Are Phase Change Material–Concrete Assemblies in Building Envelopes Fire Safe? Experimental Validation and Numerical Modelling

Abstract

Phase change materials (PCMs) are increasingly incorporated into façades and wall systems to enhance passive thermal regulation; however, their fire safety remains poorly understood. While PCMs effectively reduce cooling loads, limited data exist on their behaviour under real fire exposure. In this study, the thermal response of PCM-integrated concrete panels was investigated through two-dimensional finite element modelling using an apparent heat-capacity formulation that accounts for phase change, latent-heat absorption, and encapsulation softening. Simulations were performed under the ISO 834 standard fire curve and constant furnace exposures between 200 °C and 800 °C for 60 min to evaluate insulation performance and encapsulation stability. Results show that PCM melting at approximately 31 °C provides a 20–25 min delay in rear-face temperature rise under moderate fire exposure (≤400 °C), maintaining the rear-face temperature increase below 180 °C for one hour. Beyond 500 °C, the acrylonitrile butadiene styrene (ABS) encapsulation softens near 95 °C, suppressing latent-heat storage and leading to rear-face temperatures between 260 °C and 360 °C. Comparative analyses indicate that organic PCMs lose effectiveness rapidly unless protected by at least a 25 mm concrete cover, whereas inorganic PCMs exhibit superior stability owing to their non-combustibility and endothermic dehydration behaviour. The results identify performance trends, thermal limitations, and design considerations for the investigated PCM–ABS–concrete assembly under the studied fire exposure conditions. The validated experimental–numerical framework provides insight into the thermal response of PCM-integrated concrete assemblies and supports future development of fire-resilient building-envelope components.

1. Introduction

Buildings are at the centre of the global decarbonisation challenge, accounting for roughly 34% of final energy use worldwide, much of it in the form of heating and cooling demand. Phase change materials (PCMs), which store and release large amounts of latent heat during solid–liquid transitions, have emerged as a promising solution for moderating indoor temperatures, shifting peak loads, and enhancing thermal inertia in lightweight envelopes [1,2]. Over the past two decades, hundreds of experimental and numerical studies have demonstrated the potential of PCMs integrated into gypsum boards, concrete, plaster, insulation layers, and ventilated façades to reduce energy consumption and improve comfort [3,4]. Large European research efforts, such as the STOREPET and 3D-EE-Struct project, have advanced PCM–fibre composites towards market-ready solutions, and systematic reviews confirm their broad energy performance benefits [5,6].

Arinaitwe et al. [7] systematically assessed more than 50 PCM-integrated building materials and found that most organic PCM composites achieve only Euroclass D–E classifications, even when combined with flame retardants, primarily due to paraffin-based fuel contribution and rapid heat release. Similarly, Gnanachelvam et al. [8] demonstrated through small-scale AS 1530.4 fire tests that PCM plasterboards reduced the insulation fire-resistance level (FRL) by up to 30–35% relative to conventional gypsum boards, with rear-face temperatures exceeding 180 °C within 40–45 min of ISO exposure. However, bio-based PCMs [9] and inorganic formulations containing hydrated salts and metallic oxides have shown markedly improved performance, retaining insulation capacity and non-combustibility without contributing additional fuel. This distinction underscores the necessity of shifting toward hybrid or inorganic PCM systems when targeting façade applications requiring Euroclass C or higher.

Furthermore, recent material innovations offer insights into balancing latent heat storage and flame retardancy. Li et al. [10] reported that biochar-supported composite PCMs, enhanced with ammonium polyphosphate (APP) and expanded graphite (EG), achieved UL-94 V-0 flame-retardant classification, reduced total heat release (THR) by 20%, and improved limiting oxygen index (LOI) from 18.3% to 23.7%. These findings align with the present study’s observed benefits of graphite-doped PCMs, which exhibit moderated temperature rise and delayed ignition. Comparable outcomes were also observed in PCM-filled double-glazing systems analyzed by Wang et al. [11], where the addition of a paraffin interlayer delayed critical glass fracture temperature by ≈55 s under fire loading due to transient endothermic melting. In a related context, numerical fire-response modelling of framed glazing systems under fire exposure has also been reported [12], further highlighting the importance of validated assembly-level thermal models. Collectively, these results support the hypothesis that PCM inclusion can remain compatible with fire safety standards if encapsulated with thermally stable matrices, as verified by inorganic and bio-based encapsulation technologies.

Yet, despite this maturity on the energy side, fire safety remains the critical missing piece. The dominant organic PCMs, such as paraffins and fatty acids, are chemically stable and recyclable but highly flammable [13]. Their low ignition thresholds, dripping behaviour, and high heat release rates pose serious risks when incorporated into walls or ceilings [14]. Recent systematic reviews of PCM fire behaviour show that most commercially available PCM composites achieve only Euroclass D or E ratings, far below the thresholds required for regulated façade or insulation products [15]. Even when flame retardants are added, improvements are modest and often compromise latent heat storage capacity.

Experimental studies further illustrate these limitations. PCM-enhanced gypsum plasterboards lose encapsulation integrity at elevated temperatures, leading to melt leakage and increased effective fire load. Cone calorimetry tests confirm that paraffin-based PCMs ignite easily, with critical heat fluxes around 17–20 kW/m² and rapid heat release rates [3,4]. Light steel frame walls lined with PCM plasterboards show reduced fire resistance compared to conventional boards. By contrast, inorganic PCMs such as salt hydrates are non-combustible and can even provide endothermic dehydration, but they suffer from segregation, supercooling, and durability issues [8].

The regulatory and practical consequences of these findings are profound. Following recent façade fire disasters, from Grenfell Tower to Torre Windsor, combustible materials in façades are under unprecedented scrutiny [16]. Reviews of PCM in ventilated façades highlight that while PCMs could greatly improve thermal comfort, their contribution to fire spread risk in cavities is essentially unstudied [17,18]. Without systematic evidence, PCM-based products are unlikely to gain acceptance from regulators, insurers, or façade engineers [19].

This body of work points to a clear research gap: while the energy benefits of PCMs are well established, their fire performance has only been probed in small-scale experiments, with limited data on full assemblies. No systematic modelling framework currently exists to predict how PCM-integrated walls or façades respond under realistic fire exposure, whether under ISO 834 [20,21] standard heating or non-standard façade fire fluxes [21,22].

To address this gap, we ask:

• How do organic and inorganic PCMs differ in their contribution to insulation failure and fire spread in wall assemblies?
• How do thickness, placement, and encapsulation strategies influence time-to-failure under fire exposure?
• Can two-dimensional finite element modelling capture the key thermal processes of PCM melting, latent heat absorption, and potential containment rupture in realistic fire scenarios?
• What design guidelines can ensure PCM integration without compromising fire safety requirements?

In this paper, we develop a 2D finite element fire model of PCM-integrated wall systems using an apparent heat-capacity formulation. We simulate both ISO 834 standard fire exposure and façade-representative heat fluxes derived from fire dynamics simulations, comparing organic and inorganic PCMs under varied placements and thicknesses. By quantifying insulation failure times, temperature evolution, and potential protective strategies, we establish performance thresholds and propose design rules for safe PCM deployment in façades and walls. In doing so, this work provides the first systematic, assembly-level modelling framework to evaluate PCM fire safety—bridging the gap between latent heat innovation and regulatory acceptance. The numerical framework was calibrated and validated using experimental temperature data from furnace tests on PCM–concrete assemblies.

