher compression forces during heating, limiting rotational capacity when tensile forces developed. Connection performance was examined via the spring components. For ISO fires, connection failure occurred only after beam failure, confirming that deformation governed structural resistance. For parametric fires, connection failure occurred only once (FS1), during cooling, when tensile forces increased due to thermal contraction. This highlights that the cooling phase may govern connection safety even when deformation limits are not exceeded. Overall, the results confirm that FREEDAM joints enhance fire robustness by reducing axial restraint, improving rotational capacity, and delaying or preventing failure. These effects support the use of joints with high rotational capacity in performance-based fire design. REFERENCES [1] Sunder SS, Gann RG, Grosshandler WL, Lew HS, Bukowski RW, Sadek F, et al. Final report on the collapse of the World Trade Center towers. Gaithersburg, MD: 2005. https://doi.org/10.6028/NIST.NCSTAR.1. [2] Ahmadi MT, Aghakouchak AA, Mirghaderi R, Tahouni S, Garivani S, Shahmari A, et al. Collapse of the 16-Story Plasco Building in Tehran due to Fire. Fire Technol 2020;56:769–99. https://doi.org/10.1007/s10694-019-00903-y. [3] Santiago A, Simões da Silva L, Vaz G, Vila Real P, Lopes AG. Experimental investigation of the behaviour of a steel sub-frame under a natural fire. Steel and Composite Structures, vol. 8, Techno Press; 2008, p. 243–64. https://doi.org/10.12989/scs.2008.8.3.243. [4] Francavilla AB, Latour M, Piluso V, Rizzano G. Design criteria for beam-to-column connections equipped with friction devices. J Constr Steel Res 2020;172. https://doi.org/10.1016/j.jcsr.2020.106240. [5] Santos, AF, Santiago A, Craveiro H, Simões Da Silva L. Performance of freedam joints under fire-experimental assessment, Coimbra, Portugal: SiF 2024 – The 13th International Conference on Structures in Fire; 2024

ABSTRACT. Fiber-reinforced polymers (FRP) are widely used for strengthening reinforced-concrete (RC) structures, but their performance deteriorates rapidly in fire because the polymer resin loses mechanical strength at high temperatures. Once the glass transition temperature (Tg)—typically 60–100 °C—is exceeded, the polymer matrix softens, causing a sharp reduction in stiffness and bond strength. Common cement-based fire-protection materials (e.g., mortars of sprays consisting of vermiculite, perlite or mineral-fiber) protect steel and concrete but provide only limited protection to FRP laminates. Although these mortars effectively block heat transfer by conduction, moisture inside them heats up, evaporates, and transfers heat by convection, allowing FRP to reach about 100 °C within minutes. While these protection materials are very effective in slowing the temperature rise beyond 100 °C, the FRP may lose much of its bond strength by that stage.

This paper presents the development and testing of a new cement-based vermiculite mortar capable of providing adequate fire resistance to FRP-strengthened members. The goal is to block convective heat flow and delay the temperature rise in the FRP layer. Several moisture-barrier coatings were combined with the vermiculite mortar. The system was evaluated on RC slabs strengthened with externally bonded (EB) FRP laminates. The test procedure included: Slab-1 – Benchmark: An RC slab strengthened with two layers of externally bonded CFRP laminates was tested in three-point bending at ambient temperature to find out its flexural strength. The slab resisted a maximum load of 158 kN of load before failure. Slab-2 – Vermiculite mortar (VM): A CFRP-strengthened slab like Slab-1 but protected with a 45 mm thick cement vermiculite mortar was loaded with a sustained 63 kN (≈40% of Slab 1 strength and greater than the unstrengthened capacity). Exposed to a standard fire, it failed in flexure after 26 min, showing that enough heat passed through the mortar to raise the FRP temperature above Tg. The average FRP temperature reached nearly 70°C at 26 min. It is hypothesized that convection, not conduction alone, caused this rapid temperature rise. Slab-3 – Vermiculite mortar with layers of waterproofing (VML): A similar slab protected with a modified fireproofing system (three intermittent layers of 1 mm thick moisture-barrier slurry combined with vermiculite cement plaster of total thickness 45 mm) survived 40 min of fire before failure. The average FRP temperature was close to 70 °C at 40 min. Since the overall thickness matched that of Slab 2, this confirms that blocking hot-vapor flow increases FRP fire resistance. Slab-4 – Vermiculite mortar with mixed waterproofing (VMM): The same amount of moisture-barrier slurry used in Slab 3 was uniformly mixed into the vermiculite mortar rather than applied in layers. Under identical loading and fire exposure, this slab survived 31 min (compared with 40 min for Slab 3 and 26 min for Slab 2). Fig. 1 shows the schematic of the specimen and the testing protocol. Fig. 2 shows a photograph of the test of slab 2 at elevated temperatures. Fig. 3 plots the average temperatures rise with time at the FRP layer in slabs 2, 3, and 4. Fig. 4 shows force vs time curves and the respective failure durations for slabs 2, 3, and 4. It is concluded that: (i) moisture barriers reduce heat flow through the mortar layer and, when used in multiple layers, can increase fire resistance significantly (40 min versus 26 min); and (ii) discrete moisture-barrier layers are more effective than mixing the waterproofing material uniformly into the mortar.

