Carbon reinforced concrete exposed to high temperatures

The material behavior of carbon reinforced concrete exposed to high temperatures was examined in 118 tensile and 81 bond tests using two material combinations. This analysis served to determine resilient material values under temperature influence for current carbon reinforcements and to show starting points for further test series. In this article, the test specimens for the tensile and bond tests are described in detail as well as the used test setup. Special attention is given to the physical–chemical behavior of the impregnated material, which were analyzed namely with dynamic mechanical thermal analysis (DMTA) and thermal gravimetric analysis (TGA). Finally, the results of the tensile and bond tests under different temperature exposures are presented and classified.


| INTRODUCTION
For several decades, fiber reinforced polymers (FRPs) as reinforcement for concrete have been investigated under different load and environmental conditions. But because of the variety of different fibers and polymers, which could be used to fabricate FRP, it is not possible to describe the material in general. That is why nearly every reinforcement must be tested under diverse load and environmental conditions to ensure that the material properties in an application meet expectations and calculations.
In the lifecycle of a building, it is possible that it is exposed to a fire. In general, fire is a load condition considered in the calculation of the building and its construction elements. There are various ways to experimentally determine material properties or the structural behavior of an element at elevated temperatures. A steel reinforced beam for example will most likely be tested in a furnace exposed to the Cellulosic fire curve according to ISO-834. 1 In such a fire-resistance test, one determines the duration for which the beam can endure the fire. It is possible to identify the temperature profile of the construction element and the time of failure, however it is not possible to calculate the element under the influence of elevated temperature. Therefore, high temperature tests are needed to be able to calculate construction elements exposed to high temperature in a building.
For steel reinforced concrete, material values when exposed to fire and high temperatures are defined in EN 1992-1-2. 2 For FRP materials, this is not yet specified in the standards in general. For example, the fire load case is excluded in the guideline ACI 440. 3 However, experimental values with a wide range of materials and investigated temperatures are known, for example, References 4-6. Within the evaluation of one's own test series, the values of previous investigations are presented and used for comparison. The high temperature behavior was investigated on two different material combinations (MCs) consisting of two diverse two-dimensional FRP reinforcements in various concrete matrices.
The focus of the research was to determine design values of the considered materials exposed to high temperature. With the knowledge of the design values under temperature it is possible to calculate carbon reinforced construction elements which are exposed to fire. Therefore, tensile and bond test were executed to determine the tensile strength and the bond behavior under defined temperatures. The tensile properties of the investigated material were tested in stationary and transient tests, the bond behavior only in stationary tests. According to EN 1992-1-2, 2 a heating rate was used ranging between 2 and 50 K/min for both types of tests.
A detailed documentation of all examinations contains. 7 2 | MATERIALS AND EXPERIMENTAL PROGRAM