2. Experimental Details

Building upon our previous work [23], where we demonstrated that sandwich composite panels incorporating PCMs can significantly reduce peak heat flux and delay temperature rise under cyclic thermal loads, this study extends the application domain toward fire safety. In that earlier investigation, the hybrid composite configuration exhibited not only improved energy efficiency but also retained acceptable mechanical performance under ambient and moderate thermal conditions. The observed delay in thermal transmittance and the buffering effect of the PCM core offered promise for passive thermal regulation in buildings. Motivated by these results, the current study presents the next logical progression—evaluating the fire performance of the same PCM-integrated composites under standardized high-temperature conditions. This transition from thermal efficiency to fire resilience aims to validate the material system for broader safety-critical applications.

2.1. Materials and Sample Preparation

The PCM-integrated concrete panels were designed to represent a typical façade-scale composite element with embedded latent heat functionality. The system combined ordinary Portland cement (OPC)-based concrete as the structural matrix and macro-encapsulated organic PCM enclosed within 3D-printed acrylonitrile butadiene styrene (ABS) shells serving as thermal reservoirs. ABS was selected as the PCM encapsulation material to serve as a representative thermoplastic shell for macro-encapsulated PCM systems. ABS is commonly employed in PCM encapsulation studies due to its compatibility with fused-deposition modelling (FDM), dimensional stability, and ease of fabrication, which allow precise control over encapsulation geometry and PCM mass fraction. Importantly, ABS exhibits a well-defined glass transition temperature (T_g ≈ 95 °C), making it suitable for investigating encapsulation softening, loss of containment, and degradation under elevated temperatures.

Alternative encapsulation materials such as polyethylene, polypropylene, epoxy-based shells, aluminium capsules, or fully inorganic containment systems offer higher thermal stability or non-combustibility. However, these materials introduce trade-offs related to manufacturing complexity, brittleness, cost, or reduced adaptability for façade-scale integration. In the present study, ABS was intentionally selected as a conservative polymeric encapsulation to quantify the fire-induced limitations of organic PCM systems and to establish thermal safety thresholds. This choice enables a clear comparison with inorganic PCM systems, which are subsequently evaluated numerically as fire-resilient alternatives.

The PCM-filled ABS macro-encapsulation moulds (PCM container) were first prepared separately and allowed to solidify at ambient temperature. During panel fabrication, the filled ABS mould was positioned at the centre of the panel thickness using a temporary wooden frame, which served only as an alignment and positioning aid.

The assembled unit was then placed inside a rigid steel mould with internal dimensions of 300 × 300 × 50 mm³, which acted as the external formwork for concrete casting. Fresh concrete was poured into the steel mould in a single casting operation, fully embedding the ABS macro-encapsulation within a monolithic concrete panel. The wooden frame was removed after initial setting and was not part of the final panel.

After casting, the panels were demoulded following standard curing procedures. The final test panels therefore consisted of a single concrete panel with an internally embedded PCM-filled ABS macro-encapsulation, rather than a sandwich of separate concrete layers.

Commercial paraffin-based RT35 (PCM melting point ≈ 31 °C, latent heat ≈ 190–200 kJ kg⁻¹) was selected owing to its widespread use in building envelope applications and prior thermal reliability in energy-efficiency studies by the authors. To enhance thermal conductivity and stability, 5 wt% expanded graphite (EG) was mixed into the PCM prior to encapsulation, yielding a composite PCM (RT35–EG) with improved heat diffusion and reduced leakage tendency. An EG content of 5 wt% was selected as a practical compromise to enhance PCM thermal conductivity and stability while maintaining acceptable molten-fill viscosity and limiting the reduction in latent heat capacity that would occur at higher EG loadings.

The PCM containers were fabricated via fused-deposition modelling 3D printing using ABS filament (glass transition temperature (T_g) ≈ 95 °C, melting ≈ 105 °C). The PCM container refers to the 3D-printed ABS macro-encapsulation unit (ABS wall thickness: 2.5 mm) filled with PCM. After printing, molten RT35–EG was injected into the macro-encapsulation at 50 °C under vacuum and sealed using a solvent-bonded ABS lid. The nominal PCM mass fraction per panel was 8 ± 0.5 wt%.

Panels were cast as monolithic concrete panels using rigid steel moulds (300 × 300 × 50 mm³), which define the final panel dimensions. During casting, a temporary wooden frame was used only as a positioning aid to hold RT35 PCM-filled ABS macro-encapsulation at the intended location within the mould; the wooden frame was removed after initial setting and was not part of the final panel. Figure 1 summarises the overall fabrication, instrumentation, and testing workflow.

The PCM–EG composite (RT35 paraffin PCM modified with 5 wt% expanded graphite) was injected in molten state at 50 ± 2 °C into the 3D-printed ABS macro-encapsulation mould cavities using a vacuum-assisted filling procedure to minimise air entrapment and ensure complete cavity filling. A vacuum level of approximately −0.08 MPa was applied during injection, followed by gradual pressure equalisation. The typical filling time for each mould cavity was 3–5 min depending on volume. After filling, the moulds were sealed using solvent-assisted ABS bonding and allowed to cool to ambient temperature prior to embedding into the concrete panels.

The injection temperature of 50 °C was selected to ensure complete liquefaction and low viscosity of RT35 while remaining well below the glass transition temperature of ABS (≈95 °C), thereby enabling uniform filling without inducing thermal deformation or degradation.

Concrete was prepared using OPC (42.5 R), river sand, and 10 mm crushed granite aggregate, with a water–cement ratio of 0.45 and a target density of 2300 kg m⁻³. The mixture exhibited an average compressive strength of 37 MPa (28 days). Concrete casting was performed in two successive steps to ensure correct placement of RT35 PCM-filled ABS macro-encapsulation and to obtain uniform concrete cover. An initial concrete layer was placed and lightly compacted, the ABS macro-encapsulation was positioned, and the remaining concrete was poured to reach the final thickness; each step was lightly tamped and vibrated to minimise air entrapment.

Panels were demolded after 24 h and water-cured for 28 days at 23 ± 2 °C. Prior to fire testing, panels were oven-dried at 40 °C for 48 h to remove residual moisture. Final test panels measured 300 × 300 × 50 mm³. The numerical model represents an equivalent cross-sectional (through-thickness) domain consistent with the panel thickness (50 mm) and the embedded PCM-container configuration, rather than the full 300 × 300 mm plan area. Thermocouple ports (Ø 2 mm) were drilled at 5 mm from the hot face, mid-plane, and 5 mm from the rear face to facilitate direct comparison between experimental and simulated temperature histories. Thermocouples were embedded in the concrete at 5 mm from the exposed face, at mid-thickness, and 5 mm from the rear face. The PCM core temperature was not measured directly, as instrumenting the PCM within the sealed ABS mould would require breaching the encapsulation and could influence leakage/heat-transfer behaviour. PCM thermal response was therefore inferred from the measured concrete temperatures and the validated numerical model. The sample preparation is further detailed in ref. [23].

Table 1. Thermophysical properties of materials used in PCM-integrated concrete panels [23].

Material	Density, ρ (kg m−3)	Thermal Conductivity, λ (W m−1 K−1)	Specific Heat Capacity, cp (kJ kg−1 K−1)	Phase-Change Temperature Range (°C)	Latent Heat (kJ kg−1)	Remarks
Concrete (OPC-based)	2300	1.45	0.88	–	–	Reference matrix material
PCM (RT35)	820	0.20	2.0–3.5	28–35	190–200	Organic paraffin PCM (melting ≈ 31 °C)
PCM–EG composite	870	0.52	2.2–3.8	28–35	185–195	5 wt% expanded graphite for enhanced λ
ABS (encapsulant)	1040	0.18 → 0.12	1.3	–	–	Tg ≈ 95 °C; softens > 90 °C
Air gap/interface	1.2	0.026	1.0	–	–	For convection/radiation boundary estimation

summarizes the thermophysical properties used in both experimental evaluation and modelling. The PCM–EG composite showed a thermal conductivity of 0.52 W m⁻¹ K⁻¹, a density of 870 kg m⁻³, and an effective specific heat varying between 2.2 and 3.8 kJ kg⁻¹ K⁻¹ across the phase-change range. The concrete matrix exhibited λ = 1.45 W m⁻¹ K⁻¹ and ρ = 2300 kg m⁻³. ABS encapsulation softened above 95 °C, and its conductivity decreased from 0.18 to 0.12 W m⁻¹ K⁻¹ during heating, consistent with manufacturer data and differential scanning calorimetry (DSC) analysis.

