Effects of Elevated Temperatures and Cooling Regimes on the Mechanical Properties and Toughness of Glass Fiber-Reinforced Geopolymer Concrete

Abstract

In this study, an eco-friendly geopolymer concrete (GPC) was synthesized using fly ash, slag, and rice husk ash as precursors, and glass fibers were incorporated to enhance its mechanical properties. And then this study investigates the residual mechanical properties and microstructure evolution of glass fiber-reinforced geopolymer concrete (GFGPC) following elevated temperature exposure and subsequent cooling. Specimens incorporating varying glass fiber volume fractions (0–2.5%) were subjected to temperatures ranging from 25 °C to 800 °C, followed by either natural cooling or water-spraying cooling. The uniaxial compressive strength, Brazilian splitting tensile strength, and three-point flexural strength of the glass fiber-reinforced GPC were experimentally determined. Furthermore, fracture performance indicators—including the energy absorption capacity at failure, characteristic length, and double-K fracture parameters—were systematically analyzed. Results indicate that a glass fiber content of 1.5% optimally enhances the composite’s mechanical performance. Under natural cooling, splitting tensile and flexural strengths exhibit a non-monotonic trend, peaking at 200 °C. Conversely, water-spraying cooling induced thermal shock generally degrades tensile and flexural properties. However, at extreme temperatures (600 °C and 800 °C), water-spray cooling facilitates matrix densification and secondary geopolymerization, resulting in a residual compressive strength increase of 12.16% and 20.77% compared to natural cooling. Furthermore, based on composite damage theory, a binary nonlinear prediction model was developed to accurately capture the coupled effects of temperature and fiber characteristics on the residual compressive strength (R² > 0.90). Coupled with scanning electron microscopy (SEM) observations, the profound effects of elevated temperatures and thermal shock on the GPC gel matrix were elucidated, and the microscopic mechanisms underlying the failure of the fiber-bridging effect at high temperatures were thoroughly investigated. The findings of this study provide a solid theoretical foundation and scientific reference for the performance assessment and repair decision-making of GPC structures post-fire exposure.

1. Introduction

As the concept of sustainable development gains global traction, utilizing appropriate technological approaches to recycle industrial and agricultural solid waste has emerged as a primary research focus in the engineering sector, aiming to conserve natural resources and mitigate environmental pollution [1,2,3,4]. Compared to ordinary Portland cement (OPC) concrete, Geopolymer concrete has increasingly attracted significant research interest, owing to its cement-free preparation process while still maintaining high mechanical strength, favorable durability, and excellent potential for high-temperature resistance [5,6,7].

Currently, the mainstream precursor choices for GPC include fly ash [8], blast-furnace slag [9], and metakaolin [10]. Numerous scholars have conducted meticulous and outstanding research based on these materials. For instance, Rickard et al. [11] selected fly ash produced by three Australian power stations and synthesized geopolymer concrete using each type. Their findings indicated that the 28-day compressive strength of geopolymers largely depends on the fraction of fly ash particles smaller than 20 μm. Furthermore, the strength loss following elevated temperature exposure was primarily governed by the iron content in the fly ash and the Si:Al ratio of the geopolymer mixture. Hager et al. [12] found that partially replacing fly ash with slag could enhance the early mechanical strength of GPC under ambient curing conditions. They noted that with the addition of 50% ground granulated blast-furnace slag, the 90-day compressive strength reached 105.0 MPa, whereas the mortar without fly ash exhibited a 90-day compressive strength of 40.8 MPa. Concurrently, several researchers have actively explored the utilization of other waste materials as full or partial precursors [13,14,15]. Zhang et al. [16] investigated the variations in the mechanical properties and microstructure of GPC incorporating waste ceramic powder at elevated temperatures. Their study demonstrated that the compressive strength of GPC specimens increased with higher contents of waste ceramic powder, with all samples achieving their maximum compressive strength when exposed to 300 °C.

In practical engineering applications, extreme high-temperature events such as fires can induce the evaporation of free water, dehydration and shrinkage of the gel matrix, and a thermal expansion mismatch between the matrix and aggregates, leading to severe deterioration of mechanical properties [17]. In addition, the post-disaster cooling method (e.g., natural cooling or fire-hose water spraying) exerts a profound impact on the residual performance of the concrete [18]. In recent years, scholars worldwide have conducted extensive explorations into the variations in mechanical properties and microstructure of GPC under elevated temperatures and different cooling regimes [19,20,21]. For example, Valencia Saavedra et al. [22] pointed out that OPC concrete decomposes at 1100 °C. In contrast, alkali-activated concrete (with fly ash partially replacing cement, or fly ash and blast-furnace slag fully replacing cement) retained residual strengths of 5.5 MPa and 15 MPa, respectively. These results indicate that alkali-activated concrete exhibits superior performance at elevated temperatures. Another study [23] revealed that as the temperature increased, the number of pores with diameters between 10 and 100 nm in GPC specimens decreased, while the number of pores larger than 1000 nm increased. The crack width also expanded with rising annealing temperatures. Furthermore, materials based solely on fly ash (FA) exhibited smaller crack widths and lower crack frequencies compared to those incorporating GGBFS. Regarding the investigation of cooling methods, Huang et al. [24] confirmed that water cooling has a more pronounced effect on the flexural strength of ultra-high-performance concrete (UHPC) than on its compressive strength. For instance, at 1000 °C—the stage of most severe strength loss—the difference in flexural performance caused by the cooling method reached 63.6%, whereas the difference in compressive performance was 14.2%.

Furthermore, the GPC matrix inherently exhibits significant quasi-brittleness, characterized by relatively high compressive strength but comparatively low tensile and flexural strengths [25]. It is well known that incorporating appropriate fiber reinforcement is an effective strategy to enhance the toughness of concrete, retard crack propagation, and improve overall mechanical properties [26,27,28,29]. Tanyildizi et al. [30] investigated the high-temperature resistance of geopolymer concrete reinforced with polyvinyl alcohol (PVA) fibers, noting that the compressive strength of specimens cured at 60 °C and exposed to 400 °C increased by 29% relative to the control samples (CR). A similar proportional increase was observed in their flexural strength at 400 °C. Tu et al. [31] summarized the melting points of steel fibers, basalt fibers, and carbon fibers, which are 1370 °C, 1500–1700 °C, and above 3000 °C, respectively. When applied in fiber-reinforced geopolymers, these fibers can maintain their strength under elevated temperatures. Rodgers Bangi et al. [32] found that the interconnected channels formed in the matrix after the melting of PVA fibers helped reduce the probability of spalling in high-strength ordinary Portland cement concrete at high temperatures by alleviating internal vapor pressure.

However, most existing studies are limited to ambient temperatures or single heating conditions. There remains a lack of systematic and in-depth research on the evolution laws of static mechanical properties (uniaxial compression, splitting tensile, and three-point bending) and the microscopic damage mechanisms of GPC under the multi-factor coupled effects of elevated temperatures, fiber reinforcement, and varying cooling methods (especially severe thermal shock from rapid quenching). Against this background, GPC specimens with different glass fiber volume fractions were prepared in this study to systematically evaluate their residual mechanical responses and macroscopic evolution after exposure to elevated temperatures ranging from 25 °C to 800 °C, followed by either natural cooling or water-spraying cooling. Combining SEM microstructural characterization with macroscopic fracture mechanics parameters, this study comprehensively analyzes the superimposed effects of high-temperature thermal damage, thermal shock, and fiber bridging/pull-out mechanisms. Furthermore, a strength prediction model accounting for multi-factor coupling is established, aiming to provide a theoretical basis for the fire-resistant structural design and post-disaster safety assessment of this novel class of geopolymer high-performance concrete (GHPC) materials.

2. Materials and Methods

2.1. Raw Materials

In this study, three types of industrial and agricultural solid wastes were selected as precursors for the GPC specimens (Figure 1): Class I fly ash with a density of 2.54 g/cm³, a specific surface area of 0.409 m²/g and a 45 μm sieve residue of 13%; Grade S105 ground granulated blast-furnace slag (GGBS) with a density of 2.8 g/cm³, a specific surface area of 0.431 m²/g, a 45 μm sieve residue of 3.5% and a 28-day activity index of 98.4%; and rice husk ash (RHA) with a density of 2.06 g/cm³ and a specific surface area of 0.259 m²/g. The chemical compositions and loss on ignition (LOI) of the fly ash, GGBS, and RHA are detailed in

Table 1. Chemical content (%) and LOI of fly ash, ultra-fine BF slag and RHA.

Composition	SiO2	CaO	Al2O3	Fe2O3	MgO	SO3	LOI
Fly ash	60.90	4.90	24.47	6.70	0.68	1.50	2.00
Ultra-fine BF slag	34.50	34.00	17.70	1.03	6.01	1.64	0.84
RHA	89.16	1.45	0	0.125	0.70	0.38	1.04

.

