Influence of Cooling Methods on the Residual Mechanical Behavior of Fire-Exposed Concrete: An Experimental Study

This work reports the main conclusions of a study on the mechanical behavior of concrete under ISO 834 fire with different cooling methods. The aim of this research was to provide reliable data for the analysis of structures damaged by fire. The experimental program used cylindrical concrete test specimens subjected to ISO 834 heating in a furnace up to maximum gas temperatures of 400, 500, 600, 700, and 800 °C. The compressive strength was measured in three situations: (a) at the different temperature levels reached in the furnace; (b) after a natural cooling process; and (c) after aspersion with water at ambient temperature. The results indicate that the concrete residual compressive strength is fairly dependent on the maximum temperature reached in the furnace and revealed that concrete of a lower strength preserved relatively higher levels of strength. The cooling method significantly influenced the strength, albeit at a lower intensity. In all cases, the residual strength remained in the range of 38% to 67% of the strength at ambient temperature. The statistical analysis showed that the data obtained by the experimental program are significant and confirmed the influence of the conditions imposed on the residual strength.


Introduction
In recent decades, the study of concrete during and after a fire has been scant in Brazil. In the mid-1970s, after two buildings were damaged by fire, the development of fire safety technology became a challenge for engineers and fire authorities. A technical standard for the design of concrete structures under fire was published in 1976. It required fire resistance times for building structures, depending on their height and occupancy. This standard prescribed adequate concrete covering according to the required times for fire resistance. In general, considering low-and medium-rise buildings in Brazil (up to 15 floors), normal coverings of 25 mm were sufficient to achieve the desired fire resistance. In Europe and the United States, design guidelines are available in the Eurocodes (EN 1992-1-2) and affects strength. Tanaçan, Ersoy, and Arpacıoglu [30] investigated aerated concrete and concluded that an abrupt water cooling regime soon after heating increases cracking, with a subsequent strength drop. Lee, Xi, and Willian [31] found that the stiffness, strength, and permeability are influenced by heating and cooling cycles. Karakoç [32] evaluated the compressive strength of light aggregate concretes and demonstrate that it is influenced by the cooling methods. Kim et al. [33] concluded that the cooling rate affects the dimensional variation of the aggregates and that the mechanical properties are deteriorated by accelerated cooling. Xiang et al. [34] realized that, for slow cooling, the heating temperature influences the residual cracking pattern and failure mode during compression testing.
In this context, it is essential to understand the behavior of concrete structures affected by fire and the influence of heat on building materials. Several studies have evaluated the residual properties of cement-based materials after a high temperature, whilst others have investigated the physical and chemical changes occurring at the microstructural level. However, comprehensive studies evaluating the influence of cooling methods on the mechanical properties of concrete are scarce in the literature. Therefore, the research reported here aims to broaden the knowledge on the subject, providing new results, and was designed to answer two questions: How is the residual strength of concrete under ISO 834 fires affected by the strength at ambient temperatures? To what extent does the cooling of hot concrete specimens affect the residual strength? These questions are particularly important for engineers and fire brigade officials when examining fire-damaged concrete structures.

Materials and Mix Proportion
The coarse aggregates were limestone with a maximum diameter of 25 mm and fineness modulus of 7.05. A sand of fineness modulus 2.68 and a cement ASTM Type III [35] with a compression strength of 38.1 MPa were used. The dosage process considered theoretical compressive strengths (f ck ) of 15, 21, 25, and 35 MPa with water/cement ratios of 0.620, 0.517, 0.458, and 0.338, respectively. To determine the optimal ratio between sand and gravel, the modified Reilly [36] method as exposed in Carvalho [37] was used. The result was a minimum porosity with a mixture of 65.5% coarse aggregates and 34.5% sand. The water/dry materials ratio (A%) was related to the fineness modulus using 11 distinct dosages whose slump tests were in the range of 55 to 92 mm. The water/cement ratio was determined for concrete of slump 70 mm. The A G characteristic of aggregates following the Reilly [36] method was 0.474. The Reilly constants were M 1 = 4.2031 and M 2 = 0.3281. All specimens were kept at ambient humidity for 90 days until testing.