To ensure the reliability of results, all thermal measurements were performed in triplicate using identical panels fabricated under the same materials, encapsulation, and curing conditions. Temperatures were recorded with calibrated K-type thermocouples (±0.5 °C accuracy) at one-second intervals, and surface temperatures were verified with a Testo 834 infrared camera (thermal sensitivity = 0.06 °C). The standard deviation of repeated tests remained below 1.2%, confirming the excellent repeatability of PCM melting and overall thermal response, and validating the robustness of the experimental–numerical methodology.

2.2. Fire Exposure Setup and Conditions

Fire testing was conducted using a programmable vertical furnace designed to replicate both steady-state and ISO 834 standard heating conditions. The furnace comprised a refractory-lined steel chamber (0.6 × 0.6 × 0.6 m³) with uniformly distributed electrical heating elements, capable of maintaining stable temperatures up to 900 °C. Temperature control was achieved through a closed-loop PID system, with feedback from shielded thermocouples located near the panel’s exposed surface (Figure 1).

Two heating regimes were employed:

• Constant-temperature exposures at 200 °C, 400 °C, 600 °C, and 800 °C for 60 min each, representing controlled façade fire scenarios of increasing severity.
• Standard ISO 834 fire exposure, following the temperature–time relationship (Equation (1)):

$$T_g=20+345{\log }_{10}\left(8t+1\right)$$

(1)

where T_g is the furnace temperature (°C) and t is the exposure time (min). All fire exposure tests were conducted in an electrically heated furnace (electric resistance heating) operated under closed-loop PID temperature control. The ISO curve was executed for 60 min, reaching approximately 840 °C at the end of the test. The furnace was controlled to follow the ISO 834-1 standard time–temperature curve. The furnace was programmed to follow the ISO 834 time–temperature curve, for which the theoretical value at 60 min is approximately 945 °C according to the ISO expression. In this study, the furnace temperature was regulated using a closed-loop PID controller based on internal control thermocouple feedback within the furnace chamber. However, because the tests were conducted in a small-scale configuration with a partially open furnace interface and unavoidable heat exchange at the panel boundary, the measured exposed-surface temperature of the panel was lower than the ISO setpoint. “Here, ‘open furnace configuration’ refers to the panel being mounted at the furnace opening (front-face exposure) without a fully sealed enclosure around the panel perimeter. The furnace was electrically heated and controlled via a closed-loop PID system to follow the ISO 834 time–temperature curve based on internal thermocouple feedback (gas/control temperature). Owing to the open-front interface and associated air exchange and heat losses, the effective heat flux to the panel and the measured exposed-surface temperature can be lower than the theoretical ISO gas temperature; this does not indicate a lack of ISO curve control, but reflects panel-level thermal coupling. This reduction is attributed to combined effects of (i) radiative losses and reduced radiation view factor at the opening, (ii) convective losses and air exchange/leakage near the panel interface, (iii) spatial gradients between the furnace control location and the panel surface, and (iv) the panel acting as a heat sink during transient heating. Therefore, the reported ≈840 °C corresponds to the measured panel exposed-surface temperature at 60 min, while the ISO value of ≈945 °C refers to the standard furnace temperature defined by ISO 834 (EN 1363-1 [24]).

Each PCM-integrated concrete panel (300 × 300 × 50 mm³) was mounted vertically with its hot face exposed to the furnace cavity and lateral faces insulated using ceramic fiber wool to enforce near one-dimensional heat transfer. Type K thermocouples (±0.5 °C accuracy) were embedded at 5 mm from the exposed face, mid-thickness (25 mm), and 5 mm from the rear face to measure transient temperature profiles. Readings were acquired at 1 Hz using a 16-channel NI-9213 data acquisition system. The heating regimes employed in the experiments, including constant-temperature exposures and the ISO 834 standard fire curve, are illustrated in Figure 2.

The ISO 834 standard fire curve defines a continuous transient furnace temperature profile, with characteristic reference times of 30, 60, 90, and 120 min corresponding to increasing fire-severity levels commonly used in fire-resistance classification. In the present study, the ISO 834 exposure was applied directly according to the standard equation, and the thermal response was analysed primarily up to 60 min. This duration was selected because it represents a typical façade and wall fire-resistance benchmark (EI 60) and allows direct experimental validation without compromising panel integrity. Although longer durations such as 90 and 120 min were not investigated experimentally, the validated numerical framework can be readily extended to these durations in future studies.

To verify thermal uniformity, the furnace temperature was continuously logged and cross-checked with the programmed profile. Surface temperatures were independently monitored using a Testo 834 infrared thermal imaging camera (thermal sensitivity = 0.06 °C), calibrated with a reference blackbody. The emissivity of concrete (0.92 ± 0.02) was used to ensure reliable readings.

After exposure, panels were cooled to ambient temperature before UPV testing to assess internal cracking and residual stiffness. The combined use of embedded thermocouples, IR imaging, and controlled heating profiles provided high-fidelity thermal data for validation of the numerical simulations presented in Section 4.

2.3. Numerical Modelling Framework

A two-dimensional transient heat transfer model was developed using the finite element method (FEM) implemented in FEniCS (v2023.1) to simulate the coupled thermal response of the PCM-integrated concrete panels during fire exposure. The computational domain replicated the experimental panel (300 × 50 mm), consisting of a concrete matrix with embedded PCM container arranged at mid-depth. The model aimed to capture temperature-dependent conductivity, latent heat absorption during melting, and the softening behavior of the ABS encapsulant under elevated temperatures.

The governing heat-transfer equations employed in this study are based on standard transient heat conduction and apparent heat-capacity formulations for phase change materials. All numerical constants and material parameters appearing in the equations are reported in Table 1 and in the boundary-condition descriptions, and were obtained from experimental measurements, manufacturer data, and calibration against the present experimental results. The numerical solution was implemented by the authors in FEniCS using these established formulations. The source codes are available from the corresponding author upon reasonable request.

The transient heat conduction equation with phase change is expressed in Equation (2).

$$\rho c_p{\displaystyle \frac{\mathsf{\partial }T}{\mathsf{\partial }t}}=\nabla ·\left(k\nabla T\right)+\rho L{\displaystyle \frac{\mathsf{\partial }f}{\mathsf{\partial }t}}$$

(2)

where $\rho$ is the material density (kg·m⁻³), $c_p$ is the specific heat capacity (J·kg⁻¹·K⁻¹), $T$ is temperature (K), $t$ is time (s), $k$ is the thermal conductivity (W·m⁻¹·K⁻¹), $L$ is the latent heat of fusion of the PCM (J·kg⁻¹), and $f$ is the liquid fraction of the PCM, defined in Equation (3).

$$f\left(T\right)=\left\{\begin{array}{cc}\hfill 0,& T<T_{solid}\hfill \\ \hfill {\displaystyle \frac{T-T_{solid}}{T_{liquid}-T_{solid}}},& T_{solid}\le T\le T_{liquid}\hfill \\ \hfill 1,& T>T_{liquid}\hfill \end{array}\right.$$

(3)

This formulation explicitly accounts for latent heat absorption and release through the liquid fraction function, ensuring a physically consistent representation of the phase transition during melting and solidification.