River sand with a maximum particle size of 0.6 mm and a fineness modulus of 2.4 was utilized as the fine aggregate, while limestone gravel with a particle size ranging from 5 to 10 mm was employed as the coarse aggregate. The alkaline activator for the GPC was formulated by mixing a 12 mol/L sodium hydroxide (NaOH) solution with a sodium silicate solution having a modulus of 3.3. Glass fibers, manufactured by Shandong Senhong Engineering Materials Co., Ltd. (Zibo, China), were utilized for reinforcement. The visual morphology, mechanical properties, and geometric dimensions of these fibers are presented in Figure 1 and

Table 2. Mechanical properties and dimensions of glass fiber.

	Elastic Modulus (GPa)	Tensile Strength (MPa)	Equivalent Diameter (mm)	Length (mm)	Density (kg/m3)
Glass fiber	75	1824	12	12	2.68

, respectively.

Given that ground granulated blast-furnace slag exhibits higher reactivity at ambient temperatures compared to low-calcium fly ash, a higher proportion of GGBS facilitates the geopolymerization process, thereby contributing to early-age strength [33,34]. Simultaneously, the incorporation of fly ash refines the microstructure of the concrete, enhancing its compactness and fire resistance at elevated temperatures [35,36]. However, despite its superior reactivity, the cost of GGBS is notably higher than that of fly ash. Therefore, based on a comprehensive evaluation of reaction kinetics, workability, durability, and cost-effectiveness, corroborated by preliminary experimental verifications, a fly ash-to-GGBS mass ratio of 2:3 was adopted in this study. This ratio was selected to guarantee optimal performance while mitigating material costs. Furthermore, derived from preliminary trial mixings of geopolymer mortar and concrete, and aligning with the findings of Patel et al. [37], incorporating 5% RHA by mass of the binder can significantly enhance the mechanical properties of the concrete. Consequently, geopolymer high-performance concrete specimens with five different glass fiber volume fractions (0%, 1%, 1.5%, 2%, and 2.5%) were fabricated. The detailed compositions and mix proportions are summarized in

Table 3. Mixture proportions of GHPC (kg/m³).

Specimen	FA	Slag	RHA	NaOH	Na2SiO3	Water	Sand	Aggregate	Glass Fiber Volume Fraction (%)
P	152	228	20	30	105	105	671	1033	0
G1.0	152	228	20	30	105	105	671	1033	1.0
G1.5	152	228	20	30	105	105	671	1033	1.5
G2.0	152	228	20	30	105	105	671	1033	2.0
G2.5	152	228	20	30	105	105	671	1033	2.5

.

2.2. Sample Preparation

To ensure optimal workability of the concrete and a homogeneous dispersion of the fibers, the GHPC preparation procedure was strictly executed as illustrated in Figure 2. Fibers were incorporated during the dry mixing process, utilizing inter-aggregate friction to facilitate uniform dispersion and minimize agglomeration. Given that mechanical vibration predominantly applies horizontal forces and GHPC exhibits lower workability than conventional concrete, the resulting fiber distribution is assumed to be statistically random, with a negligible risk of fiber settlement within the casted cubes [38]. In accordance with ASTM C192 [39], following the casting process, the vibration-compacted GHPC specimens were cured under ambient conditions for 24 h prior to demolding. Subsequently, the demolded specimens were transferred to a curing chamber maintained at a temperature of 20 ± 5 °C and a relative humidity exceeding 90% for a 28-day curing period. Upon completion of the curing process, the initial concrete specimens were cut into cylinders (50 mm in diameter and 100 mm in height) and discs (50 mm in diameter and 25 mm in thickness). These were utilized for uniaxial compressive tests and Brazilian splitting tests, respectively. Meanwhile, the prismatic specimens, cast in 40 × 40 × 160 mm molds, were kept intact and employed for three-point bending tests. Ultimately, a total of 225 specimens were prepared for this experimental program, comprising 75 cylinders, 75 discs, and 75 prisms.

The specimens were exposed to target peak temperatures of 25 °C, 200 °C, 400 °C, 600 °C, and 800 °C at an average heating rate of 4 °C/min. The temperature range of 25–800 °C was selected because it comprehensively covers the complete evolutionary trajectory of the material, spanning from the ambient baseline state, through structural dehydration and phase transitions, to irreversible degradation under extreme elevated temperatures. This specific heating rate was selected to mitigate the thermal gradient between the surface and the core of the concrete, preserve the stability of the matrix, facilitate the gradual release of internal moisture, and consequently reduce the probability of thermal spalling. The time–temperature heating curves of the muffle furnace are depicted in Figure 3.

To ensure a uniform internal temperature field within the specimens, a constant temperature was maintained for 2 h upon reaching the target peak temperature. Following this thermal equilibration process, the specimens were cooled to ambient temperature using two distinct cooling regimes: natural cooling and water-spraying cooling (Figure 4). Natural cooling: The specimens were left inside the muffle furnace chamber and allowed to cool slowly in situ by opening the furnace door. This regime was designed to simulate the natural temperature decay profile of concrete structures in a post-fire ambient environment. Water-spraying cooling: The specimens were placed in a water collection tray and subjected to a continuous water spray from an overhead distribution apparatus, which was equipped with uniformly distributed orifices and connected to a constant water supply. This method aimed to simulate the transient thermal shock effect induced by fire-hose spraying during active firefighting operations.

The water-spraying process was sustained for 2 h. Upon completion of the cooling phase, the specimens were left to air-dry under ambient indoor conditions for 12 h. Finally, uniaxial compressive strength tests and microstructural characterization analyses were conducted on the treated specimens.

2.3. Testing Method

2.3.1. Uniaxial Compression Test

In accordance with ASTM C39 [40], uniaxial compressive tests were conducted using a Hualong WAW-300 microcomputer-controlled electro-hydraulic servo testing machine with a capacity of 300 kN, (Shanghai Hualong Testing Instrument Co., Ltd., Shanghai, China) as schematically illustrated in Figure 5a. To significantly mitigate the end effects induced by frictional constraints between the loading platens and the specimen ends, a silicone oil lubricant was applied to the contact interfaces. Cylindrical specimens with a height-to-diameter ratio of 2.0 were utilized to ensure that the central region of the specimen remained in a quasi-uniaxial compressive stress state, thereby enhancing the precision of the mechanical response measurements [41]. The loading process was executed under fully automatic closed-loop control using a displacement-controlled mode at a constant rate of 0.12 mm/min. The loading was terminated when the specimen exhibited macroscopic cracking or a significant drop in load. Each experimental condition comprised three replicate specimens, and the arithmetic mean of the test results was reported to ensure statistical reliability.

The uniaxial compressive strength of the specimens was calculated using the following Equation (1):

$${\sigma }_c={\displaystyle \frac{P}{A}}$$

(1)

where σ_c is the uniaxial compressive strength of GHPC (MPa); P is the peak load at failure (N); and A is the cross-sectional area of the specimen under compression (mm²).

2.3.2. Splitting Tensile Test

The Brazilian splitting test is an indirect testing method widely utilized to determine the tensile strength of brittle materials, including rocks and concrete. By applying diametrically opposed linear compressive loads to a standard cylindrical specimen, this test induces an approximately uniform tensile stress along the loading diameter within the specimen. Once this tensile stress reaches the ultimate tensile strength of the material, splitting failure occurs along the central loading axis.

The Brazilian disk splitting tests were conducted using the same testing apparatus (as illustrated in Figure 5b). A displacement-controlled mode was employed at a constant loading rate of 0.1 mm/min. The loading process was terminated upon the occurrence of a sudden through-splitting fracture or a sharp drop in load following the peak value. Furthermore, strain gauges were affixed to the center of the discs (as shown in Figure 6a) to accurately record the strain evolved during the splitting failure of the specimens.

In accordance with ASTM C496 [42], the indirect tensile strength of the material can be calculated using the following Equation (2):

$${\sigma }_t={\displaystyle \frac{2P}{\pi Dt}}$$

(2)

where σ_t is the splitting tensile strength (MPa); D is the diameter of the specimen (mm); and t is the thickness of the specimen (mm).

2.3.3. Three-Point Bending Test

The three-point bending test is a classical mechanical testing method widely applied to determine the flexural mechanical properties of materials such as rocks, concrete, ceramics, and composites. In this test, a concentrated vertical load is applied at the mid-span of a simply supported beam specimen with a specific span length, inducing compressive stresses in the upper portion of the cross-section and tensile stresses in the lower portion. As the load gradually increases and the maximum tensile stress at the bottom edge reaches the ultimate flexural tensile strength of the material, cracks typically initiate on the tension surface at mid-span and rapidly propagate parallel to the loading direction until fracture failure occurs. Compared to direct tensile tests or four-point bending tests, the three-point bending test offers significant advantages, including simpler specimen preparation, reduced material consumption, less stringent alignment requirements for the loading apparatus, and overall operational convenience. The experiments were continued on the identical testing machine employing a displacement-controlled mode with a constant loading rate of 0.2 mm/min. The loading protocol was halted upon the initiation of a distinct major crack or a substantial decline in load following the peak. The loading fixtures were appropriately replaced (as shown in Figure 5c), and strain gauges were affixed to the midline of the bottom surface of the prismatic specimens (as shown in Figure 6b) to accurately measure the deflection during bending failure.