ABSTRACT. Atria and open stair connections between floors form part of the design aspirations for many modern office buildings. These features introduce a risk of fire affecting multiple floors simultaneously, which can threaten the structure in different ways to a single floor fire, particularly for steel frame structures. This study investigated the challenges, the practical considerations and possible approaches to evaluating structural fire behaviour in multi-floor fires, as well as design measures which can be implemented to address the risks. The study was conducted based on a case study from a recent Arup commercial project.

ABSTRACT. This paper presents the real-world application of performance-based structural fire design involving steel-timber hybrid construction for the proposed expansion of an iconic cultural landmark located in the US.

The expansion includes a hybrid roof structure featuring large-span timber-steel construction. Due to the geometrically challenging design of the roof structure, including a diverse array of steel-timber connections, the project demanded advanced performance-based assessments to accommodate a technically challenging design over conventional prescriptive fire protection strategies.

The thermal response of critical steel-timber hybrid connections were assessed using the 3D solid capabilities of the specialist non-linear finite element program, SAFIR. The structural fire assessments demonstrated that the critical connection categories satisfied the established criteria. The assessments also served to optimize the fireproofing requirements, which were achieved using strategically applied intumescent coatings and supplemental wood blocking. The paper also provides insight on how advanced structural fire engineering approaches utilizing simulations can be leveraged on time-critical projects as an alternative to fire testing, while maintaining code-equivalent safety performance and supporting broader adoption of mass timber construction in hybrid structural systems.

ABSTRACT. In this study, seven full-scale furnace tests were conducted on unprotected steel tubular columns filled with steel-fibre-reinforced concrete. The test setup consisted of multiloaded specimens comprising two-storey columns that were subjected to a compressive load. The lower column was exposed to an ISO 834 standard fire, and a displacement-control horizontal force was applied linearly to the top of the column. All specimens were subjected to compressive and horizontal loads, considering the end restraints provided by the cooler adjacent frame members and forced displacement caused by the thermal elongation of the heated beam during fire conditions. The failure time of the specimen filled with steel-fibre-reinforced concrete was longer than that of the specimen filled with plain concrete, as reported in a previous study, owing to the increase in the high load-bearing capacity of the filled steel-fibre-reinforced concrete. The estimated flexural rigidity decreased with time and approached zero when the column failed. The bending moment at the top of the heated column exceeded that at the bottom. This indicates that the flexural and axial rigidities at the top of the heated column decreased earlier than those at the bottom.

ABSTRACT. This paper presents eleven full-scale furnace tests on roof systems formed by trapezoidal steel profiled sheeting under standard fire exposure. The programme included uninsulated, mineral-wool, PIR, and hybrid mineral-wool/XPS roof build-ups, with spans of 3.0 m and 4.5 m and service loads from 85 to 284 kg/m². The aim was to examine the deformation-controlled fire behaviour of thin-walled profiled roof decks and the influence of load level, thermal regime, and roof build-up on measured fire resistance. In most insulated specimens, the steel-sheet temperature lagged behind the furnace temperature during the first 10 min and then stabilised at approximately 0.85–0.95 of the average furnace temperature once steel temperatures exceeded about 500 °C. Rapid deformation growth was generally observed at average steel temperatures of about 550–700 °C. Two principal deformation regimes were identified: a sudden local-buckling regime with rapid loss of stiffness, and a more gradual regime with continued deformation development and limited redistribution. In one lightly loaded specimen, a tension-dominated large-deformation response was clearly observed beyond the conventional deflection threshold. The results confirm that the fire resistance of profiled roof decking is governed by the interaction of thermal expansion, local buckling, restraint, and force redistribution, rather than by steel temperature alone.