The latent heat of the PCM was represented using the apparent heat capacity method, which distributes $L$ over a small temperature interval $\left[T_s,T_l\right]$ (solidus to liquidus) (Equation (4)):

$$c_p^{*}\left(T\right)=c_p\left(T\right)+{\displaystyle \frac{L}{T_l-T_s}}$$

(4)

for $T_s\le T\le T_l$.

Outside this range, $c_p^{*}\left(T\right)=c_p\left(T\right)$.

For the RT35–EG PCM, $T_s=28$ °C and $T_l=35$ °C, with $L=190 \mathrm{kJ}{ \mathrm{kg}}^{-1}$. This approach allows smooth energy absorption and release during melting and solidification without tracking phase interfaces explicitly.

Phase change is captured via an apparent heat capacity $c_p^{*}\left(T\right)$ that embeds latent heat within a narrow “mushy” interval around $T_m$. For the PCM subdomain (Equation (5)):

$$\begin{gathered} \phi \left(T\right)=½\left[1+tanh\left({\displaystyle \frac{T-T_{\mathrm{m}}}{\frac{\Delta T}{4}}}\right)\right],{\displaystyle \frac{d\phi }{dT}}=½·\left({\displaystyle \frac{4}{\Delta T}}\right)\left[1-tan{h}^2\left({\displaystyle \frac{T-T_{\mathrm{m}}}{\frac{\Delta T}{4}}}\right)\right] \\ {c}_{\mathrm{p}}\ast \left(T\right)=(1-\phi )c_{\mathrm{p},\mathrm{s}}+\phi c_{\mathrm{p},\mathrm{l}}+L·(d\phi /dT). \end{gathered}$$

(5)

Here, φ ∈ [0,1] is a smooth surrogate for the liquid fraction, $\Delta T$ the mushy-zone width (we use $\Delta T\approx 1.5\mathrm{K}$ unless otherwise reported), and $L$ the latent heat. Thermal conductivity and density are interpolated similarly (Equation (6)):

$$k\left(T\right)=\left(1-\phi \right)k_{\mathrm{s}}+\phi k_{\mathrm{l}},\rho \left(T\right)=\left(1-\phi \right){\rho }_{\mathrm{s}}+\phi {\rho }_{\mathrm{l}}$$

(6)

This regularization provides (i) a differentiable response for robust nonlinear solves and (ii) a physically consistent plateau in temperature–time histories during melting.

The thermal and mechanical response of the ABS encapsulation was temperature-dependent. Softening was incorporated by reducing its thermal conductivity $k_{ABS}\left(T\right)$ and density ${\rho }_{ABS}\left(T\right)$ above the glass transition temperature $T_g\approx 95$ °C using exponential degradation functions (Equation (7)):

$$k_{ABS}\left(T\right)=k_0 e^{-\alpha \left(T-T_g\right)},{\rho }_{ABS}\left(T\right)={\rho }_0\left[1-\beta \left(T-T_g\right)\right]$$

(7)

with $\alpha =0.012$ °C⁻¹ and $\beta =1.5\times {10}^{-4}$ °C⁻¹, based on manufacturer data and DSC analysis. This representation allowed the model to reproduce the loss of encapsulant stiffness and increased heat transfer once the PCM container integrity began to degrade.

To represent macro-encapsulation failure under fire, we introduce a smooth softening indicator $S\left(T\right)$ based on the PCM-adjacent temperature (Equation (8)):

$$S\left(T\right)=½\left[1+tanh\left({\displaystyle \frac{T-T_{soft}}{\delta }}\right)\right]$$

(8)

with $\delta$ a small smoothing constant (e.g., 2 K). The effective PCM properties used in simulation are then blended as (Equation (9)):

$${{\rho }_{\mathrm{pcm}}}^{\mathrm{eff}}=\left(1-S\right)\rho \left(T\right)+S{\rho }_{\mathrm{l}},{k_{\mathrm{pcm}}}^{\mathrm{eff}}=\left(1-S\right)k\left(T\right)+{Sk}_{\mathrm{l}},{c_{\mathrm{p},\mathrm{pcm}}}^{\mathrm{eff}}=\left(1-S\right)c_{\mathrm{p}}\ast \left(T\right)+{\mathit{Sc}}_{\mathrm{p},\mathrm{l}}.$$

(9)

Physically, for $T\ll T_{\mathrm{soft}}$, the PCM behaves with full latent-heat capability; for $T\gg T_{\mathrm{soft}}$, the latent term vanishes and the PCM conducts/stores heat as liquid only, mimicking leakage/rupture or the practical loss of latent buffering once containment softens.

Why does this matter for fire safety? This switch is intentionally conservative for the insulation criterion: once softening begins, the model stops crediting the latent-heat buffer, so rear-face temperatures rise faster, matching the risk-focused aim of this study.

At the exposed (hot) surface, a combined convective–radiative heat flux boundary condition (Neumann type) was applied. The furnace temperature (constant or ISO 834 profile) was used as the reference temperature in the convection and radiation terms driving the heat flux into the panel. The opposite (rear) surface was assigned a convective–radiative boundary (Equation (10)):

$$-k{\displaystyle \frac{\mathsf{\partial }T}{\mathsf{\partial }n}}=h\left(T-T_{\mathrm{\infty }}\right)+\epsilon \sigma \left(T^4-T_{\mathrm{\infty }}^4\right)$$

(10)

Here, $T$ is the surface temperature of the panel and $T_{\mathrm{\infty }}$ is the ambient temperature (both in K); the terms $T^4$ and $T_{\mathrm{\infty }}^4$ represent the fourth powers of temperature, corresponding to radiative heat transfer as per the Stefan–Boltzmann law, where $h=10 {\mathrm{W} \mathrm{m}}^{-2}{\mathrm{K}}^{-1}$, $\epsilon =0.9$, and $\sigma =5.67\times {10}^{-8} {\mathrm{W} \mathrm{m}}^{-2}{\mathrm{K}}^{-4}$. All other surfaces were assumed to be adiabatic. The initial temperature of the entire domain was set to 25 °C.

At the hot face (furnace side), the net heat flux is the sum of convection to furnace and thermal radiation (Equation (11)):

$${q^{″}}_{hot\left(T\right)}=h\left[T_{\mathrm{s}}-T_g\right]+\epsilon \sigma \left[T_{\mathrm{s}}^4-T_g^4\right].$$

(11)

with $h=25 {\mathrm{W} \mathrm{m}}^{-2}{\mathrm{K}}^{-1}$ and $\epsilon =0.8$ baseline values, and $T_g$ either a constant (200/400/600/800 °C plateaus for 60 min) or ISO 834 $T_g\left(t\right)$. At the rear face, we apply natural convection to ambient: $q_{rear}^{″}=h_{nat}(T_s-T_{\mathrm{\infty }})$ with $h_{nat}=5 {\mathrm{W} \mathrm{m}}^{-2}{\mathrm{K}}^{-1}$, $T_{\mathrm{\infty }}=293.15 \mathrm{K}$. Laterals are adiabatic.