In accordance with ASTM C293 [43], and premised on the linear elastic failure of the specimen, the flexural strength of the material can be calculated using the following Equation (3):

$${\sigma }_f={\displaystyle \frac{3PL}{2b{h}^2}}$$

(3)

where σ_f is the flexural strength (MPa); L is the span length between the two lower supporting rollers (mm); b is the cross-sectional width of the specimen (mm); and h is the cross-sectional height of the specimen (mm).

3. Experimental Results

3.1. Mass Loss

Mass loss serves as a crucial indicator for characterizing the phase evolution of concrete following exposure to elevated temperatures [44]. The mass loss rate, denoted as I_m, is defined as the ratio of the mass lost during high-temperature exposure to the initial mass prior to heating. It reflects the physical and chemical alterations within the internal microstructure and is calculated using the following Equation (4):

$$I_m={\displaystyle \frac{m_0-m_t}{m_0}}\times 100\%$$

(4)

where m₀ is the initial mass of the specimen before heating, and m_t is the residual mass of the specimen after high-temperature exposure.

After heating the GHPC cylindrical, disc, and prismatic specimens with identical fiber volume fractions, their average mass loss rates were calculated, and the results are plotted in Figure 7. The experimental results demonstrate that at the same temperature, the mass loss rates (I_m) of the concrete across different fiber volume fraction are highly comparable, and all consistently increase with ascending temperature. This indicates that the fiber content exerts a minor influence on the thermally induced mass loss of GHPC, and the mass evolution is predominantly governed by the degradation of the concrete matrix itself.

At temperatures below 200 °C, the mass loss rates of the fiber-reinforced specimens were slightly higher than those of the plain concrete (averaging approximately 9.8% higher). This phenomenon is attributed to the relatively porous interfacial transition zone (ITZ) at the fiber–matrix interface, which induces early-stage microcracking and consequently forms interconnected pathways for vapor release. Conversely, at the extreme elevated temperature of 800 °C, the three-dimensional restraint effect provided by the dispersed fibers effectively mitigates the surface spalling of the matrix, thereby macroscopically reducing the overall mass loss (averaging approximately 6.4% lower).

3.2. Epigenetic Feature Analysis

Figure 8 illustrates the macroscopic morphological evolution of the specimens after experiencing different peak temperatures and cooling treatments (specimens of different dimensions exhibited similar apparent transformation patterns under elevated temperatures).

200 °C regime: At this temperature, no significant difference in visual appearance was observed between the high-temperature state and the post-cooling state. The matrix generally presented a bluish-gray color, and the surface structure maintained a high degree of integrity without any obvious macroscopic cracks. This indicates that the interfacial bonding between the glass fibers and the matrix is tight, with no evident crack development.

400 °C regime: Upon heating to 400 °C, the color of the heated specimens significantly lightened to a pale gray. After natural cooling, the hue of the specimens evolved into grayish-yellow, and a small number of fine microcracks emerged on the surface, though localized spalling had not yet occurred. Conversely, the specimens subjected to water-spraying cooling displayed a color similar to that of the natural cooling group but exhibited slight signs of flaking at the corners and edges. This situation suggests that internal cracks within the matrix gradually propagate due to thermal damage, although large macroscopic cracks have not yet formed. Furthermore, water-spraying cooling exacerbates the generation of these cracks.

600 °C regime: As the temperature increased to 600 °C, the color of the heated specimens faded further, and the density of microcracks continuously increased. The surfaces of the naturally cooled specimens appeared yellowish-white and developed distinct reticular cracks. In contrast, the thermal gradient stress induced by water-spraying cooling resulted in significant blocky spalling at the corners of the specimens. This apparent phenomenon demonstrates that the fiber–matrix interfacial bonding begins to deteriorate, with cracks developing and gradually interconnecting.

800 °C regime: When the peak temperature reached 800 °C, the heated specimens turned grayish-white. The naturally cooled specimens exhibited severe physical damage characteristics, including surface pulverization, extensive spalling at the corners, and the formation of penetrating cracks, leading to a profound deterioration in structural integrity. Although the water-spraying cooled specimens did not exhibit large-scale spalling macroscopically, after the subsequent drying process, the surface cracks severely intersected and coalesced to form a complex crack network, accompanied by conspicuous fiber exposure. This observation signifies that the internal cracks within the GFGPC matrix have fully developed—even coalescing into partial macroscopic cracks—resulting in the severe degradation of the fiber–matrix interfacial bonding.

3.3. Compression Performance Parameters

3.3.1. Uniaxial Compressive Strength

The average uniaxial compressive strength values of the specimens under various experimental conditions, obtained from the uniaxial compressive tests, are summarized in

Table 4. Uniaxial compressive strength values of test samples.

Sample	Maximum Loading Temperature for Testing fc/MPa
	25 °C	σ	200 °C	σ	400 °C	σ	600 °C	σ	800 °C	σ
NP	60.43	2.08	68.52	1.07	32.71	2.34	16.33	0.98	10.42	0.77
WP	60.43	2.08	58.49	2.90	28.07	1.74	19.16	1.40	13.21	0.87
NG1.0	65.45	2.83	74.39	1.94	36.04	2.39	19.98	1.76	11.79	0.98
WG1.0	65.43	2.83	60.75	2.87	29.64	1.76	22.77	1.65	14.25	1.07
NG1.5	70.72	1.76	80.23	1.83	40.35	2.01	22.12	1.37	14.57	0.99
WG1.5	70.72	1.76	65.19	2.86	33.12	1.76	25.16	1.23	18.33	1.67
NG2.0	68.61	3.83	75.93	3.87	36.53	2.34	20.85	1.97	12.98	1.03
WG2.0	68.61	3.83	61.40	2.76	30.33	2.73	23.55	1.76	15.70	0.83
NG2.5	63.29	4.83	69.83	3.36	34.47	2.67	17.55	1.33	10.26	0.89
WG2.5	63.29	4.83	53.98	4.87	28.20	2.76	20.49	1.57	12.25	0.96

Note: The first letter of the specimen designation denotes the post-heating cooling regime (‘N’ for natural cooling and ‘W’ for water-spraying cooling). The second letter indicates the fiber type, ‘G’ for glass fiber. The trailing number represents the fiber volume fraction. f_c is the experimentally measured uniaxial compressive strength (MPa), and σ is the standard deviation of the strength values for the corresponding test group.

. Additionally, the corresponding load–displacement curves are plotted in Figure 9.

Figure 10 illustrates the evolutionary trend of the residual compressive strength of GHPC with respect to temperature under different cooling regimes. In conjunction with the analysis in Table 4, it is evident that under natural cooling conditions, the strength of the GHPC specimens exhibited a non-monotonic evolutionary characteristic of an initial increase followed by a subsequent decrease, reaching a peak residual strength at 200 °C. Taking the NG1.5 group—which demonstrated the optimal compressive performance in the figure—as an example, its strength at 200 °C increased by 13.44% compared to the ambient temperature baseline. However, as the temperature further escalated to 400 °C, 600 °C, and 800 °C, the strength reduction rates reached 42.94%, 68.73%, and 79.40%, respectively. In stark contrast, the specimen strength under the water-spraying cooling regime exhibited a monotonically decreasing trend with rising temperature. For the corresponding WG1.5 group, the strength reduction rates at 200 °C, 400 °C, 600 °C, and 800 °C were 7.82%, 53.16%, 64.43%, and 74.08%, respectively.

Furthermore, a comparative analysis of the compressive strength values at identical temperatures reveals the existence of a significant strength inversion temperature threshold in the experiment. In the 25–400 °C regime, the residual strength of the naturally cooled specimens was generally superior to that of the water-spraying cooled group. Conversely, in the 600–800 °C regime, the water-spraying cooled group demonstrated a higher strength retention rate. Using the 1.5% volume fraction glass fiber-reinforced group as an example, at 200 °C and 400 °C, the strength of the water-spraying cooled group was only 81.26% and 82.09% of that of the naturally cooled group. Conversely, at 600 °C and 800 °C, the strength of the water-sprayed specimens surpassed that of their naturally cooled counterparts by 13.76% and 25.82%, respectively.

