Gypsum plasterboards under natural fire — Experimental investigations of thermal properties

The use of fire protection materials is a common approach to ensure the fire resistance of steel elements exposed to fire. In this article, experimental investigations regarding the thermal behavior of gypsum plasterboards for steel elements exposed to natural fires are presented. Material properties, such as the specific heat, the thermal conductivity, and the density of gypsum plasterboards, have been investigated yet, but not especially for natural fire scenarios with different heating rates and cooling phases. For this purpose, experimental investigations of gypsum plasterboards under natural fire exposure are presented. Based on our own experimental investigations, the thermal properties of the investigated gypsum plasterboard for both the heating and cooling phases are demonstrated. Additionally, results from a large-scale fire test on loaded steel beams as well as unloaded steel columns protected with gypsum plasterboards are presented. The test results show a clear dependency on the heating and cooling rate. Furthermore, the thermal material properties change within the heating and cooling phase. So, the main objective of this article is to provide the thermal properties of selected gypsum plasterboard exposed to natural fires in the heating and cooling phases.

To meet a fire resistance steel components usually have to be protected for a severe fire exposure. In addition to reactive systems passive fire protection materials (FPM), such as gypsum plasterboards according to EN 520 (2009) 1 are commonly used for theses purpose.
The fire protection behavior of FPM leads to a delayed heating of the protected steel element and thus guarantees a sufficiently long fire resistance. If the calculation methods of the Eurocodes are to be used for the design of steel elements under fire exposure, for the thermal material properties of the FPM values must be known.
The fire protection behavior of FPM is typically derived by fire tests (EN 1363-1 2 ) according to the ISO 834 standard fire curve. 3 Therefore, the assessment of the fire protection behavior of FPM is solely related to this fire scenario. Depending on the thickness of the FPM and the section factor of the steel element, fire resistances of R 30 to R 180 can be achieved. Based on the desired fire resistance class, some product manufacturers give values for the thickness of FPM depending on the steel section factor. In addition to the standard fire curve, the centralized European rules for the fire design of structural elements in fire (EN 1993-1-2 4 ) also allow fire safety design with natural fire scenarios. A natural fire scenario is a more realistic and often more economical way of design that deviates from the standard fire curve. In contrast to the standard fire curve, which assumes a rapid and continuous rise in temperature, a natural fire curve consists of realistic heating and cooling phases. Natural fire curves depend on various parameters such as ventilation, room configuration and fire load (quantity and quality). Therefore, the heating and cooling rate can vary widely. As such, it is of great importance to know the fire protection behavior and likewise the thermal material properties of FPM in the case of natural fire.
In the German national annex of EN 1993-1-2 (2010), 5 only constant temperature-independent material properties for selected FPM are provided, determined on the basis of component tests using the ISO 834 standard fire curve. 3 The draft of EN 1995-1-2 6

| Thermal decomposition of gypsum plasterboards
In general, gypsum plasterboards consist of a porous solid core of primarily calcium sulfate dihydrate CaSO 4 Á 2H 2 O between two layers of paper.
During fire exposure, the chemically bounded water evaporate and a reversible decomposition process of the core material takes place. Up to 80 C, the reversible dehydration process of calcium sulfate dihydrate to calcium sulfate hemihydrate and then to calcium sulfate anhydrite occurs at 125 C and 150 to 225 C, respectively. 10 The first and second dehydration process is given as a chemical Equations (1) and (2): Both dehydration processes are endothermic. At 375 C to 400 C, there is a third enxothermic reaction from CaSO 4 (III) to CaSO 4 (II) (Equation (3)). The crystalline structure of gypsum transforms from a soluble to an insoluble anhydride CaSO 4 , where energy Q released.
Above 700 C calcium carbonate changes to calcium oxide and carbon dioxide. The exothermic reaction is given in Equation (4): The dehydration processes and the chemical reactions characterize the temperature-depended thermal properties of gypsum plasterboards.