| Material and test specimens
Two different MCs were determined in this investigation. The first combination (MC 1) consisted of an inflexible carbon mesh impregnated with an epoxy resin and a high strength concrete. The inflexible carbon mesh was purchased from the company solidian GmbH with the product designation GRID Q95/95-CCE-38-E1 (2017), 8 see Figure 1, left side. The carbon rovings have a crosssectional area of 3.62 mm 2 in warp and weft direction. According to the manufacturers' specifications, the fabric has a characteristic tensile strength in the warp direction (in the weft direction) of 2300 (2000) MPa and a modulus of elasticity of 220,000 (205,000) MPa. The concrete mixture was especially developed for carbon reinforced concrete. 9 The maximum grain size of the grit is 16 mm. This mixture has a high compressive strength of 125.5 N/ mm 2 , determined on 100 specimens for 28 days. The reinforcement tended to split the concrete in the reinforcement plane. 10 This effect was caused by the stiffness of the impregnation and the conditional manufacturing profiling of the rovings, for more details see Reference 10. Therefore, the relatively high tensile strength of the concrete was advantageous.
Because of the self-compacting properties, the mixture could be cast into the formwork where one layer of reinforcement was fixed on both sides. First, slabs with dimensions of 99 cm (length) Â 38 cm (width) Â 3 cm (height) were cast. Then the single test specimens were cut out of the plates with the dimensions of 99 cm (length) Â 10 cm (width) Â 3 cm (height) for tensile tests and 65 cm (length) Â 8 cm (width) Â 3 cm (height) for bond tests. The reinforcement was embedded centrically with regard to the thickness of the slab and parallel to its surface. The MC 1 was demolded out of the formwork after 1 day and was stored in a water tank until the 7th day. Before the specimens were tested, they were stored from day 7 for at least 90 days in the climate chamber with 20 C and 65% relative humidity.
The second material combination (MC 2) comprised a carbon textile with an acrylate impregnation. This mesh is even more flexible than a mesh impregnated with an epoxy resin. The reinforcement was ordered from the company V. Fraas GmbH (now Wilhelm Kneitz solution in textile GmbH) with the product name Sitgrid040 (2017), for the geometry see Figure 2. The carbon rovings have a cross section area of 1.81 mm 2 in warp direction and of 0.45 mm 2 in weft direction. The concrete mixture of this MC is a sprayable concrete Pagel TF10 11 with a maximum grain size of 1 mm. Only water needs to be added to the mixture on site or in the laboratory. According to Reference 11, the minimum concrete compressive strength after 28 days is 80 MPa and the modulus of elasticity is 25,000 MPa.
For the test specimens, thin slabs were manufactured in a hand-lamination process by layering concrete and textile reinforcement alternately in a steel formwork. The slabs had a dimension of 120 cm (length) Â 70 cm (width) Â 1.6 cm (height) and the specimens were cut out of them subsequently. The geometry of the tensile specimens was 110 cm (length) Â 6 cm (width) Â 1.6 cm (height); the bond specimens were smaller with 65 cm (length) Â 5 cm (width) Â 1.6 cm (height).
The plates were stripped after 3 days and then stored in the water tank until day 7. Analogous to MC 1, the specimens were stored from day 7 until at least 90 days in the climate chamber with 20 C and 65% humidity and cut in between.

| Heating chamber
In order to be able to test the carbon reinforced concrete specimens exposed to high temperature, a heating chamber was needed. The chamber was closed so that occurring gases could be exhausted and the temperature was shielded from the surrounding climate. The temperature impact was applied to the test specimen with the help of twin-tube infrared emitters. The construction of the heating chamber was based on the structure of Ehlig 12,13 which was, however, designed open. The frame of the chamber consisted of aluminum profiles, which was cladded with fire protection panels. To observe the test specimens during the test, openings were provided in the fire protection panels, see Figure 3.
The centrally positioned specimens were heated from both sides. Two twin-tube infrared emitters per side were used for tensile specimens and only one per side for bond specimens because the bond specimens were narrower in size. The basic interior structure of the chamber is shown in Figure 4. To control the infrared emitters, one thermocouple is required on the test specimen for each infrared emitter. The thermocouples were attached to the test specimen with a heatconducting paste in order to determine the surface temperature as best as possible.
To control the temperature exposure, thermocouples were embedded in the plane of the reinforcement in some samples. In general, there was a good agreement between the temperatures measured on the outside of the test specimen and the temperatures in the plane of the carbon mesh.