The underlying mechanisms for these phenomena can be analyzed as follows: At relatively low elevated temperatures (e.g., 200 °C), this moderate thermal condition provided activation energy for the abundant unreacted fly ash or GGBS particles from the ambient curing stage, prompting them to undergo secondary geopolymerization with the residual alkaline activator within the GHPC matrix. Simultaneously, the slow temperature decay of the natural cooling process allowed the internal stresses—induced by moisture evaporation and slight thermal expansion—to be fully released. The matrix densification gain brought by this thermal excitation outweighed the micro-pore defects caused by moisture loss; consequently, the residual compressive strength after natural cooling paradoxically exceeded that of the unheated room temperature (RT) specimens. Conversely, for the water-spraying cooled specimens, the sudden temperature drop from 200 °C to the ambient water temperature (approximately 20 °C) created a severe transient thermal gradient between the rapidly cooled surface layer and the high-temperature core. The resulting thermal stress was sufficient to exceed the tensile strength of the material in this temperature regime, tearing microcracks in the inner surface layer and the transition zone. This severe thermal shock effect overshadowed the strengthening effect of the secondary geopolymerization on the matrix, causing the strength to exhibit continuous deterioration from the low-temperature stage onwards.

However, under extreme high-temperature conditions (e.g., 600 °C and 800 °C), the naturally cooled specimens remained in the high-temperature zone for an extended period, subjecting the matrix to more prolonged dehydration shrinkage and thermal stress redistribution. Particularly around 573 °C, the phase transformation of quartz aggregates occurred, and the long-term accumulation of thermal mismatch strains led to irreversible and severe degradation of the microstructure in the interfacial transition zone (ITZ). Water-spraying cooling, by arresting the continuous accumulation of thermal damage, paradoxically preserved relatively superior structural integrity compared to slow natural cooling. Concurrently, at these extreme temperatures, a fraction of unreacted precursor particles (such as fly ash) still persisted. The influx of external moisture induced by water-spraying cooling facilitated the production of a minor amount of N-A-S-H gel, which filled the fine microcracks within the concrete. Nevertheless, this healing effect offered minimal mitigation against the massive penetrating cracks caused by the severe thermal damage at this stage. Therefore, the strength of the water-spraying cooled specimens only exhibited a modest inversion relative to the naturally cooled specimens, and compared to the previous temperature gradient, the magnitude of strength decline remained highly pronounced.

Generally, the coefficient of variation (COV)—calculated as the ratio of the standard deviation to the mean—can be employed to evaluate the dispersion of experimental data. As presented in Table 4, the COV values for compressive strength are all below 10%, indicating a low level of dispersion and confirming the reliability of the experimental results. Furthermore, it is interesting to note that for GFGPC specimens with an identical glass fiber volume fraction, the COV of compressive strength generally exhibits an increasing trend with rising temperature.

The residual compressive strength (f_cu) of fiber-reinforced concrete following high-temperature exposure and cooling is the combined outcome of temperature-induced damage, fiber reinforcement effects, and the specific cooling regime. Based on the mechanics of composite materials and damage evolution theory, this study proposes that the residual strength can be characterized by the superposition of a matrix–fiber coupled strength term (f_cu(T,λ_f)) and a cooling method modification factor (ξ_cool). Its generalized expression is formulated as follows:

$$f_{cu}=f_{cu}\left(T,{\lambda }_f\right){\xi }_{cool}$$

(5)

To quantitatively evaluate the contribution of fiber geometric characteristics to the reinforcement effect, the fiber characteristic value (λf) is introduced:

$${\lambda }_f=V_f{l}_f/d_f$$

(6)

where V_f represents the fiber volume fraction, l_f is the fiber length, and d_f is the fiber diameter.

By integrating the architectural logic of the response surface methodology (RSM) with the physical evolution laws of material performance degradation, the following binary nonlinear prediction model is established to describe the coupled effects of elevated temperatures and fiber reinforcement on the strength of GHPC:

$$f_{cu}\left(T,{\lambda }_f\right)=f_{cu,0}\times \mathsf{\Phi }\left(T\right)\times \mathsf{\Psi }\left({\lambda }_f\right)\times \mathsf{\Omega }\left(T,{\lambda }_f\right)$$

(7)

The physical connotations and mathematical constructs of each term in the model are detailed as follows: Temperature evolution term Φ(T): Fitted using a modified Gaussian function, this term characterizes the non-monotonic evolutionary behavior of the matrix strength as a function of temperature. Fiber contribution term Ψ(λ_f): Constructed as a quadratic polynomial, this term aims to reflect the nonlinear mapping relationship between the fiber dosage and the matrix strength [45]. It effectively captures the strength reduction effect caused by fiber agglomeration at high dosages. Coupling correction term Ω(T, λ_f): Utilized to represent the compensatory effect of fibers in arresting thermal crack propagation under high-temperature fields. This term reflects that a higher fiber content decreases the material’s sensitivity to high-temperature thermal damage; that is, the strength reduction rate decelerates with an increase in λ_f.

$$\mathsf{\Phi }\left(T\right)=a\times \exp \left[-{\displaystyle \frac{{\left(T-T_{opt}\right)}^2}{2b^2}}\right]+c$$

(8)

$$\mathsf{\Psi }\left({\lambda }_f\right)=1+d{\lambda }_f-e{\lambda }_f^2$$

(9)

$$\mathsf{\Omega }\left(T,{\lambda }_f\right)=1+\gamma {\lambda }_f{\displaystyle \frac{T-20}{T_{\max }}}$$

(10)

where T_opt is the temperature corresponding to the maximum compressive strength, which is 200 °C in this experiment; and a, b, c, d, e, γ are empirical fitting parameters.

The modification factor ξ_cool is employed to quantify the additional damage or gain induced by different cooling rates. For the natural cooling regime, a standard reference value of ξ_cool = 1.0 is defined. For the water-spraying cooling regime, considering the initial damage caused by the transient thermal shock, the value of this modification factor typically ranges from 0.75 to 0.85.

Step compensation mechanism: Given the rehydration effect induced by water cooling at temperatures above 600 °C, a step compensation term Δη based on the mass gain rate is introduced into the model:

$$\Delta \eta =k{\displaystyle \frac{m_{ab}}{m_{in}}}$$

(11)

where m_ab is the mass gain of the specimen after water spraying, and m_in is the initial mass of the specimen.

Synthesizing the aforementioned components, the comprehensive fitting equation for the high-temperature compressive strength of GHPC under varying cooling regimes is obtained as follows:

$$f_{cu}\left(T,{\lambda }_f\right)=f_{cu,0}\left[a{e}^{-{\displaystyle \frac{{\left(T-200\right)}^2}{2b^2}}}+c\right]\left(1+\alpha {\lambda }_f-\beta {\lambda }_f^2+\delta T{\lambda }_f\right){\xi }_{cool}$$

(12)

Based on the experimentally measured data, a nonlinear regression analysis was performed on the above formula. The results demonstrate that the correlation coefficient (R²) for the fitted model of this fiber-reinforced GHPC exceeds 0.90, indicating a high degree of statistical reliability.

By substituting the parameters from

Table 5. Fitting parameters and correlation coefficients of concrete strength.

	a	b	c	α	β	δ	m	k	R2
Glass fiber	0.207	1.00	0.403	0.126	2.12 × 10−03	1.27 × 10−04	0.812	23.452	0.911

into the proposed equations, the predicted compressive strength values of GHPC under various fiber characteristic values and elevated temperatures were calculated. Subsequently, by plotting the experimentally measured strength values on the x-axis against the corresponding model-predicted strength values on the y-axis, and incorporating additional experimental data compiled from previous studies by other researchers [9,46], the correlation plot depicted in Figure 11 was generated.

An analysis of the correlation plot between the predicted and experimental compressive strengths (Figure 11) reveals that the data points are uniformly distributed along the line of equality (y = x). Furthermore, the correlation coefficient reaches 0.9754, robustly demonstrating that the proposed model can accurately predict the residual mechanical properties of GHPC across various fiber characteristic values and temperature gradients.

3.3.2. Compressive Energy Absorption

The energy absorbed during compressive failure is typically characterized by the integral area under the load–displacement curve. This metric inherently reflects the deformation capacity and the level of energy dissipation of the material prior to ultimate failure. Its value can be calculated using the following integral expression:

$$W_c={\displaystyle {\int }_0^{{\delta }_c}P}d\delta$$

(13)

where W_c is the energy absorbed up to peak failure (J); P is the applied compressive load; and δ_c is the corresponding displacement at peak failure.

The specific calculated values of the compressive energy absorption for the specimens, along with the comparative analysis plot, are presented in

Table 6. Compressive failure energy of specimens.