ABSTRACT. Over the past 30 years, fires have caused bridge failures in the U.S. at nearly the same annualized rate as seismic hazards and construction defects. However, fire has garnered significantly less consideration than these other hazards in the design and maintenance of bridge structures. Common overpass bridges with steel plate girders have been particularly susceptible to fire-induced collapse and damage. In June 2023, for example, a gasoline tanker truck crashed and caught fire while crossing under the I-95 overpass at Cottman Avenue in Philadelphia, Pennsylvania, USA. Against the backdrop of the I-95 fire event, this study demonstrates the use of performance-based structural-fire engineering to determine passive fire protection that can achieve varying response objectives for common steel girder overpass bridges. Thermo-structural evaluation of the steel girder bridge superstructure (including the contributions of the composite deck slab) is conducted in three stages: fire modeling, heat transfer modeling, and structural modeling. Per NFPA 502, engineering analysis can be applied to design fire protection for bridges toward preventing the progressive collapse of primary structural elements, with the first priority being the life safety of users, first responders, and bystanders. Minimizing economic impact is the next objective, which can be achieved by enabling a potential return to service following post-fire repair. Performance-based fire protection levels that are tailored to varying levels of post-fire outcomes can then be correlated to hourly ratings, thus facilitating the selection and specification of passive fire protection materials in consultation with suppliers. The potential for application of intumescent paint is a major focus of this study, since those materials present a market-ready solution for durable and weather-resistant passive fire resistance for steel bridge elements.

ABSTRACT. Design of steel structures in fire is traditionally governed by prescriptive provisions and simplified thermal actions defined in EN 1993-1-2, where the structural performance is evaluated under nominal fire curves such as the ISO-834 standard fire. Although these methods provide a consistent procedure for assessing structural fire resistance, they do not capture the complexity of real fire exposures, especially in scenarios involving localised fire, non-uniform heating, or significant interaction between primary structural members and cladding systems. As a result, prescriptive approaches may lead to conservative design of individual steel members. Performance-based fire design offers a more realistic framework by combining natural fire exposure, temperature-dependent material degradation, and nonlinear structural analysis. The use of numerical models in structural design is newly formalised in EN 1993-1-14, which sets the requirements for model verification and validation. Despite increasing research interest, comprehensive full-scale data for steel portal frames under natural localised fires remain scarce. This gap presents a challenge for validating advanced numerical models, including those that explicitly represent the behaviour of semi-rigid joints and temperature-dependent stiffness degradation. A large-scale fire experiment performed within the project OP TAK – Steel hall 4.0 provides a rare opportunity to address this need. The test involved a 24 × 12 m industrial frame structure exposed to a localised fire, with extensive measurements of gas temperatures, steel temperatures, and global displacements, as shown in Fig. 1. The tested frame consisted of IPE 400 columns and an IPE 330 beam connected by bolted endplate joints. The structure was clad with trapezoidal sheeting (TR100/275/0.75 mm) supported by Z-purlins, representing a typical configuration of a light industrial building. The fire source consisted of a 4 m³ wood crib placed beneath frame no. 4, covering an area of 7 m². The ignition method ensured rapid transition to flaming combustion, producing peak gas temperatures exceeding 900 °C. In addition to dead loading, a horizontal force of 15 kN was applied at the frame corner to simulate wind action on a thermally weakened structure. Gas temperatures, steel temperatures, and horizontal displacements were measured throughout the experiment. Additional thermal data were obtained through drone-based infrared imaging and a cooled high-temperature camera. The highest measured steel temperatures ranged between 460–520 °C on the column and up to 620 °C on the beam above the fire. The frame exhibited a global horizontal displacement of approximately 40 mm at peak heating. These measurements provide a detailed representation of the thermo-mechanical behaviour of a real portal frame subjected to a localised fire and serve as a unique dataset for numerical model validation. A key challenge in the fire design of unbraced steel frames is the appropriate representation of joints, whose stiffness and strength significantly influence the global stability of the structural system at elevated temperatures. Simplified member-based approaches in EN 1993-1-2 cannot fully capture these effects, especially when joints experience differential thermal expansion, local yielding, and interaction between individual components. The Component-Based Finite Element Method (CBFEM) provides a suitable framework for addressing these aspects because it models joint components explicitly and accounts for their temperature-dependent stiffness degradation. By combining shell-element representations of members with detailed joint modelling, CBFEM enables a more realistic prediction of system-level behaviour under non-uniform thermal exposure. The present study uses the full-scale experiment to validate a detailed CBFEM model of the tested frame within a performance-based fire design framework. The objectives are to assess the capability of CBFEM to reproduce the measured thermo-mechanical response, to evaluate the influence of thermal gradients and joint stiffness on the global behaviour, and to demonstrate the applicability of validated numerical models in accordance with EN 1993-1-2 and EN 1993-1-14. A detailed numerical model of the tested frame was developed using the Component-Based Finite Element Method (CBFEM) implemented in IDEA StatiCa Member, as shown in Fig. 2. The model explicitly represents the geometry of the columns, beam, and joints, enabling realistic prediction of stiffness degradation and load redistribution at elevated temperatures. Members were modelled using shell elements corresponding to the IPE 400 column and IPE 330 beam profiles. Joint regions—including endplates, bolts, and column bases—were represented as assemblies of individual components following the CBFEM formulation. A materially nonlinear (MNA) analysis was performed to evaluate the thermo-mechanical behaviour. The validation is based on a time-synchronised comparison between measured and computed frame deformations. At each evaluation point, the experimentally recorded temperatures were used as input for the numerical analysis so that the predicted displacements represented the same temperature state as observed during the test.