Specimen Preparation and Curing
The experimental program included 252 cylindrical concrete specimens that were 100 mm in diameter and 200 mm in height. Initially, the materials were separated and weighed according to the predefined proportions. Then, the components were mechanically mixed in the following order: coarse aggregates, sand, half of the water content, Portland cement, and the remaining water. After obtaining a homogeneous mixture, the fresh concrete was poured into the cylindrical molds. Manual densification was performed to reduce voids and air bubbles, ensuring a uniform mass. After 24 h of molding, the stiffened specimens were removed from the molds and allowed to wet cure for 28 days, which was a sufficient time for cement hydration reactions and strength gain.

Heating and Cooling Methods
The concrete test specimens were subjected to ISO 834 heating in a cylindrical furnace of a 770 mm diameter and 800 mm height, up to maximum gas temperatures (θ gmax ) of 400, 500, 600, 700, and 800 • C (48 specimens at each temperature). A typical time-temperature curve of the furnace is shown in Figure 1. For each temperature in the furnace, three situations have been considered: hot tested (HT), naturally cooled (NC), and cooled after spraying with room temperature water (WAC). As for the HT situation, the model is rated still hot according to the heating temperature. In the second methodology, the material is naturally cooled in air for 24 h. Finally, in WAC (Figure 2), the cooling is done abruptly, simulating a fire situation where water is applied to the structure at a high temperature, in which there is a great thermal shock in the material. is done abruptly, simulating a fire situation where water is applied to the structure at a high temperature, in which there is a great thermal shock in the material.

Test Set-Up
The strength was evaluated by the compression test, with a slow load application, to configure a quasi-static condition. The tests were performed on universal servo hydraulic testing equipment, permitting accurate load measurements to be obtained. All conditions were analysed (theoretical strength, temperature, and cooling method). The control group was tested at room temperature for each theoretical strength.

Statistical Approach
The consistency of the data obtained was evaluated by statistical analysis. The significance level adopted was 5% (0.05). Therefore, the analyses had a confidence interval of 95%. Univariate analysis of variance (ANOVA) using the F-ratio was chosen as the analysis technique. The impact of the conditions imposed on the compressive strength was measured by the size of the effect, using the variances and the parameter eta-square (η 2 ). The null effect was identified by the value zero, and values close to or greater than 1 indicated a broad effect. The observed power represented the robustness of the tests in identifying the differences between the groups analysed. Values close to one indicated a high test power. is done abruptly, simulating a fire situation where water is applied to the structure at a high temperature, in which there is a great thermal shock in the material.

Test Set-Up
The strength was evaluated by the compression test, with a slow load application, to configure a quasi-static condition. The tests were performed on universal servo hydraulic testing equipment, permitting accurate load measurements to be obtained. All conditions were analysed (theoretical strength, temperature, and cooling method). The control group was tested at room temperature for each theoretical strength.

Statistical Approach
The consistency of the data obtained was evaluated by statistical analysis. The significance level adopted was 5% (0.05). Therefore, the analyses had a confidence interval of 95%. Univariate analysis of variance (ANOVA) using the F-ratio was chosen as the analysis technique. The impact of the conditions imposed on the compressive strength was measured by the size of the effect, using the variances and the parameter eta-square (η 2 ). The null effect was identified by the value zero, and values close to or greater than 1 indicated a broad effect. The observed power represented the robustness of the tests in identifying the differences between the groups analysed. Values close to one indicated a high test power.

Test Set-Up
The strength was evaluated by the compression test, with a slow load application, to configure a quasi-static condition. The tests were performed on universal servo hydraulic testing equipment, permitting accurate load measurements to be obtained. All conditions were analysed (theoretical strength, temperature, and cooling method). The control group was tested at room temperature for each theoretical strength.