Sample	Maximum Loading Temperature for Testing Wc/J
	25 °C	σ	200 °C	σ	400 °C	σ	600 °C	σ	800 °C	σ
NP	90.19	6.32	70.49	5.34	35.35	2.24	20.49	1.34	5.67	0.45
WP	90.19	6.32	56.41	4.56	31.92	2.84	16.50	1.24	4.52	0.37
NG1.0	102.55	6.37	89.42	5.37	43.85	2.77	24.20	1.64	6.20	0.33
WG1.0	102.55	6.37	60.01	3.45	33.53	2.34	20.99	1.76	4.88	0.43
NG1.5	126.61	5.28	95.90	5.43	53.78	3.74	26.51	1.88	6.76	0.55
WG1.5	126.61	5.28	85.25	4.88	46.55	3.46	22.46	2.04	6.33	0.44
NG2.0	98.71	6.83	93.49	6.33	38.54	3.34	25.41	1.45	6.68	0.53
WG2.0	98.71	6.83	39.77	3.41	34.41	2.05	20.51	1.65	4.73	0.35
NG2.5	95.54	9.32	79.79	5.76	39.91	2.67	19.22	1.45	5.72	0.34
WG2.5	95.54	9.32	56.43	4.84	30.56	2.44	17.55	1.27	4.55	0.21

and Figure 12, respectively.

Figure 12 reveals the evolutionary laws of the energy dissipation capacity of GHPC during compressive failure under different cooling regimes. The experimental results indicate that, unlike the non-monotonic evolutionary characteristic of compressive strength, the energy absorbed (W_c) by the specimens during the compressive failure process exhibits a monotonically decreasing trend with increasing exposure temperature, regardless of whether natural cooling or water-spraying cooling was applied. Taking the NG1.5 group as an example, its residual energy dissipation values after exposure to elevated temperatures of 200 °C, 400 °C, 600 °C, and 800 °C were only 75.75%, 42.47%, 20.94%, and 5.32% of the ambient temperature baseline, respectively. The corresponding water-spraying cooled group (WG2.0) exhibited a more pronounced attenuation, with corresponding proportions of 67.33%, 35.76%, 17.74%, and 5.00%, respectively.

Across the entire temperature range and all fiber volume fractions, the energy absorption level of the water-spraying cooled specimens remained consistently lower than that of the naturally cooled group. Particularly at 200 °C, the difference in energy dissipation between the two regimes was remarkably significant: the maximum discrepancy occurred in the G2.0 group, where the energy dissipation of the naturally cooled specimens was 61.83% higher than that of the water-spraying cooled counterparts. In conjunction with the load–displacement curves presented in Figure 9, it is evident that the severe thermal shock effect induced by water-spraying cooling not only reduced the peak failure load of the concrete but also advanced its displacement at failure, thereby resulting in a substantial reduction in overall energy dissipation.

3.4. Splitting Tensile Performance Parameters

The splitting tensile performance parameters of the various GHPC specimens, obtained through the Brazilian splitting tests, are statistically summarized in

Table 7. Splitting tensile performance parameters of the GHPC specimens.

Sample	Parameter	Maximum Loading Temperature and Standard Deviation
		25 °C	σ	200 °C	σ	400 °C	σ	600 °C	σ	800 °C	σ
NP	ft	4.44	0.34	4.96	0.43	1.78	0.12	1.15	0.09	0.68	0.04
	Gt	927.56	78.32	988.78	65.34	347.68	22.12	228.71	12.43	128.01	12.10
	lch	175.62	13.48	183.44	12.34	197.05	14.48	212.92	17.34	230.63	20.35
WP	ft	4.44	0.43	3.93	0.23	1.53	0.14	1.01	0.10	0.48	0.03
	Gt	927.56	78.32	781.97	54.34	266.62	21.34	178.77	14.34	121.87	10.23
	lch	175.62	13.48	172.11	10.56	195.35	14.77	211.07	18.45	218.59	20.40
NG1.0	ft	5.04	0.45	5.31	0.43	2.29	0.12	1.97	0.16	1.14	0.11
	Gt	1454.65	110.34	1171.38	81.23	404.68	23.34	384.01	12.34	212.67	20.34
	lch	220.09	17.88	210.01	19.45	250.88	23.45	283.88	25.89	291.53	27.99
WG1.0	ft	5.04	0.45	4.91	0.34	2.13	0.21	1.45	0.11	0.94	0.05
	Gt	1453.65	110.34	1015.07	99.45	373.569	21.34	338.50	23.56	173.63	19.45
	lch	220.09	17.88	201.85	18.56	241.25	20.89	254.39	22.44	259.20	24.78
NG1.5	ft	5.79	0.44	6.87	0.34	2.62	0.16	1.70	0.11	0.96	0.06
	Gt	1472.13	70.56	1261.61	68.54	499.85	20.56	294.72	10.87	264.16	12.46
	lch	230.50	12.34	214.26	18.43	247.44	16.34	256.58	15.34	304.53	18.45
WG1.5	ft	5.79	0.44	4.89	0.42	2.27	0.21	1.62	0.12	0.89	0.05
	Gt	1472.13	70.56	1115.87	50.46	477.84	21.55	279.43	18.48	213.63	14.45
	lch	230.50	12.34	208.90	10.45	210.55	15.34	245.53	12.45	246.53	12.45
NG2.0	ft	5.21	0.49	6.36	0.43	2.48	0.21	1.90	0.14	1.15	0.09
	Gt	1466.76	100.56	1206.44	86.45	452.59	42.45	359.77	29.56	236.02	19.56
	lch	221.51	17.34	213.69	19.88	246.94	20.54	271.02	20.98	284.44	22.87
WG2.0	ft	5.21	0.49	4.35	0.34	2.34	0.22	1.56	0.14	0.70	0.05
	Gt	1466.76	100.56	1112.91	101.65	396.22	31.56	294.87	27.56	180.81	14.66
	lch	221.51	17.34	204.56	18.34	244.73	19.36	263.62	19.56	273.68	24.66
NG2.5	ft	5.12	0.45	6.21	0.42	2.38	0.19	1.74	0.13	1.05	0.08
	Gt	1312.52	125.45	1162.74	98.43	463.16	24.54	383.79	34.35	216.47	20.13
	lch	217.07	18.45	209.24	19.45	254.80	15.34	266.58	20.77	308.43	25.78
WG2.5	ft	5.12	0.45	4.68	0.33	2.10	0.20	1.46	0.12	0.64	0.04
	Gt	1312.52	125.45	959.57	88.34	420.59	32.54	309.51	27.34	190.32	16.34
	lch	217.07	17.34	200.68	17.45	253.80	19.85	255.16	20.91	277.29	23.34

Note: f_t is the experimentally measured Brazilian splitting tensile strength (MPa); G_t is the splitting energy (J/m²); and l_ch is the characteristic length (mm).

. Furthermore, based on the experimental test results, the splitting load–displacement curves of the specimens under different peak heating temperatures and cooling regimes are plotted in Figure 13.

3.4.1. Splitting Tensile Strength

A comparative analysis of the data in Table 7 reveals that, across all investigated elevated temperatures and cooling regimes, the specimens incorporating a 1.5% glass fiber volume fraction consistently exhibited the optimal splitting tensile strength. Specifically, under natural cooling conditions at 25 °C, 200 °C, 400 °C, 600 °C, and 800 °C, the strengths of these optimized specimens were 30.39%, 38.52%, 46.66%, 47.18%, and 42.47% higher than those of the plain GHPC control group (without fibers), respectively. These findings compellingly demonstrate that the incorporation of glass fibers significantly enhances the tensile performance of the concrete.

Figure 14 illustrates the variation in the splitting tensile strength of the G1.5 group with respect to the heating temperature and cooling regime. The GHPC specimens with other fiber volume fractions exhibited similar evolutionary trends. Similar to the aforementioned variation in compressive strength, the strength of the naturally cooled specimens at 200 °C significantly increased by 18.74% compared to the ambient temperature (25 °C) baseline. This enhancement can likewise be attributed to the thermal excitation of the residual precursors and the subsequent occurrence of secondary geopolymerization.

However, as the temperature exceeded 200 °C, the strength of the specimens exhibited a decreasing trend regardless of whether natural cooling or water-spraying cooling was applied. At 400 °C, the strength of the naturally cooled specimens plummeted by 61.81% compared to that at 200 °C, while the water-spraying cooled specimens also experienced a 53.56% reduction. The strength of both groups continued to decline at 600 °C and 800 °C. Although the residual strength under natural cooling remained slightly higher than that under water cooling throughout the entire high-temperature regime, the strengths of both groups essentially converged at the extreme temperature of 800 °C, both plummeting below 1 MPa (with respective decline rates of 43.21% and 45.36%).

3.4.2. Splitting Energy and Characteristic Length

To comprehensively evaluate the toughness and deformation capacity of GHPC during the fracture process, the splitting fracture energy and characteristic length of the concrete were calculated.