ABSTRACT. Introduction The assessment of structural performance under fire conditions is becoming increasingly important in modern structural engineering, especially in a context where the prescriptive approach shows significant limitations in defining realistic scenarios. Performance-based fire engineering allows for a more accurate and flexible analysis of the thermo-mechanical behaviour of structural members and for the identification of specific and cost-effective solutions. In this context, the applicability of fragility curves to fire resistance problems represents a methodological advancement of great interest [1]. Although being a consolidated tool in seismic engineering, they have only recently been introduced in the field of fire safety, offering a probabilistic interpretation of the structural response that goes beyond the traditional deterministic assessment. This work fits into this perspective, proposing a performance analysis process aimed at estimating the fire vulnerability of steel structures and identifying, where necessary, the most appropriate protection measures. This work fits into this perspective, proposing a performance analysis process aimed at estimating the fire vulnerability of steel structures and identifying, if necessary, the most appropriate protective measures. Through the realistic fire scenarios, thermal analyses and advanced thermomechanical simulations, fire fragility curves are obtained to describe the probability of reaching specific damage states as a function of fire intensity measures. This approach is an effective tool for supporting more informed design decisions oriented towards the overall safety of the structure. Case Study Application The typical case identified for the purposes of the study is a steel industrial warehouse used for storage. The structural system consists of five portal frames, with single spans repeated in the longitudinal direction at regular spacing. The structure has a plan dimension of (30mx32m) with a height of 9 meters. The double-pitched roof consists of 6° inclined rafters and connecting purlins positioned every 2.5 meters, which both support the covering and transfer the actions to the beams. The roof bracing are arranged in the external spans in order to transfer the horizontal forces to the vertical bracing. Hot-rolled I or H sections are used for beams, columns and purlins. Cold-formed steel sections are chosen for the vertical bracing, while hot-rolled L-shaped sections are used for the roof bracing. All sections of the structural elements are made of S275 steel, with a characteristic yield strength fyk = 275 N/mm², characteristic tensile strength ftk = 430 N/mm² and Young's modulus E = 210000 N/mm². To determine the natural fire curves in the defined fire scenarios, a validation of the zone fire model developed using CFAST was carried out against the CFD model developed using FDS. For this purpose, a simplified portion of an industrial warehouse was modelled. The models simulate the propagation of ignition sources, consisting of pallets stored on racks, according to the travelling fire concept, with the aim of generating a reliable fire curve for subsequent thermo-mechanical analyses. The configuration consists of two adjacent compartments connected by an internal opening, within an unconfined environment where the lateral walls are represented as wall vents, while the floor and ceiling are included to account for thermal stratification and vertical energy exchange. In the first compartment, two overlapping ignition sources are defined, representing multiple or sequential ignition scenarios, whereas in the second compartment a single fire source is introduced as a propagation scenario and activated when a specified autoignition temperature is reached. This setup enables the simulation of fire spread between compartments and the analysis of propagation timing and characteristics. The results from the zone model are then benchmarked against the CFD simulations in FDS to assess model reliability and to identify potential discrepancies in the representation of fire propagation phenomena. This comparison is shown in Figure 2. The thermomechanical analyses were carried out in SAFIR® [2] to evaluate the response of the building under fire conditions. More specifically, all sections of the structural members were modelled in order to perform thermal analyses, which provide the thermal field within the section as a function of time. In the Structural 3D module, the structure is modelled with the constraint conditions and design loads, assigning the sections defined in the previous thermal phase. Thermomechanical analysis provides the evolution of internal forces (N, V, M) and deformations during the fire event. Some results of the advanced analyses are shown in Figure 2, which highlights: the applied natural fire curves; the heated sections of the structural members and an overview of the structural model. Finally, the new methodology described in Figure 1, was applied and through the cloud analysis the fragility curves were derived for different performance levels. The selected intensity measures were the specific fire load and the maximum Heat Release Rate, while, as Engineering Demand Parameter-EDP, the maximum and residual displacements were chosen. All the details will be presented in the full paper. References [1] Ma C., Gernay T.,(2025) Fragility curves for structural fire performance of various composite floor designs under natural fire, Reliability Engineering & System Safety,https://doi.org/10.1016/j.ress.2025.110820. [2] Franssen, J.M., & Gernay, T. SAFIR, a computer program for analysis of structures subjected to fire.