In the model, the exposed surface was subjected to a combined convective–radiative heat flux boundary (Neumann condition) interacting with a furnace following the ISO-834 temperature–time curve. The surface temperature therefore evolved naturally during the simulation and was not fixed, preventing over-specification of boundary conditions. The numerical model geometry and applied thermal boundary conditions used for the finite element simulations are illustrated in Figure 3. The numerical geometry is an equivalent 2D cross-sectional representation of the experimentally embedded ABS macro-encapsulation mould. The concrete–ABS–PCM–ABS–concrete arrangement preserves the actual through-thickness thermal path (concrete cover → ABS wall → PCM cavity → ABS wall → concrete cover) that governs rear-face temperature evolution under furnace exposure. The equivalent 2D representation was configured to remain consistent with the experimental PCM/ABS fraction (including the measured PCM mass fraction) and was used because the side faces were insulated during testing, promoting predominantly one-dimensional through-thickness heat transfer.

Although the experimental specimen possesses a three-dimensional embedded PCM geometry, the thermal response during furnace exposure is governed primarily by through-thickness heat transfer due to: (i) the relatively large in-plane dimensions compared with panel thickness (300 × 300 × 50 mm³), (ii) lateral insulation applied during testing, and (iii) symmetric placement of the PCM insert within the panel. Accordingly, an equivalent two-dimensional cross-sectional representation was adopted to preserve the dominant thermal resistance path between the exposed and rear surfaces. The equivalence of this simplification is supported quantitatively by the strong agreement between simulated and experimental temperature histories. Nevertheless, the model does not explicitly capture localized three-dimensional phenomena such as edge heat losses, corner effects, or preferential leakage channels, which may contribute to the small deviations observed at prolonged exposure durations.

A structured quadrilateral mesh with local refinement around the PCM container was used to accurately resolve steep temperature gradients. A mesh independence study ensured that the difference in rear-face temperature between successive mesh refinements was less than 1 °C, corresponding to ~18,000 elements in the final model. Temporal discretization employed an implicit backward Euler scheme with adaptive time-stepping (Δt = 0.1–1 s), providing numerical stability at high temperature gradients. Mesh sensitivity analysis confirming the spatial convergence of the finite element model is presented in Figure 4, showing that temperature deviations between successive refinements remain below 1 °C throughout the simulation.

The model was validated against experimental data for all exposure regimes, showing close agreement in temperature evolution and insulation performance (see Section 3.4).

The following points represent model options for sensitivity (fire-focused):

• Mushy-zone width $\Delta T$: 1.0–3.0 K (affects the sharpness of melting plateau).
• Softening threshold $T_{\mathrm{soft}}$: ±10 K about the nominal value to bracket filament variability.
• Surface coefficients: $h$ and $\epsilon$ ±20% to bound furnace flow/emissivity uncertainty.
• Concrete $k\left(T\right),c_p\left(T\right)$: include/exclude moisture-peak in $c_p$ to assess its impact on rear-face rise at 200–400 °C plateaus.

2.4. Experimental Programme and Validation Plan

The experimental programme and numerical simulations were designed in parallel to enable direct cross-validation of temperature fields, thermal gradients, and overall insulation performance. The PCM-integrated concrete panels prepared as described in Section 2.1 were subjected to the fire exposure conditions detailed in Section 2.2. For each configuration, identical thermocouple positions were implemented in both the physical panels and the finite-element model to ensure point-to-point comparability. The validation strategy followed a three-stage approach:

Stage 1—Thermal response mapping: Transient temperature data from the embedded thermocouples (front, mid-plane, rear) were time-synchronized with corresponding nodal outputs from the FEM simulation. Furnace temperature profiles and boundary conditions in the model were assigned directly from recorded experimental data to eliminate discrepancies from idealized heating assumptions.

Stage 2—Calibration of material parameters: Key thermal properties—effective conductivity, specific heat, and latent-heat range of the PCM—were iteratively adjusted within ±5% of measured laboratory values until the simulated mid-plane temperature history matched the experiment within ±2 °C. No arbitrary scaling was introduced; all parameters remained within experimentally verified limits (

Table 2. Summary of thermal insulation and structural performance of PCM–ABS composite panels under 60 min furnace exposures.

Exposure (°C)	tΔ140 (min)	tΔ180 (min)	Rear-Face Temp. at 60 min (°C)	PCM State	Observed ABS Integrity
200	>60	>60	87 ± 6	Melted, retained	Intact
400	48 ± 3	>60	165 ± 10	Melted/liquid	Slight softening
600	31 ± 2	43 ± 3	262 ± 15	Liquid	Partial leakage
800	21 ± 3	36 ± 2	352 ± 18	Liquid/decomposed	Cracked, charred
ISO 834	29 ± 2	40 ± 3	305 ± 15	Liquid	Moderate degradation

Note: In Table 2, the parameters ${\mathrm{t}}_{140}$ and ${\mathrm{t}}_{180}$ represent the time required for the rear-face thermocouple(s) to reach absolute temperatures of 140 °C and 180 °C, respectively. However, per EN 1363-1 [24], insulation failure is defined by a rear-face temperature rise (ΔT) of 140 K average and 180 K maximum, relative to the initial temperature.

).

Stage 3—Quantitative validation metrics: Agreement between experimental and numerical temperature histories was evaluated using the root-mean-square error (RMSE), mean bias error (MBE), and coefficient of determination (R²) (Equation (12)):

$$RMSE=\sqrt{{\displaystyle \frac{1}{n}}{\displaystyle {\sum }_{i=1}^n}(T_{num,i}-T_{exp,i}{)}^2},R^2=1-{\displaystyle \frac{\sum (T_{num,i}-T_{exp,i}{)}^2}{\sum (T_{exp,i}-{\overline{T}}_{exp}{)}^2}}$$

(12)

Consistency across all three thermocouple depths and time intervals was used as the criterion for model acceptance.

Each validated model case (ISO 834 and constant-temperature exposures at 200–800 °C) served as a benchmark for assessing PCM melting progression and encapsulation stability. The same experimental–numerical workflow was repeated for both organic and inorganic PCMs, ensuring transferability of the methodology. The validated datasets form the foundation for the comparative analysis discussed in Section 3.

3. Results and Discussion

3.1. Experimental Fire Performance of PCM–ABS Composite Assemblies

The fire resistance of PCM–ABS-integrated composites was evaluated under 60 min constant-temperature exposures (200–800 °C) and standard ISO 834 heating. Table 2 summarizes key outcomes, including rear-face temperatures, and t_Δ140 and t_Δ180 represent the time required for the rear-face temperature rise relative to the initial temperature to reach 140 K and 180 K, respectively, consistent with EN 1363-1 insulation criteria.

At 200 °C, the rear-face temperature remained well below both thresholds, never exceeding 90 °C during the 60 min test. This indicates t₁₄₀ and t₁₈₀ > 60 min, demonstrating exceptional insulation performance. The PCM likely melted early in the exposure, but remained fully contained and effective in buffering heat transfer. The composite retained its structural integrity with no deformation or leakage. Thermographic imagery of the rear surface (Figure 5) supports this, showing dominant cool zones (≤60 °C).

At 400 °C, rear-face temperatures reached 165 ± 10 °C, crossing the 140 °C threshold at 48 ± 3 min, but never reaching 180 °C within the test duration (t₁₈₀ > 60 min). The PCM was in a fully molten state throughout, absorbing substantial latent heat. Minor softening of the ABS matrix was observed visually, but the panel remained structurally stable. Thermographic data shows warm regions up to 65 °C, consistent with delayed but advancing heat penetration.

At 600 °C, both insulation thresholds were crossed within the hour: t₁₄₀ = 31 ± 2 min, t₁₈₀ = 43 ± 3 min, and final rear-face temperature reached 262 ± 15 °C. PCM leakage became evident, attributed to softening of ABS above 100 °C and onset of degradation around 370 °C. This is reflected in rear-face thermal images, with temperatures exceeding 70 °C and showing localized heating patterns indicating compromised thermal resistance.