| Thermal properties of gypsum plasterboards-Literature data
The temperature distribution of a protected steel element under natural fire can calculated with the Fourier Equation (5): Therefore, the thermal conductivity λ, the specific heat capacity c p and the density ρ must be known. In the literature, the thermal material properties for gypsum plasterboards are mostly given for the heating phase with different heating rates. The temperaturedependent material properties are determined experimentally, calibrated on fire tests with ISO 834 standard fire curve or usually modified or optimized with numerical simulations. In addition, various gypsum plasterboards (like Type F or Type X) are examined using various testing methods and procedures. According to chemical formulation of gypsum plasterboards, the thermal properties vary. 11 This results in different courses of the thermal properties, which leads to a limited comparability. In Figure 1, the functions for the temperaturedependent thermal properties in the heating phase derived from literature data are shown. Furthermore, the own results are shown, which are explained further in detail.
For measuring the thermal conductivity λ, for example, a thermal conductivity meter, TPS or a Laser flash apparatus (LFA) were used. 8,11,12 Depending on the temperature, λ differs from 0. Furthermore, thermogravimetric analysis (TGA) measurements with heating rates of 10 C and 20 C/min were performed to determine the mass loss. 7,11,16 The literature data shows a decrease of mass at 200 C and 600 C, 7 which is mainly caused by evaporation of chemically bounded water and the dehydration process. Depending on moisture content and temperature, mass loss up to 29% was F I G U R E 1 Literature data of thermal properties of gypsum plasterboards in the heating phase determined. Another mass reduction up to 6% occurs after 650 C following the decompositions of impurities of gypsum. 17 At 1000 C, gypsum plasterboard density ρ is 71% to 90% of their value at ambient temperature. So, the literature data shows the correlation between mass loss and moisture content. While Hollmann et al 8  The simultaneous thermal analysis (STA) was also used by References 8 and 9 to determine mass loss and specific heat capacity c p simultaneously. In addition to STA, DSC was used with different heating rates (2, 5, 10, 20 K/min) to determine c p . The literature data of c p shows a peak at 150 C to 200 C up to 30 000 J/(kgÁK), which results from the phase transformation (endothermic reaction). 10 The higher the moisture content of the gypsum plasterboard, the higher the latent heat and the energy required for the evaporation process.
The evaporation of the chemically bounded water withdraws energy, which is expressed in an increase of c p . According to König, 12 the dehydration processes can be determined more precisely by using DSC with a heating rate of <5 K/min. Schleifer 7 determines c p with a heating rate of 20 K/Min, so the first peak of approx. 25 000 J/(kgÁK) occurs later then the peak of 30 000 J/(kgÁK), which was determine with a heating rate of 5 K/min (cf. König 12 ). Furthermore, the percentage of calcium sulfate dehydrate, can vary for different plasterboards depending on the manufacturer and affects the peak. The larger calcium sulfate dehydrate content, the higher the c p peak at 150 C to 200 C. 15 In the temperature range from 200 C to 400 C, c p is almost constant. At a temperature of 400 C, a short-term increase of c p is due to the transition of the crystalline structure of CaSO 4 . 9 Above 600 C, a further peak appears, which can be explained by the decomposition of calcium carbonate (CaCO 3 ) with release of carbon dioxide. Depending on the temperature, c p differs from 750 to 30 000 J/(kgÁK).
In total, the literature data shows the characteristic functions of the temperature-dependent thermal properties only in the heating phase. Accordingly, there is a lack of information, how the thermal properties of gypsum plasterboards behave in the cooling phase.

| Specimens and experimental test procedures
To determine the fire protection behavior of protected steel elements under natural fire curve, it is necessary to know the temperaturedependent material properties for both the heating and cooling phases, since the thermal material properties deviate according to the temperature. Using thermo-analytical measurement methods and procedures, λ, c p , and ρ are determined as a function of temperature.
For the experimental investigations gypsum plasterboards (Typ F, , ρ 20 C = 780 kg/m 3 ) were used. The λ was measured with the Transient Plane Source (TPS) method according to DIN EN ISO 22007-2 (2015). 13 The differential scanning calorimetry (DSC) analyses in accordance with DIN 51007 (1994) 19 are performed in order to determine c p . Additionally, the temperature-dependent mass loss of the gypsum plasterboards are determined by TGA according to DIN 51006 (2005). 20 Previously, fire simulations with computational fluid dynamicsmodels and zone models were carried out to determine the heating and cooling rates in different room and use configurations. 21 The results were evaluated statistically. The mean values of the heating rate (20 K/min) and cooling rates (6 K/min) were selected from a static distribution. Due to the thermo-analytical measuring methods and procedures, the heating rate of 10 K/min, 40 K/min and the cooling rate of 10 K/min were additionally selected. The experimental results from the thermo-analytical measuring methods and procedures are provided in the following.