| Experimental program
The experimental program varies between tensile and bond tests because of the different temperature properties which were expected in both tests.
The tensile behavior under high temperature exposure was examined with stationary and transient tests with a temperature ranging between 100 and 600 C. In stationary tests, the stress-strain curve was determined at a defined temperature, see Figure 5 on the left side. For this purpose, the specimen was heated before loading until failure. The MC 2 was preloaded before heating, because the preferred use was the strengthening of reinforced concrete structures. In transient tests, the creep deformations under temperature exposure were considered. For this purpose, a load was first applied up to a predefined value. At this constant load level, the temperature was now increased, also up to a target value, see Figure 5 on the right side. The transient tests ended with the tensile failure of the carbon reinforcement due to load and temperature exposure.
To determine the tensile load-bearing behavior of the two MCs exposed to high temperature, stationary tests were carried out at temperatures 100, 200, 300, 400, 500, and 600 C. Transient tensile tests were carried out at load levels of 50%, 60%, 70%, and 80% of the tensile strength at room temperature to expand material properties. In order to determine a sufficient database, at least five tests were carried out per temperature and load level. If the tests showed wide scattering, the number of tests was increased.
In the case of the bond tests, only stationary tests were carried out because the information which was determined by the transient tests, was unreliable for the chosen test setup because of the really short bond length. The stationary tests were carried out in a significantly lower temperature range than the tensile tests as a consequence of the temperature behavior of the impregnation, which had more impact on the bond strength than on the tensile strength. The following temperature levels were tested: 100, 150, 200, 300, and 400 C. Again, at least five tests were carried out.

| Tensile test setup
To determine the tensile strength under thermal exposure, tensile tests were carried out on the composite materials, for more details see Schütze et al. 14 The single layer reinforced test specimens, which were produced for this purpose, had a length of 0.99 m (MC 1) or 1.1 m (MC 2) and varied in width and thickness depending on the MC, see Section 2.1. The temperature was applied on a length of 30 cm from the infrared emitters in the middle of the specimen. The heating chamber had a dimension of 46 cm, see Figure 6. Thus, the test specimen protruded 26 cm (MC1) and 32 cm (MC 2) on both sides out of the chamber. In order to protect the load introduction construction from the heat, 7 cm space was left between the clamping and the heating chamber.
The fixture consisted of profiled steel plates that were screwed together with six screws to prevent the test specimen from slipping out of the clamping. In order to prevent bond failure out of the anchorage area, the test specimen protruded 13 cm beyond the fixture. The temperature was only applied to a limited area in the middle of the sample in order to initiate a planned tensile failure there. The load introduction construction was connected to the testing machine in such a way that the specimen can rotate in plane of the reinforcement, thus reducing any constraining stresses.
The deformations were documented by using linear variable differential transformers (LVDT). Measuring in the heated area was not possible because of the closed  heating chamber. Therefore, the LVDTs were placed on the clamping. The measurement data therefore contained the deformation of the specimen in and outside the heating area and of the area in the clamping. Therefore, the recorded deformation cannot be generalized and can only be used to compare the deformations in this test program ( Figure 6).

| Bond test setup
There are a large number of tests that can be used to determine the bond behavior of carbon reinforced concrete. Essentially, these are the two end anchoring tests, 15 the overlapping test 16 and the single-sided pull-out test (SPO) 17 with its different optimizations. 18 The end anchoring test is not suitable for elevated temperature, because the clamping of the bond area has to be heated. The overlapping test is a possible alternative, but correct documentation of the deformations is difficult, because of the shuttered heating chamber. Thus, only the SPO remains as feasible.
For this purpose, an odd number of rovings was inserted into the test specimen, in this case three. Only one roving was tested to determine the bond properties. The other two rovings stabilized the test specimen and represented the load-bearing effect of the twodimensional reinforcement. The bond length of this test specimen was defined by a saw cut dividing the two rovings, which are not tested, and a saw cut at the end of the bond length of the tested roving, see Figure 7, and for more details see References 17,19. During the SPO test it is possible to measure the pullout force and opening of the pull-out crack with LVDT on the load introduction which is caused by the pull-out of the roving from the concrete, see Figure 7-provided that no further cracks have occurred in the specimen apart from the predetermined breaking point (saw cuts) in the specimen.
The challenge or risk of this test is the tendency of the reinforcement material to split the concrete specimen. During manufacturing process, the reinforcement is flattened and deformed, which happens perpendicular to the reinforcement plane and leads to the typical concrete cracks in the reinforcement plane. 10 This phenomenon was a major challenge, especially in the reference tests at room temperature, because the high rigidity of the reinforcement caused a high transverse tensile stress occurring during the extract. With the decrease in the stiffness of the impregnation material exposed to temperature, the transverse tensile stress decreased during extraction and the roving pulled out of the concrete.