In fracture mechanics, the splitting fracture energy generally refers to the total energy consumed to generate a unit area of crack during the splitting process. Its fundamental calculation formula is defined as follows:

$$G_t={\displaystyle \frac{W_t}{A_t}}$$

(14)

where W_t is the total energy dissipated during complete fracture (J), which can be determined by integrating the enclosed area under the load–displacement curve; and A_t is the cross-sectional area of the splitting crack (m²). In the Brazilian splitting test, the crack propagates through the entire height of the specimen upon failure; thus, A_t is calculated as the product of the specimen diameter and the disc thickness.

The characteristic length (l_ch) integrates the stiffness, fracture toughness, and tensile capacity of the material. Generally, a smaller characteristic length indicates a higher degree of material brittleness. The specific calculation formula is expressed as follows:

$$l_{ch}={\displaystyle \frac{E_t{G}_t}{f_t^2}}$$

(15)

where E_t is the elastic modulus (MPa); G_t is the splitting fracture energy (J/m²).

The splitting fracture energy and characteristic length values for each GHPC specimen are presented in Figure 15 and Figure 16, respectively.

As illustrated in Figure 15, with the increment of heating temperature, the splitting energy of both the plain geopolymer (NP group) and the fiber-reinforced modified groups (NG series) exhibited a significant and irreversible declining trend. Notably, when the temperature escalated from 200 °C to 400 °C, a precipitous drop in splitting energy was observed across all groups. The most substantial reduction occurred in the WG2.0 group, with a decrease magnitude of up to 64.40%. At extreme temperatures of 600 °C and 800 °C, the splitting energy diminished to low levels, and the rate of decline tended to stabilize. Furthermore, compared to the NP group at identical peak temperatures and cooling regimes, the fiber-reinforced groups maintained significantly higher splitting energy. This suggests that the fibers effectively performed bridge-cracking and energy-dissipating pull-out roles within the matrix. However, this toughening effect gradually attenuated at higher temperatures. For instance, the splitting energy of the NG1.5 group was 48.88% higher than that of the NP group at 25 °C, but this margin decreased to 28.87% at 600 °C. This degradation can be attributed to the glass fibers reaching their melting point at 600 °C, leading to the failure of the bridging mechanism.

Upon concrete cracking, the crack tip is not a discrete point but rather a region characterized by microcracking, aggregate interlock, and friction, known as the Fracture Process Zone (FPZ). The characteristic length is proportional to the physical dimensions of the FPZ; a larger l_ch implies a more extensive FPZ, indicating that the material can undergo more pronounced nonlinear deformation prior to fracture. This aligns with the load–displacement curves in Figure 13, where nonlinear deformation before GHPC fracture increased significantly with rising temperature.

As shown in Figure 16, the characteristic length of naturally cooled specimens exhibited a trend of an initial increase followed by a subsequent decrease with rising temperature. At 200 °C, the splitting strength of GHPC specimens improved while the splitting energy decreased. This is due to the thermal activation of the geopolymer, where unreacted precursors continued to undergo polycondensation, resulting in a denser and harder gel matrix. This densification enhanced the brittleness of the material, causing cracks to propagate more rapidly and concertedly under stress, thereby shrinking the FPZ and leading to a slight decrease in l_ch. When the temperature exceeded 400 °C, the geopolymer gel underwent severe dehydration, shrinkage, and even decomposition, while the thermal expansion mismatch between the aggregate and the paste generated substantial internal stress. This resulted in the initiation of an extensive and dense microcrack network. Under splitting stress, the original dense structure was compromised, and damage was no longer confined to a single sharp macroscopic primary crack but rather dispersed, propagated, and dissipated energy along the widespread microcrack network. This pervasively damaged internal state significantly expanded the FPZ, macroscopically manifesting as a marked increase in the characteristic length.

Compared to naturally cooled specimens, the l_ch of water-spraying cooled specimens increased continuously with temperature; however, it remained consistently lower than that of naturally cooled specimens at the same temperature and fiber content. This is because the water-quenching process induced severe thermal shock, causing rapid surface shrinkage and the instantaneous formation of localized macroscopic primary cracks or fatal defects. In subsequent mechanical testing, failure occurred extremely rapidly along these pre-existing weak planes. Due to the rapid development of the primary crack, the surrounding material had insufficient time to participate in the microcrack energy dissipation process, resulting in highly localized fracture, a smaller FPZ, and a consequently lower characteristic length.

3.5. Three-Point Bending Performance Parameters

The flexural performance parameters of the GHPC specimens obtained from the three-point bending tests are statistically summarized in

Table 8. Flexural parameters of GHPC specimens.

Sample	Parameter	Maximum Loading Temperature and Standard Deviation
		25 °C	σ	200 °C	σ	400 °C	σ	600 °C	σ	800 °C	σ
NP	ff	6.31	0.45	7.36	0.45	4.55	0.23	3.37	0.32	2.16	0.12
	KICQ	0.53	0.04	0.69	0.03	0.46	0.03	0.39	0.02	0.20	0.02
	KICS	5.00	0.32	5.47	0.47	3.90	0.28	2.82	0.19	1.41	0.24
WP	ff	6.31	0.45	3.94	0.34	2.96	0.23	2.49	0.19	2.06	0.17
	KICQ	0.53	0.04	0.34	0.02	0.37	0.02	0.29	0.02	0.17	0.01
	KICS	5.00	0.32	3.12	0.31	1.87	0.29	1.19	0.21	0.82	0.18
NG1.0	ff	12.24	0.89	14.39	0.78	8.92	0.56	6.32	0.45	3.67	0.24
	KICQ	0.92	0.05	1.03	0.03	0.78	0.06	0.63	0.04	0.31	0.03
	KICS	9.68	0.78	10.69	0.78	7.64	0.63	5.65	0.54	2.80	0.35
WG1.0	ff	12.24	0.89	8.77	0.76	6.47	0.54	4.15	0.23	3.57	0.21
	KICQ	0.92	0.05	0.59	0.04	0.50	0.03	0.34	0.02	0.21	0.02
	KICS	9.68	0.78	6.94	0.54	4.33	0.49	3.02	0.32	2.09	0.33
NG1.5	ff	12.46	0.74	16.03	1.02	9.99	0.56	7.39	0.34	4.07	0.23
	KICQ	1.04	0.07	1.28	0.07	0.76	0.06	0.66	0.04	0.29	0.02
	KICS	9.86	0.65	11.91	0.98	8.56	0.74	6.67	0.54	3.43	0.53
WG1.5	ff	12.46	0.74	10.49	0.89	8.42	0.66	4.87	0.34	3.52	0.22
	KICQ	1.04	0.07	0.79	0.07	0.65	0.06	0.37	0.02	0.24	0.01
	KICS	9.86	0.65	8.30	0.32	5.01	0.64	3.46	0.43	2.20	5.31
NG2.0	ff	10.64	0.74	14.21	0.88	8.37	0.65	5.60	0.43	3.35	0.21
	KICQ	0.80	0.03	1.02	0.08	0.74	0.06	0.57	0.04	0.29	0.02
	KICS	9.20	0.74	10.62	0.86	7.64	0.54	4.32	0.49	2.40	0.44
WG2.0	ff	10.64	0.74	8.54	0.34	6.62	0.54	4.65	0.32	2.80	0.12
	KICQ	0.80	0.03	0.59	0.03	0.45	0.04	0.33	0.03	0.21	0.02
	KICS	9.20	0.74	6.96	0.34	4.00	0.35	2.83	0.54	1.85	0.37
NG2.5	ff	10.64	0.74	14.21	0.88	8.37	0.65	5.60	0.43	3.35	0.21
	KICQ	0.80	0.03	1.02	0.08	0.74	0.06	0.57	0.04	0.29	0.02
	KICS	8.42	0.73	10.56	0.86	5.17	0.54	3.33	0.49	2.09	0.44
WG2.5	ff	10.64	0.74	8.54	0.34	6.62	0.54	4.65	0.32	2.80	0.12
	KICQ	0.80	0.03	0.58	0.03	0.50	0.04	0.32	0.03	0.20	0.02
	KICS	8.42	0.73	6.76	0.34	3.30	0.35	2.40	0.54	1.43	0.37

Note: f_f is the flexural strength (MPa); K_IC^Q is the initial fracture toughness of the specimen (MPa·mm^0.5); and K_IC^S is the unstable fracture toughness (MPa·mm^0.5).

. Furthermore, based on the experimental results, the flexural load–deflection curves of the specimens under different peak heating temperatures and cooling regimes are plotted in Figure 17. The x-axis in Figure 17 denotes the crack mouth opening displacement (CMOD).