| Temperature-dependent thermal conductivity
For measuring λ a sensor, which works as a thermocouple (TC) and plane heat source at the same time, was placed between two identical In the cooling phase, the measured values for the thermal conductivity are lower (see Figure 3). It can be assumed that the gypsum plasterboards were subjected to irreversible material changes due to temperature exposure during the heating phase.  19 and ISO 11357-1 (2016). 22 For the determination of c p , a Mettler Toledo DSC 822e with a temperature sensor of À180 C to 700 C was used. The samples were placed in the sample holder with small quantities in the mg range (9.6-11.22 mg, in total three samples). Therefore, the gypsum plasterboards were crushed and pul- Based on the results, it can be seen that c p is temperature-and heating-rate dependent. Additionally, c p diverges in the heating and cooling phases. Accordingly, the assumption of a constant value does not justify the temperature-dependent c p .
For the gypsum plasterboard (see Figure 3) During the cooling phase, c p was nearly constant between 0.45 and 0.9 kJ/(kgÁK). The measurement results in the cooling phase demonstrated that the cooling rate affected c p . In the cooling phase, the measured values for c p are lower at a cooling rate of 10 K/min than at a cooling rate of 6 K/min. However, the direct influence of the cooling rate on the measured c p cannot be clearly derived from these data, because c p had been previously determined at different heating rates.
Overall, the test results of the cooling phase depend on the maximum temperature and heating rate of the heating phase and the cooling rate itself. Nevertheless, the values can be used to show that the c p decreases or stays nearly constant as the temperature is reduced. At ambient temperature, the temperature and heating rates dependent on c p are scattered between 0.40 and0.78 kJ/(kgÁK).

| Mass loss and temperature-dependent density
A total of 24 samples were measured at ambient temperature to determine the density of the investigated gypsum plasterboard. The initial ρ of the investigated gypsum plasterboard was determined to be 807 kg/m 3 (moisture content 2.63%). As the initial ρ of the investigated gypsum plasterboard changes due to mass loss and is affected by the moisture content, the curves given in Figure 5 show a dependency on the heating rates. In order to calculate the temperature The temperature-dependent mass loss and the density decreased during the heating phase. The irreversible maximum mass loss was determined to be 23% at 1000 C for 10 K/min, 20 K/min and 40 K/ min, respectively. Therefore, ρ was reduced to 621 kg/m 3 at 1000 C.
The lower the heating rate, the earlier the mass loss occurred. The mass loss correlated to the chemical and physical processes in the gypsum plasterboard. The chemical and physical processes were shown in the course of the mass loss. Therefore, the two dehydration reactions in the gypsum plasterboard, which took place with the increase in temperature, were shown as a mass loss of 18% at 100 and 200 C (see Figure 5).
In the cooling phase, the results showed that the cooling rate is irrelevant. After the mass loss occurs in the heating phase, a constant value of the mass and the density can be assumed. The lower the temperature of the heating rate, the lower the mass loss and the higher ρ in the cooling phase. properties. In addition, scaled effects were investigated more closely since they were not taken into account by the thermo-analytical tests.
In particular, the crack behavior, the joint formation and the fire protection behavior of the gypsum plasterboard with and without loads were investigated. One unloaded I-section columns (HEA 240) with a length of 1000 mm was used as test specimens as well as one loaded I-section beam (HEA 240, length 4900 mm). Gypsum plasterboard was applied to protect the steel elements. In the following the results of the unloaded and loaded specimens with gypsum plasterboard will be presented.
To estimate the fire protection behavior of the gypsum plasterboard, the test specimens were boxed with a single-layer of gypsum plasterboard (20 mm thickness). Other thicknesses were not investigated. To ensure a three-sided fire exposure, the beam was dispersed under the furnace ceiling, which consisted of cover panels made from aerated concrete. The column was placed on the bottom of the furnace, thus ensuring a four-sided fire exposure. The free ends of the test specimen were covered by 20 mm thick insulation boards made from vermiculite. In order to avoid heat transfer into the test specimen, the protected steel column was additionally placed on 20 mm of thick rock wool and plaster.
The loaded steel beam with gypsum plasterboard was designed as a single-span beam with a length of 4.90 m. Additionally, the loaded steel beam had two load bearing points at the third points of the length, which were applied with a constant load of 40 kN to achieve a 30% utilization of the steel beam. In order to prevent the gypsum plasterboard from falling off, the protected steel beam was loaded less. However, for the loaded specimen, the load was applied according to a two-point bending test (see Figure 6).