Statistical Approach
The consistency of the data obtained was evaluated by statistical analysis. The significance level adopted was 5% (0.05). Therefore, the analyses had a confidence interval of 95%. Univariate analysis of variance (ANOVA) using the F-ratio was chosen as the analysis technique. The impact of the conditions imposed on the compressive strength was measured by the size of the effect, using the variances and the parameter eta-square (η 2 ). The null effect was identified by the value zero, and values close to or greater than 1 indicated a broad effect. The observed power represented the robustness of the tests in identifying the differences between the groups analysed. Values close to one indicated a high test power.            Table 4 shows the ANOVA results. The corrected model and the intercept, the standard models of the analysis program, were significant (p < 0.05), which enabled the variables to be correlated analytically. The categorical variable test method (M), temperature (T), and theoretical strength (fck) were significant (p < 0.05), which indicates a relevant effect on the compressive strength. The effect size was moderate for the test method and more intense for the temperature and strength variables. However, the values obtained indicate a significant influence of the conditions on the strength. The power observed was the maximum for all situations evaluated. The adopted model can explain 88% of the observed values (coefficient of determination Rsquare = 0.880). Therefore, the obtained data are statistically significant and adequately explain the proposed problem. In the first analysis, the strength had a similar trend. There was a decrease in strength as the temperature increased, especially between room temperature and 400 • C. This result seems to be reasonable considering that 400 • C was reached in all experiments, which is the temperature which corresponds to major damage to the concrete microstructure, as observed by several authors [10][11][12]. Between 400 and 600 • C, the loss of strength was subtle, yet significant. After 600 • C, there was a sudden reduction in the strength capacity. Some tests resulted in strength gain; however, there were specific occurrences, most likely due to the variability of the data (Figure 3, f ck = 25 MPa, 500 • C and Figure 5, f ck = 35 MPa, 500 • C). For heating up to 800 • C, the residual compressive strength remained in the range of 38% to 67% of the ambient temperature strength. Natural air cooling resulted in lower decreases in strength, while the greatest variations were obtained with water spray. The hot test resulted in an intermediate condition among the others. Similar results were obtained by Chan, Luo, and Sun [29]. They concluded that abrupt cooling by immersion in water promotes greater decreases in compressive strength compared to gradual cooling inside the furnace. Furthermore, they observed that higher temperatures imply a higher porosity and consequently lower strength. The lowest relative residual strengths (%) were reached by the concretes with the highest theoretical strengths. As also verified by Botte and Caspeele [1], there was a tendency in which the difference between both cooling methods decreased when the concrete was exposed to higher temperatures. Table 4 shows the ANOVA results. The corrected model and the intercept, the standard models of the analysis program, were significant (p < 0.05), which enabled the variables to be correlated analytically. The categorical variable test method (M), temperature (T), and theoretical strength (f ck ) were significant (p < 0.05), which indicates a relevant effect on the compressive strength. The effect size was moderate for the test method and more intense for the temperature and strength variables. However, the values obtained indicate a significant influence of the conditions on the strength. The power observed was the maximum for all situations evaluated. The adopted model can explain 88% of the observed values (coefficient of determination R square = 0.880). Therefore, the obtained data are statistically significant and adequately explain the proposed problem.  Figure 6 identifies specimens in three different situations: (a) inside the oven during heating; (b) after slow cooling in the air; and (c) with and without the spalling phenomenon. The first notable point refers to the physical phenomena related to dimensional variation and cracking [3,15]. The specimens exposed to the flow of heated gases presented extensive cracking, both on the upper face ( Figure 6b) and on the lateral ones (detail I of Figure 6c). The cracks were mainly due to volumetric dimensional variation due to the different temperature gradients. Furthermore, the surface exposed to the environment dissipated heat faster than the interior of the material, leading to different dimensional variations along the volume [38]. Cementitious materials, included among ceramic materials, are fragile, they do not tolerate strains, and their performance is significantly influenced by internal imperfections (pores, cracks, and defects in the microstructural arrangement) [39][40][41]. Concrete, idealized as a particulate composite, has constituents (cementitious matrix, sand, and gravel) that have different thermal expansion coefficients [42]. Furthermore, the interfacial transition zone (ITZ) functions as a site capable of promoting crack propagation [43,44]. The distinct dimensional variation between the constituents significantly affects the ITZ, promotes cracking of the material, increases defects, and consequently reduces the strength. The internal stresses acting during loading are distributed in the material volume and are amplified at the ends of the cracks [45]. At the maximum loading stage, the material collapses and fails due to the unstable mechanism of crack propagation [46,47]. Another point concerns the degradation of the cement matrix and the particulate constituents, which deteriorate with increasing heat. Phase changes, chemical changes, the evaporation of free water, and the release of chemically bound water occur, which affect the strength drops [48]. The phenomenon of fragmentation occurred in some specimens (detail II of Figure 6c), with considerable material loss. Fragmentation results from the pressure exerted by water vapor in the capillary pores and the internal stresses generated by the thermal gradient [18].