3.5.1. Flexural Strength

Figure 18 illustrates the evolutionary trend of the residual flexural strength of GHPC under different cooling regimes. Consistent with the previously observed patterns for compressive and splitting tensile strengths, the flexural strength of the GHPC specimens under natural cooling also exhibited a non-monotonic trend—initially increasing and subsequently decreasing—reaching its peak residual strength at 200 °C. Taking the NG1.5 group, which demonstrated the superior performance in the figure, as an example: its strength at 200 °C increased by 28.66% compared to the ambient temperature baseline. However, as the temperature further escalated to 400 °C, 600 °C, and 800 °C, the residual strengths attenuated to 80.17%, 46.11%, and 25.39% of the initial values, respectively. In contrast, the strength of the specimens subjected to the water-spraying cooling regime exhibited a monotonic decrease with rising temperature. For the corresponding WG1.5 group, the strength reduction rates at 200 °C, 400 °C, 600 °C, and 800 °C were 15.83%, 32.25%, 60.92%, and 71.74%, respectively.

3.5.2. Fracture Toughness

The Double-K Fracture Model is commonly employed to evaluate the three-point bending fracture performance of quasi-brittle materials, such as geopolymer concrete. The two core parameters of this model are the initial fracture toughness (K_IC^Q) and the unstable fracture toughness (K_IC^S). Both parameters are derived from the stress intensity factor formulas in linear elastic fracture mechanics (LEFM), with the fundamental distinction lying in the specific load values and crack lengths incorporated into the calculations.

For a standard three-point bending beam (with a span-to-depth ratio of 4), the general formula for the stress intensity factor is expressed as:

$$K={\displaystyle \frac{PL}{t{h}^{3/2}}}f\left(\alpha \right)$$

(16)

where P is the applied load (N); L is the span between supports (mm); t is the thickness or width of the specimen (mm); h is the depth (height) of the specimen (mm); and α is the relative crack depth, defined as the ratio of crack length (l) to specimen depth (t). The geometric shape function, f(α), for a three-point bending beam with a span-to-depth ratio of 4, is calculated as follows:

$$f\left(\alpha \right)={\displaystyle \frac{3\sqrt{\alpha }\left[1.99-\alpha \left(1-\alpha \right)\left(2.15-3.93\alpha +2.7{\alpha }^2\right)\right]}{2\left(1+2\alpha \right){\left(1-\alpha \right)}^{3/2}}}$$

(17)

The initial fracture toughness (K_IC^Q) is determined by substituting the initiation load (P_ini) and the initial crack length (a₀):

$$K_{IC}^Q={\displaystyle \frac{P_{ini}L}{t{h}^{3/2}}}f\left({\alpha }_0\right)$$

(18)

In the P-CMOD curve, P_ini corresponds to the critical point where the curve deviates from the linear-elastic stage (i.e., where stiffness begins to decrease and microcracks start coalescing into macrocracks). This point can be accurately captured by identifying the sudden change in the tangent slope of the curve.

The unstable fracture toughness (K_IC^S) is calculated by substituting the maximum peak load (P_max) and the critical effective crack length (a_c):

$$K_{IC}^S={\displaystyle \frac{P_{\max }L}{t{h}^{3/2}}}f\left({\alpha }_c\right)$$

(19)

Since the span-to-depth ratio of the three-point bending specimens in this study is not exactly 4.0, a correction factor (X) must be applied to the shape function f(α), as shown below:

$$X=1+\left(1-4{\displaystyle \frac{h}{L}}\right)\left[0.08\left(1-{\displaystyle \frac{a}{h}}\right)+0.24{\left(1-{\displaystyle \frac{a}{h}}\right)}^2-0.28{\left(1-{\displaystyle \frac{a}{h}}\right)}^3\right]$$

(20)

The calculation results for both fracture toughness parameters are presented in Figure 19 and Figure 20, respectively.

In terms of fiber reinforcement, the specimens with a 1.5% volume fraction exhibited the optimal flexural toughness. Notably, the enhancement magnitude provided by the fibers differed between the two fracture toughness parameters. For instance, in the NG1.5 group, the initial fracture toughness (K_IC^Q) at 200, 400, 600, and 800 °C was 86.11%, 66.69%, 68.53%, and 42.53% higher than that of the NP group, respectively. Meanwhile, the corresponding increments for unstable fracture toughness (K_IC^S) were 117.89%, 119.69%, 136.81%, and 143.95%. This discrepancy suggests that glass fibers effectively exert a crack-bridging effect after the formation of macroscopic cracks. Specifically, the rapid crack propagation stage represented by K_IC^S is the most critical phase where the fibers fulfill their reinforcing role.

Regarding the cooling regime, taking the G1.5 group as an example, the K_IC^Q of naturally cooled specimens at 200, 400, 600, and 800 °C was 61.77%, 17.17%, 74.65%, and 19.73% higher than that of the water-spraying cooled counterparts, respectively. This indicates that the initial fracture toughness is highly sensitive to the cooling regime. This sensitivity arises because the thermal shock induced by water-spraying cooling tends to exacerbate the development of internal microcracks, which significantly compromises the initial fracture resistance. From the perspective of temperature gradients, as the temperature increased, the fracture toughness across all fiber dosages and cooling methods continued to decline, with the disparities between groups diminishing. This can be attributed to the accumulation of thermal damage, leading to the full propagation of matrix microcracks, the attenuation of aggregate interlocking, and the thermal softening of glass fibers at elevated temperatures.

3.6. Scanning Electron Microscopy Analysis

SEM was utilized to observe the fracture surfaces of the geopolymer concrete specimens subjected to elevated temperatures and distinct cooling treatments. Figure 21 illustrates the morphological evolution of the GPC matrix under various peak temperatures and cooling regimes.

As depicted in Figure 21a, the matrix surface is encapsulated by the N-A-S-H gel, exhibiting a smooth and dense structure without obvious microcracks. A few angular slag particles are visible, typically coated with a dense layer of gel. Compared to the naturally cooled counterparts, the most prominent characteristic of the water-spraying cooled specimens is the emergence of irregular microcracks. As shown in Figure 21d, the surface of the 200 °C specimen contracted rapidly upon contact with cold water, while the interior core remained at a high temperature. The resulting transient thermal stress induced cracking within the gel layer. Simultaneously, microscopic annular gaps were observed between the fly ash particles and the surrounding gel matrix, indicating anisotropic deformation (thermal mismatch) between the two phases under thermal shock.

In the temperature regime of 400 °C to 600 °C, the matrix of the naturally cooled GPC became rough due to drying shrinkage, and microcracks induced by differential thermal expansion began to propagate (as shown in Figure 21b). In contrast, clearly interconnected reticular cracks were visible on the matrix surface of the water-spraying cooled specimens at these temperatures (Figure 21e). In the temperature regime of 400 °C to 600 °C, the matrix of the naturally cooled GPC became rough due to drying shrinkage, and microcracks induced by differential thermal expansion began to propagate (as shown in Figure 21b). In contrast, clearly interconnected reticular cracks were visible on the matrix surface of the water-spraying cooled specimens at these temperatures (Figure 21e). At the extreme high temperature of 800 °C, the gel structure of the naturally cooled specimens collapsed severely, manifesting as a pulverized or porous honeycomb-like morphology. Conversely, owing to the profound thermal shock, penetrating macro-cracks emerged on the matrix surface of the water-spraying cooled specimens (Figure 21f), resulting in a massive loss of the specimens’ tensile and flexural strengths.

Figure 22 illustrates several typical fiber morphologies and failure modes observed under SEM. As shown in Figure 22a, fiber fracture is observed, where the fiber does not slip out of the cavity but is fractured directly. In the SEM field of view, the transversely fractured end-face of the fiber is visible, which may present an uneven, torn morphology. Furthermore, the fiber surface near the fracture point is often still tightly encapsulated by the N-A-S-H gel, indicating robust interfacial bonding at this stage. This mode is commonly observed in the GFGPC specimens at 25 °C and 200 °C. After the matrix cracks under stress, the load is effectively transferred to the fibers bridging the cracks until the fibers reach their ultimate tensile capacity and fracture. This fiber failure mode contributes most significantly to enhancing the peak flexural strength of the GPC specimens.

As illustrated in Figure 22b, fiber pull-out occurs, leaving distinct deep or cylindrical cavities within the GPC matrix upon specimen failure. Concurrently, partially pulled-out fibers are observable, with their surfaces generally appearing smooth and almost devoid of attached geopolymer gel. This suggests that the interfacial bonding strength between the matrix and the fiber is lower than the inherent tensile strength of the fiber itself. This mode is frequently encountered in GFGPC specimens at 400 °C and 600 °C. During the propagation of fracture cracks, the fibers slide and are pulled out from the matrix; this frictional sliding process dissipates a substantial amount of energy. This fiber failure mode is the primary contributor to the enhancement of the flexural toughness of the GPC specimens.