At 800 °C, the performance further declined. Rear-face temperatures rose to 352 ± 18 °C, with t₁₄₀ = 21 ± 3 min and t₁₈₀ = 36 ± 2 min, indicating rapid heat breakthrough. The PCM underwent full melting and decomposition, while the ABS exhibited visible charring and structural cracks, confirming thermal limits were exceeded. Thermograms at this stage show the highest rear-face intensity, reflecting a breakdown in insulating capacity. In addition to rear-face temperature thresholds (t₁₄₀ and t₁₈₀), visual inspection revealed that at 600 °C and above, the ABS encapsulant exhibited charring and cracking—markers of mechanical degradation. These changes indicate early signs of integrity failure, which, combined with insulation loss, could compromise the EI classification of the wall system if unprotected. According to EN 13501-2 and EN 1363-1, the “EI” fire-resistance rating denotes performance in terms of both integrity (E) and insulation (I). For example, EI 60 indicates that the test panel prevents flame penetration (E) and limits heat transfer (I) such that the average rear-face temperature rise does not exceed 140 K and the maximum rise at any point does not exceed 180 K, for a duration of at least 60 min. No explosive spalling of concrete was observed, likely due to low moisture content and slow heating rates, but surface cracking propagated along thermal gradients, especially near PCM interfaces.

Separate panels were used for each constant-temperature exposure condition (200, 400, 600, and 800 °C); after thermal exposure, panels were allowed to cool to ambient conditions before post-exposure thermography/UPV/compressive tests. Under ISO 834 exposure, the gradual heating led to similar final rear-face temperatures as in the 800 °C case: 305 ± 15 °C. Threshold times were t₁₄₀ = 29 ± 2 min and t₁₈₀ = 40 ± 3 min, suggesting a comparable thermal response due to cumulative energy input, despite the slower initial ramp. Visually, moderate ABS degradation was evident, but no catastrophic failure occurred.

In conclusion, the results demonstrate clear thermal staging: effective insulation and PCM heat absorption up to 400 °C, followed by progressive ABS softening, PCM leakage, and decomposition at higher exposures. Rear-face thermography supports this interpretation, showing increasing surface temperatures and heat localization. These findings confirm the dual-stage fire resistance behavior of PCM–ABS systems, offering high insulation performance under moderate heat loads but reduced effectiveness under extreme thermal stress.

3.2. Thermal Simulation and Mechanistic Interpretation of PCM–ABS Composite Behavior Under Fire Exposure

To rationalize the experimental trends and quantify the coupled influence of PCM melting and ABS softening, a fully transient thermal model was developed incorporating phase-dependent thermophysical properties and latent heat effects. The analysis reproduced the furnace and ISO 834 conditions, validated against embedded thermocouple data (Figure 6).

As shown in Figure 6, the simulated mid-plane and rear-face temperature curves closely track the corrected experimental data within ±5%, confirming that the apparent-heat-capacity formulation adequately represents the PCM melting process. The early segment of the curves (0–10 min) shows a nearly isothermal plateau between 28 and 35 °C, corresponding to the PCM melting interval. During this phase, the majority of input heat flux is absorbed as latent heat rather than raising the local temperature, producing a visibly flattened slope in both the experimental and numerical traces.

The zoomed-in view (Figure 7) provides greater clarity: the rear-face temperature remains almost constant until the PCM transition is complete, after which the temperature begins to rise rapidly. The second inflection, centered near 95–110 °C, marks the onset of ABS softening. This transition corresponds to the reduction in the effective thermal conductivity of the polymer network and a simultaneous increase in volumetric heat capacity. Physically, this means that while the ABS begins to lose stiffness, it continues to retard heat transfer by acting as a semi-molten thermal barrier. It should be noted that Figure 6 reports rear-face surface temperatures, whereas Figure 7 illustrates internal temperature distributions at selected depths; therefore, absolute temperature values at a given time are not directly comparable.

The transient thermal contours (Figure 8a,b) illustrate the spatio-temporal evolution of heat within the 50 mm composite slab. At 10 min of ISO 834 exposure (Figure 8a), the 31 °C isotherm (white contour) has migrated to approximately 20 mm from the hot face, indicating progressive PCM melting through the front half of the section. The ABS softening front (cyan line, ~95 °C) remains confined within the first 10 mm, suggesting that the polymer structure adjacent to the furnace surface has softened but not yet degraded.

After 60 min (Figure 8b), a broad, graded temperature field develops. After 60 min (Figure 8b), the hot-face temperature approaches the experimentally measured exposure level (≈840 °C at the exposed surface for the ISO-834 test conducted in the present furnace configuration), while the rear-face temperature evolution remains consistent with Table 2. In Figure 8, the rear-face temperature remains below 60 °C during the initial stage (up to approximately 10 min) of ISO-834 exposure, after which it rises progressively, reaching about 305 ± 15 °C at 60 min, as reported in Table 2. This near-steady temperature gradient, sustained even at one hour, highlights the significant reduction in thermal diffusivity induced by the combined effects of PCM latent-heat buffering and softened ABS conduction resistance. The absence of sharp isothermal bands confirms that the composite undergoes a controlled, progressive energy redistribution rather than abrupt thermal runaway.

The interplay between PCM melting and ABS softening is quantitatively represented in Figure 9. The sharp spike in the PCM’s apparent heat capacity near 30 °C corresponds to an effective enthalpy of 195 ± 10 kJ kg⁻¹ (consistent with DSC results [23]). This transient thermal reservoir delays the temperature escalation until the PCM is fully liquefied. Beyond this, the ABS transition curve (red dashed line) rises steeply around 95 °C, producing a secondary thermal damping stage. The combined dual-stage response explains the extended thermal inertia of the system observed experimentally—manifested as delayed attainment of both t₁₄₀ and t₁₈₀ criteria. Because RT35 melts at ~31 °C (28–35 °C range), its latent-heat buffering occurs only during the initial stage of heating. Under severe fire exposure, the latent heat is exhausted early and the subsequent thermal response is dominated by the concrete matrix and encapsulation stability rather than continued PCM ‘efficiency’.

Mechanistically, once the PCM completes its phase transition, the delayed rear-face temperature rise observed after PCM melting may be associated with transient changes in the effective thermal transport properties of the softened ABS layer; however, the present continuum model does not explicitly resolve microscale flow or pore-scale transport mechanisms. This behaviour may contribute to the delayed temperature rise observed experimentally. The minor deviations observed between experimental and numerical temperature histories at prolonged exposure times are primarily attributed to microscale physical phenomena that are not explicitly represented in the present continuum heat-conduction model. These include the development of microcracks in the cementitious matrix due to thermal incompatibility, interfacial debonding between the PCM–ABS regions and surrounding concrete, and the formation of connected pore and crack networks. Such features locally enhance heat transfer by providing preferential pathways for convective and radiative transport in addition to conduction. Furthermore, softening and partial degradation of the ABS phase can create micro-voids and local discontinuities, which modify the effective thermal conductivity and volumetric heat capacity of the composite. These microscale effects collectively promote localized acceleration of heat penetration, particularly at high exposure temperatures, thereby contributing to the small experimental–numerical deviations observed at later stages of fire exposure.

The parametric simulations performed under constant furnace exposures (200–800 °C) and the ISO 834 heating regime demonstrate the progressive evolution of heat transfer mechanisms within the PCM–ABS composite as exposure severity increases. Under moderate furnace conditions (200–400 °C), the rear-face temperature rises slowly and stabilizes below approximately 120 °C after 60 min, indicating that the latent-heat absorption of the PCM remains the primary thermal buffering mechanism. In this range, the composite’s heat transfer is dominated by transient energy storage rather than steady conduction, with the effective thermal gradient maintained across most of the section.