| CHARACTERISTICS OF THE IMPREGNATION
On the one hand, the impregnation had a significant influence on the internal bond within a fiber strand and the bond between the roving and the surrounding cement-based matrix, and on the other hand on the handling of the reinforcement in the processing procedure and also on the tensile load-bearing capacity in general. Impregnations of fiber reinforcements for concrete are nowadays mainly based on epoxy resins, acrylates or styrene-butadiene. Various tests can be carried out to determine the properties of polymers under thermal exposure. The dynamic mechanical thermal analysis (DMTA) and the thermal gravimetric analysis (TGA) are the most commonly performed tests. Another option is the differential scanning calorimetry (DSC), which is not considered here. The DMTA and the TGA were carried out on both carbon reinforcements.
First, the principle of DMTA will be described. The glass transition temperature T g is an important parameter for the characterization of polymers under thermal exposure. It marks the transition between a solid and a rubbery state. The change in properties takes place within a temperature range and not at a specific temperature value. Nevertheless, discrete characteristic values are usually derived from DMTA curves, but these can differ from standard to standard, for example, see . For the presented tests, which were carried out at the PYCO Institute, 23 a torsional load was used for load application of the fiber strand. Figure 8 shows the DMT analysis of the solidian reinforcement (MC 1). The development of the curve for the storage modulus G 0 shows a typical behavior with a pronounced plateau at the beginning and a plateau after the T g , for example, see Reference 24. The orange curve in Figure 8 presents the development of the loss factor δ, which describes the ratio between the loss module G 00 , the lossless imaginary part, and the storage module G 0 , the loss-bearing real part. The procedure for determining the T g according to the various standards, can be understood from the marked values and from Table 1. The possible glass transition temperatures range from À33 to 115 C. The À33 C according to DIN 65583 20 indicates the first possible decrease of the material properties by the decline of 2% at the beginning of the test. According to ISO 6721-11, 22 the greatest change in stiffness occurs at 115 C, at the maximum of the loss module, leading to a decrease in tensile and bond strength.
The V. Fraas reinforcement of the second MC shows a different behavior in the DMT analysis. The torsional stiffness has a major drop and so the tensile and bond strength decrease more in this temperature range, see Figure 9. At a temperature of À33 C, the torsional stiffness decreases about 2%, see Table 1. From a temperature of 99 C, the stiffness drops to zero. The impregnation completely loses its rigidity when the glass transition temperature is exceeded. In Table 1, the DMTA values for both carbon fiber grids are summarized.
In addition to the DMTA, the thermogravimetric analysis (TGA) was carried out. This analysis shows at which temperature a decomposition process of the material starts which leads to a loss of mass of the sample. In this test, the oxidation reaction of the carbon fibers could be observed when the test was carried out under atmospheric conditions. The study was conducted at the Leibniz Institute of Polymer Research Dresden. 25 The TGA of the reinforcement solidian Q95 shows that the impregnation consists of several components, which is shown by the quantity of maximums in the green curve. These components are decomposed one after the other within a temperature range between 200 and 700 C, see Figure 10. Beginning at a temperature of 700 C, the oxidation reaction of the carbon fibers begins. A tensile strength cannot be expected onward from this temperature. The impregnation of the reinforcement material of the MC 2 from the company V. Fraas consists of only a few components which is shown in the TGA. The impregnation was decomposed within a temperature range between 200 and 400 C, see Figure 11. Above a temperature of 500 C, the carbon fibers begin to decompose. Because of this fact no significant strength can be expected from this reinforcement over 500 C.