Fire Degradation and Cooling Implications
If there is no material fragmentation during heating, chemical and physical phenomena, acting in a synergistic manner, promote a decrease in strength. Under the three conditions evaluated, the mechanisms of chemical and structural degradation mostly occurred during heating. Therefore, the portion of lost strength related to cooling methods is governed by physical phenomena [4]. Abrupt cooling is the most severe condition. When the material encounters water, it experiences a marked temperature difference; thus, the heat exchanges intensify, as do the dimensional variations. Both the cementitious matrix and the aggregates are damaged; however, the ITZ is highly damaged (defects and microcracking). The result is a marked decrease in strength compared to other conditions. Slow cooling is the best condition, even with a decrease in strength. By losing energy slowly, the material accommodates the dimensional variations more efficiently, resulting in less damaged constituents and ITZs [14][15][16]. The hot test is an unfavourable thermodynamic condition due to the marked energy level of the system, thus resulting in an intermediate condition between those evaluated. The decrease in residual strength that occurs as the theoretical strength increases can be explained in a similar way. Generally, a higher strength implies a denser, more cohesive cementitious matrix with a lower porosity. However, the material ability to accommodate stresses and strains decreases, i.e., the material becomes less tolerant to strains. Therefore, considering the heating and cooling stages, the greater the strength, Materials 2019, 12, 3512 9 of 12 the greater the damage to the material due to the dimensional variation, which results in more severe cracking, degradation, and irregularities in the ITZ [27].

Conclusions
This work presents a case study on the effect of the cooling method on the mechanical behavior of concrete after a fire according to ISO 843. Based on the results of the experimental study, the following conclusions can be formulated: (1) The combined conditions (temperature, theoretical strength, and cooling method) had a significant influence on the residual strength. In general, the worst-case scenario was obtained considering high temperatures in the order of 800 °C, sudden cooling with water spray, and higher theoretical strengths. The best-case scenario was obtained with a 400 °C temperature, slow air cooling, and the lowest theoretical strengths; (2) Sudden cooling with water spraying was the most severe condition, resulting in the greatest decreases in strength. This result can be explained by the sudden dimensional variation of the particulate constituents and the cementitious matrix, which generates cracks and weakens the ITZ; (3) Concretes with a higher theoretical strength showed a lower residual strength. A higher strength implied a denser cementitious matrix, a lower porosity, and more consistent ITZ. However, the accommodation of stresses and strains was impaired, resulting in more extensive cracking and more severe degradation of the material; (4) The higher the temperature, the greater the losses of strength due to the degradation of concrete by physical and chemical phenomena. During heating, there was evaporation of free water, the release of chemically bound water, phase changes, and chemical changes; (5) Statistical analysis showed that the values obtained by the mechanical tests were significant and that the conditions evaluated (temperature, theoretical strength, and cooling method) significantly influenced the residual strength; (6) In future studies, it is suggested that changes in the structure of the concrete cement matrix are studied using a scanning electron microscope (SEM) to aggregate information about changes in the matrix as a function of heating and cooling processes.

Conclusions
This work presents a case study on the effect of the cooling method on the mechanical behavior of concrete after a fire according to ISO 843. Based on the results of the experimental study, the following conclusions can be formulated: (1) The combined conditions (temperature, theoretical strength, and cooling method) had a significant influence on the residual strength. In general, the worst-case scenario was obtained considering high temperatures in the order of 800 • C, sudden cooling with water spray, and higher theoretical strengths. The best-case scenario was obtained with a 400 • C temperature, slow air cooling, and the lowest theoretical strengths; (2) Sudden cooling with water spraying was the most severe condition, resulting in the greatest decreases in strength. This result can be explained by the sudden dimensional variation of the particulate constituents and the cementitious matrix, which generates cracks and weakens the ITZ; (3) Concretes with a higher theoretical strength showed a lower residual strength. A higher strength implied a denser cementitious matrix, a lower porosity, and more consistent ITZ. However, the accommodation of stresses and strains was impaired, resulting in more extensive cracking and more severe degradation of the material; (4) The higher the temperature, the greater the losses of strength due to the degradation of concrete by physical and chemical phenomena. During heating, there was evaporation of free water, the release of chemically bound water, phase changes, and chemical changes; (5) Statistical analysis showed that the values obtained by the mechanical tests were significant and that the conditions evaluated (temperature, theoretical strength, and cooling method) significantly influenced the residual strength; (6) In future studies, it is suggested that changes in the structure of the concrete cement matrix are studied using a scanning electron microscope (SEM) to aggregate information about changes in the matrix as a function of heating and cooling processes.