As the furnace temperature increases, the characteristic inflection associated with the onset of ABS softening occurs earlier, and the post-inflection slope steepens markedly. This shift signifies the transition from latent-heat-dominated behavior to a conduction-controlled regime, as the polymer matrix softens and its thermal conductivity increases. The simulation successfully captures this phase-coupled transition: the PCM’s latent-heat reservoir becomes depleted while the softened ABS allows accelerated heat flow through the molten composite network.

At higher exposures (≥600 °C), the numerical predictions replicate the experimentally observed transition from controlled buffering to runaway heating. The latent-heat capacity is fully utilized, the ABS matrix loses stiffness, and conductive transfer through the liquefied phase dominates. Despite this, the simulation predicts extended thermal resistance, with the rear-face temperatures reaching the ΔT = 140 K and ΔT = 180 K criteria at approximately 21 min and 36 min, respectively, consistent with experimental findings and confirming the partial retention of the barrier effect even under severe thermal degradation.

The close correlation between the simulated and measured temperature–time profiles (Figure 10), with deviations limited to within ±8%, validates the coupled representation of PCM phase change and polymer softening in the model. Minor discrepancies at longer durations likely result from micro-scale effects—such as local delamination, micro-crack propagation, and convective transport within the softened layer—which were not incorporated in the present one-dimensional heat conduction framework. Overall, the simulation accurately reproduces the composite’s two-stage thermal response and provides quantitative insight into the interplay between phase transition, polymer degradation, and residual heat insulation during high-temperature exposure.

Together, these simulations elucidate the sequential mechanisms underpinning fire resistance in PCM–ABS composites:

• Latent-heat buffering (0–10 min): PCM melting absorbs incident heat, maintaining quasi-isothermal conditions.
• ABS softening retardation (10–25 min): Softened ABS restricts heat flow while maintaining surface integrity.
• Thermal stabilization (25–45 min): Liquid PCM and softened ABS act synergistically to suppress rear-face temperature rise.
• Conduction-dominated regime (>45 min): Phase transitions completed; heat flow approaches steady-state conduction.

This multi-stage mechanism demonstrates that the inclusion of PCM and thermoplastic binder not only delays heat transfer but also introduces a self-regulating thermal resistance that evolves with temperature. Such behavior underpins the superior insulation performance observed experimentally, extending fire endurance without catastrophic structural failure. The close agreement between measured and simulated temperature histories provides an empirical bound on the modelling error associated with the equivalent 2D geometric idealisation for the present panel and boundary conditions.

3.3. Experimental Validation and Model Accuracy

Quantitative validation of the developed PCM–ABS thermal model was performed by comparing simulated and experimental rear-face temperature data under all furnace exposures and the ISO 834 standard heating regime. The results, summarized in Figure 11a–c, confirm the model’s high fidelity and its ability to accurately reproduce the complex, coupled behavior of phase change and polymer softening within the composite.

As shown in Figure 11a, the simulated and measured temperatures exhibit an almost perfect 1:1 correlation (R² = 0.996, RMSE = 3.5 °C). The close alignment across all heating conditions demonstrates that the model successfully captures both the rate and magnitude of transient heat transfer. The minimal spread along the 1:1 line suggests that the implemented temperature-dependent properties—particularly the PCM’s apparent heat capacity and the ABS softening function—provide a realistic representation of thermal behavior over the full exposure range.

The residual deviation map (Figure 11b) provides a spatial–temporal visualization of the model accuracy. For low and moderate furnace exposures (200–400 °C), deviations remain within ±2 °C throughout the 60 min exposure, confirming that latent-heat buffering and delayed conduction are precisely resolved. At higher temperatures (≥600 °C), residuals rise modestly (<5°C) toward the end of the exposure, reflecting minor underpredictions caused by dehydration of the concrete matrix, localized gas evolution, and crack-assisted conduction, phenomena not explicitly included in the simplified conduction framework. Despite these localized discrepancies, the overall deviation pattern remains symmetrical about zero, indicating that errors are random rather than systematic.

The statistical summary in Figure 11c consolidates these findings. The near-zero mean deviation (μ ≈ 0–1.5 °C) and small standard deviation (σ ≈ 1–2 °C) across all exposure levels confirm that the model consistently predicts rear-face temperatures within the intrinsic uncertainty range of the thermocouple measurement system (±2–3 °C). The widening distribution tails at higher temperatures signify physical variability rather than model bias, likely associated with heterogeneous thermal degradation and the onset of convective microchannels within the softened ABS phase.

Taken together, the quantitative metrics and residual analyses confirm that the developed phase-coupled model exhibits excellent predictive accuracy and stability. The low RMSE and high R² values verify that the model resolves both transient and steady thermal stages with negligible drift. Moreover, the statistical distributions demonstrate that predictive uncertainty remains bounded even under severe exposures, validating the robustness of the property formulations and boundary condition assumptions.

Given this level of agreement, the model is considered sufficiently accurate for façade-scale fire performance simulations, where predictive errors of ±5 °C are well within acceptable design tolerances. The established methodology therefore provides a reliable basis for scaling laboratory findings to structural-scale thermal analyses and for optimizing PCM–ABS composite configurations in future fire-resilient envelope systems.

3.4. Comparative Modelling of Inorganic PCM Fire Performance

To assess the potential of inorganic phase change materials for enhancing fire resilience, a comparative numerical simulation was conducted using representative thermophysical properties of hydrated salt-based PCMs, such as calcium chloride hexahydrate (CaCl₂·6H₂O). This class of PCM is non-combustible, with a similar melting point (~30–35 °C) and latent heat (~180–200 kJ/kg) to the organic paraffin RT35 used earlier, but with higher density (~1450 kg/m³) and thermal conductivity (~0.55 W/m·K). The same FEM domain, boundary conditions, and mesh were used, with the only change being the replacement of the organic PCM–ABS composite layer with a non-encapsulated inorganic PCM domain. The inorganic PCM analysis presented in this section is based on a validated numerical framework and is intended as a comparative and predictive assessment rather than an experimentally verified system. Experimental investigation of hydrated salt PCMs under fire exposure involves additional complexities related to dehydration reactions, moisture migration, phase segregation, and rehydration, which were beyond the scope of the present experimental programme. Consequently, the inorganic PCM results should be interpreted as indicative performance trends derived from thermophysical property-based modelling.

In the present study, three thermocouples were embedded at 5 mm from the rear face—near the center and symmetrically offset toward two corners—and were used to record rear-face temperature histories. While Table 2 reports absolute thresholds, the insulation performance discussed in the text is consistent with the EN 1363-1 ΔT definition, given that the initial temperature was close to 20–25 °C in all tests. For clarity, we acknowledge that future work should report ΔT explicitly, using average and maximum values across the rear-face sensor array in accordance with EN 1363-1.

Figure 12 compares the rear-face temperature histories for the organic and inorganic PCM panels under ISO 834 exposure. While both systems exhibit thermal buffering during the initial heating period due to latent heat absorption, the inorganic PCM panel maintains a significantly lower rear-face temperature throughout the 60 min test. The t₁₈₀ threshold is never reached in the inorganic case, in contrast to the organic PCM–ABS system which crosses this limit after ~40 min. This improved performance is attributed to the non-flammable nature of the salt hydrate, its higher thermal stability, and the absence of polymer encapsulants that soften or degrade under elevated temperatures.