ABSTRACT. Open-plan and highly ventilated compartments often produce non-uniform fire conditions, leading to uneven heating in different parts of beams, columns, and slabs. Research and full-scale fire tests have shown that such non-uniform heating can produce differential thermal gradients and travelling fire behaviour, where the fire travels across the floor plate rather than remain localised [1]. This behaviour can change how loads are redistributed within the structure compared with gravity loading or standard fire exposure assumptions. Several studies have investigated the structural response of buildings subjected to travelling fire scenarios, however, the majority have relied on idealised parametric fire curves with shifted ignition times to represent fire movement across compartments [2], [3]. Considering the highly non-uniform and transient thermal environment observed in actual travelling fire tests, this paper numerically investigates the failure mechanisms of steel frame buildings subjected to temperature distributions derived directly from the Travelling Fire Test data and compares these results with a standard parametric fire model using the same fuel load, geometry, and opening factor as the test [4]. The travelling fire scenarios analysed in this study are based on the full-scale experiments conducted at Ulster University, with a primary focus on Test 1 [4]. The computational analysis is performed using Abaqus software on a steel frame that was used during the travelling fire test. The test compartment of size 15m × 9m with a clear height of 2.9 m included several structural and dummy columns with the opening factor of 0.316 m1/2. Since in the actual fire test, the imposed loads were not considered, and the structure was highly redundant, in this study, different loading scenarios and structural arrangements are considered to understand the behaviour of the structure under imposed loads. The initial results from the computational study revealed different collapse mechanisms under the two different scenarios investigated, as shown in Figure 1. In the parametric fire, the uniform thermal exposure generates significant axial shortening and load redistribution, ultimately causing the concurrent failure of all interior columns and a global structural collapse mechanism. This resulted in the most critical scenario, as it led to the total collapse of the structure. In contrast, the travelling fire produces a localised heating front, leading to progressive deformation and collapse confined to the far-side columns within the proceeding fire zone. These findings highlight that structural failure is strongly influenced by the point of fire initiation within the compartment and how the fire develops over time. Furthermore, it is worth noting that column buckling in the parametric fire scenario initiates near the mid-height, where the temperature distribution along the column is relatively uniform. However, in the travelling fire scenario, failure occurs below the mid-height of the column due to the non-uniform temperature development along its length/height. This observation aligns with findings from our earlier study on isolated columns [5]. As shown in Figure 2a, the parametric fire results in significant axial shortening in interior columns (C-TRL5 and C-TRL7). In contrast, the travelling fire produces more localised thermal exposure, resulting in minimal shortening in C-TRL5 and only moderate shortening in C-TRL7. This reflects sequential collapse rather than simultaneous collapse, as in the case of the parametric fire scenario. A similar trend is observed in the beams, where the parametric fire causes large vertical deflections, particularly in B23. Under the travelling fire, however, beam deflections develop at different times as the fire progresses, with B12 remaining largely unaffected and B23 and B34 exhibiting lower peak deflections compared to the parametric case. Overall, the travelling fire generates less severe global deformation but introduces localised and time-dependent structural demand. Such localised, progressive failure patterns resemble the partial and non-uniform structural collapse reported during the Luton Airport car park fire [6]. Furthermore, the detailed analysis in the full paper will include the structural response of a multi-level steel frame subjected to fire exposure at different floor levels, allowing the influence of fire location on global and local behaviour to be assessed. In addition, the study investigates the performance of the composite floor slab under elevated temperatures to evaluate its deformation, load-carrying capacity, and contribution to overall structural stability during travelling fire scenarios.

ABSTRACT. Passive fire protection is essential to ensure that steel columns maintain stability during realistic fire exposures, yet current prescriptive insulation requirements based on standard furnace tests offer no explicit control over structural reliability. This study extends a recently developed reliability-based safety format for burnout resistance to protected steel columns, treating insulation thickness as a design parameter. The method defines a single design scenario through a target reliability index and incorporates sensitivity weights calibrated using most likely failure point (MLFP) analyses. Heat-transfer calculations follow EN 1993-1-2, and probabilistic evaluations account for uncertainties in fire severity, loading, and insulation properties. Case studies on pinned steel columns with sprayed fire-resistive material (SFRM) demonstrate that insulation thickness strongly influences failure probability: reliability increases with thicker insulation and decreases with higher fuel loads. Comparison with full probabilistic simulations shows the safety format to be consistently conservative. An enhanced format incorporating standardised thermal-property assessments remains conservative while improving consistency. The proposed approach offers a practical, reliability-based procedure for specifying insulation and verifying the burnout resistance of protected steel columns.