The third mode is interfacial debonding and micro-gaps, as depicted in Figure 22c. In this scenario, the fiber remains in situ—neither fractured nor pulled out—but a conspicuous annular micro-gap emerges between the cylindrical fiber and the surrounding matrix. This is a characteristic manifestation of physical damage, typically observed in specimens subjected to extreme elevated temperatures (e.g., 800 °C) or severe thermal shock (particularly under water-spraying cooling). The phenomenon is attributed to the dehydration-induced shrinkage of the matrix at high temperatures, coupled with the thermal expansion coefficient mismatch between the fiber and the gel, which leads to interfacial separation. Once such gaps form, the crack-bridging and energy-dissipating capabilities of the fibers through friction are substantially compromised.

4. Discussion

Based on the static mechanical tests and analyses presented above, a significant divergence in the mechanical performance of geopolymer concrete under different cooling regimes is observed after exposure to 600 °C and 800 °C. Compared to natural cooling, the specimens subjected to water-spraying cooling exhibited an anomalous enhancement in residual compressive strength, whereas their splitting tensile and flexural strengths suffered a drastic decline. This phenomenon indicates that the influence of the post-heating cooling rate on mechanical properties is non-monotonic. Its essence lies in the competitive mechanism between the densification of the matrix microstructure and macroscopic thermal shock damage under different loading states.

Water-spraying cooling provides a dual positive reinforcement effect on the compressive performance of the GPC matrix at high temperatures. On one hand, within the 600–800 °C range, the amorphous gel phase within the GPC undergoes dehydration, glass transition, and viscous sintering, leading to the preliminary closure of pores. The extremely high cooling rate of water spraying induces a quenching effect, rapidly freezing this densified high-temperature amorphous state. This effectively suppresses the volume expansion and secondary internal stress damage caused by crystal growth (such as the formation of feldspar mineral phases) that occurs during the slow natural cooling process. On the other hand, the high temperature significantly activates the surface reactivity of unreacted precursors. During water cooling, external moisture rapidly infiltrates the specimen, triggering intense secondary geopolymerization. The newly formed gel products heal some of the thermal-induced micropores at the microscopic scale. Under uniaxial compression, internal microcracks tend to be compacted and closed; thus, the macroscopic compressive strength depends more on the matrix density and structural integrity, resulting in the observed anomalous strength increase.

However, for tensile and flexural performance, the severe thermal shock induced by water-spraying cooling becomes the dominant failure factor. When specimens at 600–800 °C are suddenly exposed to cold water, the surface temperature drops sharply, causing a drastic contraction trend, while the core remains in a high-temperature expanded state. This immense internal–external temperature gradient induces transient tensile stresses on the surface that far exceed the material’s tensile limit, leading to the initiation of extensive macroscopic reticular cracks. The microscopic “secondary geopolymerization self-healing” is clearly insufficient to bridge these millimeter-scale macroscopic defects torn by the massive temperature differential.

Splitting tensile and three-point bending tests exhibit extreme mechanical sensitivity to such surface macroscopic defects. According to the weakest-link theory in fracture mechanics, the tensile zone at the bottom of the beam sustains the maximum tensile stress during three-point bending. The surface thermal shock cracks caused by water cooling directly evolve into potent stress concentration points. Under relatively low external loads, the crack tips can propagate rapidly, leading to brittle fracture. Similarly, in splitting tests, lateral tensile stresses accelerate the penetration of cracks along pre-existing thermal shock damage paths.

Consequently, in the 600–800 °C range, while water-spraying cooling partially optimizes the phase state and densification of the GPC matrix at the microscopic level—thereby enhancing compressive strength—it introduces fatal surface opening-mode cracks at the macroscopic level. The varying sensitivity of different loading modes to these defect characteristics ultimately leads to the diametrically opposed trends in compressive versus tensile and flexural strengths under the same cooling regime.

Furthermore, the thermo-mechanical degradation mechanisms of the geopolymer matrix elucidated in this study provide crucial insights for the advancement of structural strengthening technologies, particularly Textile Reinforced Mortar (TRM) systems. Existing durability research on conventional TRM systems predominantly focuses on ordinary Portland cement mortar [47,48], which typically experiences severe dehydration, pore expansion, and subsequent textile–matrix debonding under elevated temperatures. By conceptually excluding the coarse aggregates, the GPC matrix investigated herein presents a highly viable alternative for fire-resistant TRM applications. Unlike OPC concrete, the GPC matrix exhibits unique secondary geopolymerization and localized densification at elevated temperatures. More importantly, the severe thermal shock induced by the water-spraying cooling regime in this study rigorously simulates the extreme hydrothermal environment that TRM-strengthened structures encounter during active firefighting operations. The observed micro-damage phenomena—such as the formation of micro-annular gaps at the fiber–matrix interface (Figure 22c) and the drastic attenuation of unstable fracture toughness under thermal shock—intuitively map to the interface slip and debonding mechanisms between continuous textile grids and the mortar matrix. Consequently, the multi-factor strength prediction model and the microscopic damage evolution laws established in this work not only evaluate the residual capacity of fiber-reinforced GPC but also lay a solid theoretical foundation for assessing the durability and structural integrity of emerging inorganic TRM systems post-fire exposure.

Finally, it is worth noting that the experimental investigations in this study were conducted on relatively small-scale specimens. In quasi-brittle materials like geopolymer concrete, both mechanical properties and thermal damage evolution exhibit a pronounced size effect. In practical engineering fire scenarios, large-scale structural components (such as beams, columns, or tunnel linings) possess massive cross-sectional dimensions, leading to fundamentally different internal thermal gradients, moisture migration behaviors, and structural restraint conditions compared to small-scale samples. For instance, during water-spraying cooling, the massive, high-temperature core of a large member would exert a much stronger geometric restraint on the rapidly contracting surface layer, potentially exacerbating the localized thermal shock damage and altering the macroscopic crack propagation paths. Therefore, while the material-level damage evolution laws and strength prediction models proposed in this study provide a fundamental theoretical framework, extrapolating them directly to the structural level requires careful consideration. Future research should dedicate efforts to investigating the high-temperature mechanical response, thermal spalling behavior, and residual load-bearing capacity of large-scale or full-scale GFGPC structural members, further bridging the gap between material characteristics and practical engineering applications.

5. Conclusions

This study systematically investigated the mechanical evolutionary patterns of glass fiber-reinforced geopolymer concrete after exposure to elevated temperatures (25–800 °C), considering the effects of fiber dosage and cooling regimes (natural cooling vs. water-spraying cooling). Based on the mechanical tests and SEM microstructural analysis, the primary conclusions are as follows:

(1)

Compared to plain geopolymer concrete, the GFGPC specimens with a fiber volume fraction of 1.5% exhibited optimal static mechanical properties across all temperature gradients and cooling regimes. At ambient temperature, the compressive strength, splitting tensile strength, and flexural strength increased by 17.04%, 30.39%, and 97.43%, respectively.

(2)

Under natural cooling, the compressive, splitting tensile, and flexural strengths of GFGPC all demonstrated an initial increase followed by a decrease with rising temperature, peaking at 200 °C (with average enhancements of 12.02%, 15.8%, and 24.5% relative to ambient temperature, respectively). In contrast, under water-spraying cooling, all three strengths exhibited a monotonic decrease. Notably, while the residual compressive strength of the naturally cooled group was higher between 25 °C and 400 °C, the compressive strength retention rate of the water-sprayed group anomalously surpassed that of the naturally cooled group in the extreme temperature range of 600 °C to 800 °C, being on average 12.16% and 20.77% higher, respectively. For splitting tensile and flexural strengths, the naturally cooled group consistently outperformed the water-sprayed group across the entire temperature range.

(3)

The compressive failure absorption energy and splitting energy of both the naturally cooled and water-sprayed GFGPC groups exhibited a continuous degradation trend with increasing temperature, with the naturally cooled group consistently maintaining higher values. The characteristic length, reflecting the brittleness of GFGPC, showed a trend of initial decrease followed by an increase under natural cooling (decreasing by an average of 4.69% at 200 °C). Conversely, under water-spraying cooling, although l_ch increased monotonically due to the highly localized fracture caused by thermal shock, it remained consistently lower than that of the naturally cooled specimens under equivalent conditions.

(4)

The incorporation of fibers significantly enhanced both the initial fracture toughness and the unstable fracture toughness of GFGPC. The enhancement of K_IC^S was more pronounced due to the crack-bridging effect and the frictional energy dissipation during fiber pull-out as the crack opened. Both fracture toughness parameters decayed continuously with increasing temperature, and the values for the naturally cooled group were consistently higher than those for the water-sprayed group.

(5)

A binary nonlinear prediction model was developed based on composite damage evolution theory, which successfully mapped the coupled effects of temperature gradients and fiber characteristic values on the residual compressive strength of GFGPC. The model predictions show high agreement with experimental results, with a correlation coefficient (R²) exceeding 0.90, indicating robust statistical reliability.