These results validate the hypothesis introduced in Section 1: inorganic PCMs not only avoid adding fuel load but also maintain insulation performance at higher fire exposures, making them more suitable for façade applications targeting Euroclass C or higher. While practical implementation may require stabilization strategies to prevent supercooling and segregation, the fire modelling evidence clearly favors inorganic PCM systems from a thermal safety perspective.

The inorganic PCM results presented in this section are based on an idealized unencapsulated configuration and do not incorporate complex thermochemical effects such as dehydration, rehydration, phase segregation, supercooling, or mass transfer. These phenomena are especially relevant for hydrated salt PCMs and can significantly influence fire-side thermal performance. Therefore, the present comparison is limited to an idealized, encapsulation-free model with constant thermophysical properties. While this approach enables a first-order thermal benchmarking between organic and inorganic systems, it does not capture the detailed physical–chemical behavior of inorganic PCMs under elevated temperatures.

The inorganic PCM simulations presented herein should be interpreted as first-order comparative thermal analyses rather than predictive representations of fully coupled physicochemical behaviour under fire exposure. Important phenomena including dehydration kinetics, vapour transport, phase segregation, volumetric instability, and encapsulation interactions were not explicitly modelled. Consequently, the results primarily indicate comparative thermal trends and should not be interpreted as experimentally validated fire-resistance performance.

To enable a truly comparative evaluation, future modelling will incorporate encapsulation strategies for inorganic PCMs and include reactive transport modelling of dehydration processes, water migration, and latent enthalpy shifts. These enhancements are necessary for predictive simulations that can inform fire-safe PCM design at material and assembly scales.

3.5. Post-Fire Physical and Mechanical Assessment

The residual mechanical integrity of the PCM–ABS composite after fire exposure was evaluated through UPV measurements and compressive strength tests. The results, summarized in Figure 13, reveal a systematic degradation trend with increasing exposure temperature, reflecting the progressive deterioration of both the cementitious and polymeric phases.

At room temperature, the panels exhibited an average UPV of 4.5 km/s, indicative of dense, well-bonded microstructure and intact ABS–matrix interfaces. After exposure to 200 °C, only a marginal 5% reduction was observed, attributable mainly to transient dehydration of free water and minor matrix shrinkage. At 400 °C, UPV retention declined to approximately 84%, signaling the onset of microcracking and localized softening of the ABS inclusions. These changes coincide with the simulated temperature field, which predicted mid-thickness temperatures of 115–130 °C, sufficient to induce polymer relaxation and partial interfacial debonding while leaving the cementitious skeleton largely intact.

A pronounced degradation occurred at higher exposures (600–800 °C), where UPV values dropped to 3.1 and 2.4 km/s, respectively—representing a 47% reduction relative to unheated control panels. These reductions correlate with the experimentally and numerically observed collapse of latent-heat buffering and the complete softening and carbonization of the ABS matrix. Microstructural examination of the exposed surfaces confirmed the development of interconnected cracks aligned with the thermal gradient, providing preferential acoustic scattering paths that explain the steep loss in pulse velocity. The simulation’s steep isotherm clustering near the rear-face region at these temperatures further supports the conclusion that damage propagation originates from differential contraction between the outer degraded polymer phase and the cooler, more rigid interior.

Compressive strength measurements followed a similar trend, decreasing by approximately 12% after 400 °C and by up to 32% after 800 °C exposure. The reduction is consistent with the combined effects of binder microcracking, thermal incompatibility between the PCM–ABS domains and the cementitious matrix, and the partial loss of polymer stiffness. Despite these losses, the panels retained sufficient cohesion to prevent spalling or structural disintegration, underscoring the composite’s resilience even under severe fire loading.

The correlation between UPV degradation and simulated temperature gradients provides a quantitative link between thermal field localization and damage accumulation. The steep gradient near the mid-thickness, as predicted in the numerical model, corresponds to the zone of maximum acoustic attenuation, confirming that heat-induced phase transitions drive both thermal and mechanical degradation mechanisms. From a structural perspective, the results indicate that while the PCM–ABS composite undergoes significant stiffness loss at high temperatures, its integrity under moderate exposure (≤400 °C) remains largely preserved, qualifying it for Class EI fire performance in façade-type applications where insulation and integrity criteria dominate over load-bearing requirements. However, it should be noted that the observed polymer degradation and crack networks at high exposures suggest that while insulation criteria (t₁₈₀) were marginally met or exceeded, integrity criteria per EN 1363 (i.e., absence of through-cracks or sustained flaming) could be challenged without design modifications. These observations reinforce the importance of combining PCM integration with protective layers or high-T_g encapsulants to maintain both insulation and structural containment under severe fire conditions.

Building on the validated experimental–numerical correlation, the broader implications of these results for material design and fire safety optimization are evident. The results confirm that latent heat absorption dominates heat moderation below 400 °C, effectively delaying rear-face temperature rise and maintaining insulation performance. Beyond 500 °C, ABS softening and volatilization accelerate heat transfer, marking a transition from controlled buffering to conductive dominance.

For improved thermal resilience, increasing cover thickness or adopting high-T_g or inorganic encapsulants can delay softening and sustain barrier performance under severe exposure. Simulations indicate that even modest increases in protective cover (≈10 mm) can extend endurance by 15–20%, enabling compliance with EI fire classification under ISO 834 conditions.

These findings provide actionable guidance for façade fire design and material optimization, emphasizing the integration of PCM-based layers as thermal regulators in non-load-bearing assemblies, where extended insulation and integrity—rather than structural load capacity—govern fire resistance.

4. Conclusions

This study experimentally and numerically evaluated the fire performance of concrete panels integrated with organic and inorganic PCMs. Key findings derived from ISO 834 and constant-temperature furnace testing, as well as validated numerical simulations, are summarized below:

Thermal buffering due to PCM melting delayed rear-face temperature rise under moderate fire intensities (≤400 °C), extending the time to insulation failure.

At higher temperatures (≥500 °C), softening and collapse of ABS encapsulation reduced the effectiveness of PCM, leading to accelerated heat transfer and earlier failure.

The validated finite element model incorporating apparent heat capacity and temperature-dependent encapsulation properties accurately predicted temperature evolution and failure times across configurations.

Based on the experimentally measured rear-face temperature rise relative to the initial condition, the PCM–concrete panels achieved approximately EI 30 insulation performance under the tested ISO 834 exposure conditions.

For the investigated assembly configuration, thermal insulation performance was strongly influenced by concrete cover thickness and PCM placement. A minimum 25 mm cover delayed critical heat flux penetration.

Simulations suggest that inorganic PCMs may provide improved thermal stability under the investigated exposure conditions; however, dedicated experimental validation remains necessary. However, these results are idealized and do not yet account for chemical degradation or moisture transport.

These findings indicate that PCM integration can improve the thermal insulation performance of the investigated PCM–concrete assembly under the exposure conditions considered in this study. The identified thermal failure mechanisms—especially encapsulation softening—provide guidance for material selection and fire-safe design of energy-responsive building systems.

The conclusions of this study apply specifically to the investigated PCM–ABS–concrete assembly, material properties, and furnace exposure conditions. While the validated numerical framework provides useful insight into the thermal behaviour of PCM-integrated construction systems, broader assessment of façade-scale fire performance will require additional studies encompassing alternative PCM types, encapsulation strategies, assembly configurations, and fire scenarios.

Future work will focus on experimental fire testing of inorganic PCM-integrated concrete assemblies, including explicit consideration of dehydration reactions, mass transfer, and phase stability, to provide full experimental validation of the superior fire resistance predicted by the present numerical framework. Inorganic PCMs are predicted to offer superior fire resilience based on validated numerical simulations, highlighting their strong potential for façade applications pending dedicated experimental verification.