ABSTRACT. Identifying the post-fire mechanical properties of high strength structural steel (HSSS) is critical for evaluating the safety of the continued use of HSSS structures after a fire. In realistic fire scenarios, HSSS structures remain under load, which implies that investigations into the post-fire mechanical properties of HSSS should take account of the presence of sustained loading. Accordingly, an experimental study was conducted to explore the effects of pre-tensile stress on the post-fire mechanical properties of Q460 HSSS. Specimens were subjected to five stress ratios of pre-tensile stress, heated to temperatures ranging from 300 to 800 °C and cooled in air under constant pre-tensile stress. Static tensile tests were then conducted. The test results indicated that the effects of pre-tensile stress on the post-fire mechanical properties of Q460 HSSS were significant, depending on both the stress ratio and the maximum experienced temperature. When the maximum experienced temperature did not exceed 500 ℃, pre-tensile stress was found to enhance the post-fire strength and stiffness by up to 16%. However, after experiencing higher temperatures, it caused a reduction of up to 17%. This degradation worsened with greater stress ratio and higher maximum experienced temperature. Supplementary tests on specimens cooled in convection yielded similar results. Therefore, it is recommended to consider the adverse effects of pre-tensile stress when the maximum experienced temperature reaches or exceeds 700 ℃, especially under high stress ratios.

ABSTRACT. This study presents an experimental investigation on the post-fire material properties of SN400B steel subjected to pre-strain. The specimens were pre-strained to the strains of 7% and 10%, and then heated to target temperatures of 300°C, 400°C, 500°C, and 600°C, where they were held for 30, 60, and 90 minutes. After heating, the specimens were water-cooled and subsequently subjected to a coupon tensile test at room temperature. The test result indicated that the strain ageing at temperatures between 300°C and 400°C led to a significant reduction in toughness. In contrast, heating at 500°C and 600°C resulted in a recovery of toughness due to the stress-relieving effect. These findings provide an experimental basis for assessing the reusability of steel structures subjected to post-earthquake fire scenarios, and underscore the importance of considering the combined effects of plastic deformation and fire exposure on structural design and post-fire assessment.

ABSTRACT. The building compartmentation can be achieved by placing a non-load-bearing fire wall connected by means of fusible links to two independent steel structures. However, few studies exist on the fire performance of fusible links. This paper presents the results of a numerical model and full-scale fire test carried within the European FISHWALL research project (RFSC-funded), on fusible systems using aluminum bolts. The 3D thermal-mechanical analysis procedure implemented in the non-linear code ANSYS was carried out to interpret the fire test and to evaluate the behaviour of portal frames exposed to fire. Throughout the test, furnace temperature, specimen temperature, and global displacements of the portal frames were recorded to provide input data for subsequent numerical modelling. The combined fire-test and numerical simulation demonstrate the performance of the proposed fusible systems and validates the modelling approach on full-scale fire test, providing a robust basis for future design recommendations and parametric investigations.

ABSTRACT. “Woodcrete”, a concrete mixture utilising wood chips as a replacement for stone aggregate, offers potential benefits as a structural material including better sustainability credentials, use of timber waste, and a lower thermal conductivity and density. However, for this potential to be realised, its behaviour under fire conditions must be understood. While studies have been conducted on this material at ambient temperature, very limited research has investigated properties under fire conditions. This study investigates the compressive strength of woodcrete at elevated temperatures varying by type of aggregate replacement (coarse and fine) and replacement ratios. Cylindrical samples were heated to 300 °C, 450 °C and 600 °C in a kiln, then loaded to determine compressive strength at these temperatures. Samples with 10% fine aggregate replacement had an average ambient compressive strength of 36.5 MPa, while samples with 10% coarse aggregate replacement had a strength of 25.5 MPa. Compared to the control mix, this was a 16.3% and 41.5% decrease, respectively. Strength loss relative to ambient strength increased with high temperatures, with relative compressive strength around 0.5 at 300 °C and 450 °C, and about 0.3 at 600 °C. The results show that higher wood content, coarse replacement and higher temperatures decrease strength, with greater relative decreases in woodcrete than normal weight concrete. This gives critical initial insight into the behaviour of woodcrete in fire, paving the way for further developments so that the potential of this material can utilised in construction.