| Material behavior
At room temperature, an average short-term tensile strength of 2972.6 MPa (solidian) resp. 3128.1 MPa (V. Fraas) was determined (series with 34 resp. 59 samples), which served as a reference value for the temperature tests. To describe the material behavior of the two MCs, one exemplary test series of each combination at 400 C is shown in more detail.
The tensile strength of MC 1 exposed to 400 C is in the range between 1500 and 2000 N/mm 2 and is thus already 35% lower than the tensile strength at room temperature. At a temperature of approximately 300 C, cracks occurred in the plane of the reinforcement, see Figure 12a. This can be caused on one hand by the different coefficients of the thermal expansion of carbon and concrete, and, on the other hand, the vapor pressure due to the evaporation of the impregnation of the reinforcement. With further increased load, the crack in plane of the reinforcement opened and the concrete pieces were stripped from the reinforcement. Because of the decreasing bond under elevated temperature, only some cracks occurred in the heating area. The tensile strength of the second MC considered at 400 C lay between 1400 and 1700 N/mm 2 , that is, only 50% of the tensile strength at room temperature. In Figure 13, can be seen that no cracks occurred during the heating process and the subsequent loading. The cracking already took place during the preloading of the test specimen of MC 2, because this composite is primarily intended to be used for the strengthening of reinforced concrete structures. When looking at the test specimens after the failure, it is noticeable that the carbon fibers have a hair-like structure and thus the impregnation no longer has any effect and has already completely evaporated according to the results the TGA.
In Figure 14, the failure stresses of the stationary and the transient tests of both MCs are shown. Both materials show a clear decrease in tensile strength with increasing temperature and independent of the test procedure. So, the tensile strength under temperature exposure can be determined with both tests. The first significant strength decrease occurs at a temperature of 100 C. This is due to the glass transition temperature being exceeded, which reduces the rigidity of the reinforcement material and its tensile strength. The tensile strength of the MC 1 decreases further after a temperature of 200 C. From 500 C on, the strength drops sharply because the oxidation reaction of the carbon begins. In contrast, the strength of MC 2 drops particularly sharply in the range between room temperature and 200 C and then falls only slightly up to a temperature of 600 C.
There were efforts to predict the tensile behavior of the material considering the chemical aspects of the impregnation of the reinforcement material. 6,26 As has already been shown, the glass transition temperature had a great influence on the tensile strength under temperature exposure, but the temperature range, which is strength relevant, cannot be covered by the DMTA. The TGA described a larger temperature range. In Figure 14, F I G U R E 1 2 Stationary tensile tests of material combination 1 exposed to 400 C (graphic: K. Holz) compiled by Reference 7.
F I G U R E 1 3 Stationary tensile tests of material combination 2 exposed to 400 C (graphic: K. Holz) compiled by Reference 7. the decrease of the mass of the impregnation is compared to the decrease in tensile strength due to high temperatures. It turns out that the two parameters did not correlate with each other. While looking at MC 1, there was no change in strength when the impregnation evaporates at approx. 400 C. This phenomenon was also visible in MC 2. Here, a constant failure load level of circa 50% of the tensile strength at room temperature (transient tests)-that is, at about 1500 MPa-was reached in a temperature range between 200 and $550 C. There was no change due to the evaporation of the impregnation, which took place at temperatures between 250 and 400 C according to the TGA. The complete loss of strength of the reinforcement was only initiated when the oxidation reaction of the carbon fibers began. This temperature is shown in the TGA.
To classify the experimental results of this article, they were compared with experimental results found in the literature 6,12,27-33, see Figure 15. For this comparison, only carbon fiber reinforcements were considered. The state of knowledge was presented in more detail in References 5,7; only the values are given here. The values show the wide scattering of the tensile strengths under temperature exposure. The experimental data of the investigation described in this article are in the scatting range of the tensile strength. The decrease of the tensile strength of MC 1 and 2 is stronger than for the other shown material. Above a temperature of 400 C this effect has no influence because the impregnation was decomposed, and the oxidation reaction is dominating the material behavior.