| Test results
The measured temperatures of the steel profiles of the beam (1/2 Á length) and column (1/2 Á length) with gypsum plasterboard (20 mm thickness) are shown in Figure 9.
During the heating phase, the furnace temperature and the steel temperature rose continuously. Due to the thermal fire protection behavior of gypsum plasterboard, the steel temperatures were below the furnace temperature. The natural fire scenario reached its maximum temperature of 852 C after 31 minutes. The temperature development inside the column was more homogeneous than the beam due to the four-side fire exposure.
In the cooling phase, all test specimens showed lower cooling rates when compared to the furnace temperature, resulting in a properties λ, c p and ρ were determined up to 600 C with different heating and cooling rates. Thus, the authors had to make assumptions about the temperature-dependent functions of the thermal properties for >600 C, due to the missing measurement data. Figure 10 shows the proposal for temperature-dependent functions of the thermal properties for gypsum plaster board.
For the derivation of the proposed functions in a first step the measured values were used and secondly calibrated with the temperature results of the large-scale fire test to ensure the application for fire design of real structures. Therefore the functions of λ, c p and ρ were optimized until the best temperature-dependent curve was obtained.
Due to the irreversible material behavior of gypsum plasterboard, The temperature-dependent functions for the investigated gypsum plasterboard under natural fire of up to 1000 C are also given in Figure 10. For λ, the mean values of the TPS measurement for both the heating phase and the cooling phase have been reduced by an average of~30% due to high variation in the measured values and the fact that the TPS achieves slightly higher measured values 23 (see Figure 3). The λ values were modified until the best temperaturedependent curve was achieved. The thermally induced crack formations, which were also detected in the large-scale fire test, were simplified by increasing λ from temperatures of 500 C upwards (cf. References 8 and 9). Further, shrinkage, cracks and ablation of gypsum plasterboards is simplified modeled by increasing λ starting from 600 C. The temperature-dependent curve of c p was derived from the DSC data (up to 600 C), with a heating rate of 20 K/min (see

| CONCLUSIONS
The main objective of this article is to show the difference in the thermal properties of selected gypsum plasterboard in the heating and subsequent cooling phase. Furthermore, the fire protection performance of gypsum plasterboards under natural fire was investigated.
This was achieved by experimental fire tests at both small and large scale. The temperature and heating-rate-dependent material properties of the gypsum plasterboard, such as λ, c p , and ρ, were determined experimentally. The investigated gypsum plasterboard reveals a clear dependency on heating and cooling rates. However, currently only constant values of thermal properties for the cooling phase exist. This is due to the fact that the fire protection behavior of gypsum plasterboards, is solely assessed for fire exposure according to the standard fire curve. As the temperature regime of a natural fire scenario can deviate significantly from the standard fire curve, it is highly recommended to assess the influence of varying heating and cooling conditions on the thermal behavior of the investigated gypsum plasterboard.
The fire protection behavior was tested in a large-scale fire test under a natural fire curve, both with and without mechanical load.
The influence of the mechanical load on the thermal behavior of the gypsum plasterboard seems to be small, as an effect on the fire protection behavior of the gypsum plasterboard was not evident 14 (see Figure 9). Only an increased number of cracks (temperature and load induced) and the temperature dependence of the material properties was determined. But the beam utilization (30%) was comparatively low, so further research is needed to investigate possible correlation between fire protection behavior and load utilization.
The results of the large-scale fire test were used to calibrate the thermal material properties. Based on these thermal material properties, temperature-dependent functions of gypsum plasterboard for the heating phase and especially for the cooling phase were derived.
Due to the wide range of fire protection boards and natural fires, the experimental procedures and boundary conditions presented in this article can only be applied when the required material properties of the investigated gypsum plasterboard and the natural fire (heating and cooling rate) are comparable.

ACKNOWLEDGMENT
The authors thanks TU Braunschweig to finance this publication. The work presented in this article is part of the German research project "Test procedures for thermal material properties of fire protection claddings and intumescent coatings for the design of steel structures exposed to natural fires" (AiF 19176 N). The authors express their deep gratitude for the financial support received from the Federal Ministry for Economic Affairs and Energy.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.