ABSTRACT. Double-wythe insulated panels (DWIPs), or sandwich panels, consist of two concrete wythes separated by an insulation layer that enhances fire resilience and thermal performance. The wythes are typically linked with glass fibre reinforced polymer (GFRP) ties because the low thermal conductivity of the GFRP reduces thermal bridging associated with metallic ties and thus preserves the insulating effect of the wall. Ultra-high-performance concrete (UHPC), with its enhanced mechanical properties and fibre reinforcement, allows thinner sections without sacrificing capacity, making UHPC-DWIP walls attractive for precast production, lighter transport, and reduced panel-to-frame dimensions. Despite their growing use, most UHPC-DWIP systems remain non-load bearing because their behaviour under combined loading and fire exposure is not well understood. A major concern is UHPC’s dense microstructure, which increases its risk of explosive spalling at high temperatures. The objective of this study is to conduct a Finite Element Analysis (FEA) of an intermediate scale UHPC-DWIP wall subjected to combined fire and gravity loading. For this purpose, the numerical model will be evaluated using experimental results from previous research, in which an intermediate-scale specimen was exposed to fire on one side following the ULC-S101 procedure without external load. The overarching goal is to develop a model that can conduct parametric analyses and will be validated against future full-scale fire tests. In the long term, the model will improve the understanding and prediction of the behaviour of these structural elements under fire exposure.

ABSTRACT. The elevated temperature behavior of cement mortar modified with hybrid carbon nanotube (CNT) - reduced graphene oxide (rGO) mix has been investigated for multitude of applications in concrete and masonry walling structures. Physical, mechanical, and microstructure studies were carried out at ambient temperature, 200°C, 400°C, and 600°C with varying dosages of CNT-rGO in the cement mortar with different water/cement ratios. The findings from the extensive experimental program revealed that cement mortar modified with hybrid CNT-rGO exhibited superior mechanical performance at the range of elevated temperatures, which is attributed towards the changed pore-structure and increased amount of hydration products - calcium silicate hydrates (CSH) gel. However, notably the dosage of CNT-rGO attains saturation beyond which mechanical properties at elevated temperatures degrade due to agglomeration of CNT adversely affecting formation of the CSH gel.

ABSTRACT. This study investigates the fire behavior of filled composite beams, which are increasingly used in long-span buildings such as logistics and data centers. Although various sectional configurations have been developed in Korea, current prescriptive fire design requires individual certification based solely on steel critical temperature, creating significant practical inefficiencies. To address this, non-loaded and loaded fire tests were conducted on multiple filled composite beam types, followed by FEM-based heat transfer and thermal–stress analyses. Non-loaded fire tests were performed using ISO 834-1 (1999) to evaluate the effects of sectional shape, section width, steel plate thickness, and fire protection. Loaded fire tests applied a 40% axial load with 16 mm SFRM and 1.25 mm intumescent coating to examine thermo-mechanical behavior. The analytical heat transfer model reproduced measured temperature trends within acceptable accuracy, and thermal–stress analysis captured axial deformation behavior under fire. Results showed that the beams could be categorized into four sectional types, with larger sections exhibiting approximately 5% lower steel temperatures due to concrete heat-storage effects. Increasing steel plate thickness further reduced temperatures by 5–10%. These findings provide fundamental data for performance-based fire design of filled composite beams.

ABSTRACT. The technology of grouted sleeve splices is one of the mainly used technologies for the connection of steel rebar in precast concrete elements. This paper presents an experimental study on the post-fire performance of the grouted sleeve slices. A total of 47 groups of grouted sleeve splices specimens were fabricated and tested, considering three reinforcement strength grades (HRB400, HRB500, HRB600), two diameters (14mm and 18mm), and a range of exposure temperatures from ambient to 900 oC. Test results indicate that the surface colour of grout, steel bars, and sleeves changes distinctly with increasing temperature, providing a qualitative reference for assessing fire exposure. The failure mode of slices transitions from bar fracture outside the splice to bond failure as the temperature increases. Slices with high strength rebars showed the transition at a lower temperature. The yield strength remained relatively stable up to 600 oC, while ultimate strength began to decline after 400 oC. Total elongation at maximum force generally decreased, and residual deformation increased significantly with temperature. Predictive models are proposed for post-fire assessment of grouted sleeve splices. The findings provide essential data for post-fire assessment for precast concrete structures.