| Design values
In order to be able to calculate structures based on the investigations, design values must be determined. For this purpose, the probability distribution for the tensile strength at room temperature has to be defined. The F I G U R E 1 5 Reduction of the tensile strength with increasing temperature (graphic: K. Holz) compiled by References 5,7 tensile strength of carbon reinforcement can be assumed to have a normal distribution, compare for example, References 34,35. It is further assumed that the tensile strengths at room and high temperature are subject to the same probability distribution. This was examined using the data of MC 2 at a temperature of 80 C; this data were also normally distributed.
According to Eurocode 2, 2 the material parameters under fire exposure are subjected to a partial safety factor of 1.0. Characteristic values can therefore be used for the calculation, this means 5% quantiles of the test values per temperature. The database for the individual temperatures is relatively small, the characteristic values have been adapted to Eurocode 0. 36 With this method, the uncertainties of small test series can be compensated.
According to this method, the following reduction values related to the tensile strength at room temperature can be seen in Figure 16. It is evident that the reduced values are significantly below those of steel. However, when facing at the absolute tensile strengths, the picture turns and the carbon strengths are higher than that of reinforcement steel.

| Material behavior
Thus, the type of failure changes with the increase of temperature. At room temperature and at a temperature at MC 1 of 100 C and MC 2 of 80 C, the test specimen failed due to the tensile failure of the concrete in the SPO tests. When the temperature or rather the glass transition temperature was exceeded, the specimen failed due to bond failure.
In order to compare the bond behavior of the two MCs, the 200 C bond tests are exhibited in Figure 17. The bond flow is shown here, which represents a bond force normalized to the bond length of the specimen. 17,19 The bond flow of MC 2 is significantly lower compared to MC 1. For the MC1 more tests are presented because two different types of SPO-tests were used. The reference tests of MC 1 show a higher maximum bond flow of 117 N/ mm compared to MC 2, which has a maximum composite flow of 58 N/mm. Nevertheless, the decrease in the bond flow is approx. 90% when exposed to a temperature of 200 C.
The reduction of the bond flow is temperature-dependent, and both materials show a strong decrease of the bond characteristics, see Figure 18. If the glass transition temperature is exceeded at approx. 100 C, the bond strength of both MCs approach zero.
To evaluate the influence of the glass transition temperature on the bond characteristics, the reduction in the torsional stiffness of the roving was compared to the maximum bond flow, see Figure 18. As shown in the diagram, transferability between the decrease in stiffness from the DMTA and the reduction in bond properties matches. That is why the bond characteristics can be estimated by considering the reduction of the stiffness in the DMTA.
To classify the carried-out tests, comparable tests from the literature were compiled in this study. The database is very small so other fiber materials besides carbon were taken into account. It turns out that the polymerimpregnated reinforcements in the literature 32,37-41 show a very similar bond behavior under temperature. This is due to the fact that the behavior of the polymers exposed to high temperatures is comparable. Only steel and the mineral coated carbon reinforcement show an improved bond behavior. The ribbed reinforced steel demonstrates a similar behavior as the mineral impregnation. For further investigations of the bond behavior exposed to high temperatures, a mineral impregnation of the carbon fibers should be considered as the bond strength of polymers under temperature is insufficient ( Figure 19).

| CONCLUSIONS
For the approach of carbon reinforcement exposed to high temperature, the glass transition temperature is an essential parameter and has to be determined as it influences the tensile and the bond strength. A clear definition for determining the glass transition temperature is important for the use of this material exposed to temperature to be able to determine application limits. The TGA is indispensable for estimating the upper temperature limit of the tensile strength because it allows to determine the temperature from which point on the oxidation reaction starts.
For both MCs, stationary and transient tensile tests were carried out. The results show that there is no difference in tensile strength between the two test types. Depending on the desired information which will be needed, the type of experiment can be freely selected.
The bond properties under temperature exposure are completely dependent on the characteristics of the impregnation, especially the glass transition temperature. This is particularly evident in the fact that the decrease in bond properties of both MCs can be described by the decrease in the torsional stiffness using the DMTA.  In order to improve the material behavior of carbon reinforcements exposed to high temperatures either temperature-insensitive polymers or mineral impregnations must be considered. There is still a great need of research required. Preferably, the carbon fiber should be protected from oxygen to slow down or prevent the oxidation reaction to positively influence the tensile strength under high temperature exposure.