ABSTRACT. This study investigates the residual mechanical behavior of post-installed metal (PIM) anchors in concrete after fire exposure. PIM anchors can act as thermal bridges during fire, raising temperatures around the anchorage zone and accelerating concrete degradation. Concrete slabs with PIM anchors were heated for 30 or 60 minutes and it cooled and then tested in pull-out. Pull-out test results revealed that mandrel-driven anchors experienced notable reductions in initial stiffness and maximum pull-out strength, especially when edge distances were small. In contrast, expanded-bottom anchors showed minimal performance loss. Failure observations indicated edge cracking or cone-type failures depending on edge distance. Overall, longer heating durations intensified thermal deterioration of concrete around the anchors, significantly impairing the performance of mandrel-driven systems. These results highlight the critical role of concrete degradation in post-fire anchor behavior.

ABSTRACT. 1. Introduction

Hydrocarbon pool fires, typically resulting from tanker-vehicle incidents on transportation infrastructure, generate high levels of radiative and convective heating that pose a significant threat to unprotected steel bridge structures. The inherent complexity of large steel truss bridge geometries makes conventional analytical correlations insufficient for accurately estimating thermal exposures, as these methods cannot fully capture critical physical effects such as plume entrainment, flame tilt, shielding, or localized heat-flux variations introduced by the truss. To address this engineering challenge for a major steel truss bridge in Canada, a detailed Fire Dynamics Simulation (FDS) study was conducted as the crucial initial phase. The core objective of this study was to produce high-fidelity thermal exposure data suitable for use as accurate boundary conditions in the subsequent structural fire analysis.

2. Design Fire Scenarios and Methodology

The design fires were developed based on published data for gasoline burning, typical tanker spill volumes, and reconstructions of prior bridge-fire incidents. Three distinct design fires (DF) were selected to capture a range of intensity and duration: 100 MW, 200 MW, and 300 MW peak Heat Release Rate (HRR). The resulting theoretical flame heights ranged from 16.1 m to 23.6 m. Simulations were carried out until a steady or quasi-steady plume behaviour was established for each HRR scenario. The numerical simulations were conducted using the Fire Dynamics Simulator (FDS).

3. Mesh Sensitivity and Thermal Outputs

To ensure the numerical accuracy of the CFD results, a rigorous mesh-refinement study was performed to identify grid resolutions capable of capturing the gradients in temperature, heat flux, and velocity that govern thermal demands. Mesh resolution was assessed using mesh resolution index (D*/dx). (D*/dx) was analyzed over a range from 6 to 96. Six key parameters were assessed at different elevations above the center point of the pool fire: Fraction of modeled kinetic energy, Average resolved kinetic energy, Average “sub-grid” kinetic energy, Average velocity in the z direction, Average temperatures, and Average temperature fluctuations. Observations indicated that meshes with D*/dx values finer than 50 resulted in a diminishing return (i.e., changes in the parameters used to assess meh quality did not change much). Based on this analysis, a plume resolution index of 70 was selected for the 100 MW simulation, representing an appropriate compromise between the cost of computation and the accuracy of numerical results. The primary output of the FDS modeling was the time-dependent, high-fidelity thermal exposure data used to define the thermal boundary conditions for the structural analysis. These conditions were transferred using the Adiabatic Surface Temperature (AST). Other quantified parameters included Radiative Heat Flux (RHF), Surrounding Gas Temperature (GT), and Convective Heat Transfer Coefficient (HTC). The device outputs were averaged over the last 30 minutes of the simulation to represent continuous exposure

4. Key Results

The FDS results provided detailed insight into the complex heat transfer phenomena controlling the structural fire response. As anticipated, increasing the HRR led to predictable increases in flame height, plume temperature, and RHF. A critical finding was the identification of specific structural elements most vulnerable to direct flame impingement. Elements located closest to the fire, such as a column or portal members, were shown to be totally engulfed in flames, evidenced by similarly very high exposure values for AST, RHF, GT, and HTC registered across all four sides of the member. For instance, Column 1 in the 200 MW scenario was exposed to ASTs higher than 550 oC at all locations from all directions. This severe engulfment behavior is visually summarized in Figure 1 (AST distribution on Column 1, 200 MW). Furthermore, the plume retained significant heat even at high elevations; for the 300 MW fire (where the calculated flame height is 23.6 m), even a beam situated 6 meters higher than the flame height still registered an AST greater than 550 oC (see Figure 2).

5. Conclusion

This detailed FDS study successfully produced the accurate and non-uniform thermal fields required for evaluating open-air hydrocarbon pool fires near a complex steel truss bridge in Canada. The robust numerical methodology, verified via detailed mesh sensitivity analysis. The resultant high-fidelity thermal boundary conditions (AST, RHF, GT, HTC) are indispensable for determining the vulnerability and time-to-failure of unprotected steel members, identifying specific elements likely to experience elevated heat flux, steep thermal gradients, or early critical-temperature exceedance. This methodology supports fire-hazard evaluations for other transportation structures exposed to similar risks.