Modelling and experimental analysis of high-temperature effect on rubberized concrete

The study presents results concerning the comparison of modelling and experimental work on rubberized concrete exposed to high temperatures. Crumb rubber was used in different concentrations to replace coarse aggregate. Experiments were made on flat-cylindrical specimens extracted from larger samples in order to be analysed. Simulations have been employed to calibrate measurements and to further analyse the specimens in an effort to prove their worth as a viable mechanism in studying larger-scale phenomena.


Introduction
The road transportations development has generated an increasing number of tyres; the statistics (in countries with heavy traffic) indicate that a number of tyres equal to the number of the population is used/replaced annually.In Europe, annually, are produced about three million tyres and 70% of them are discarded when no longer used.Thus, the number of used tires increases over time generating a serious negative influence on the environment.
One of the environmental sustainability's principles frequently claimed in Civil Engineering domain is the reutilisations of the construction waste in the same domain.It could prevent the environmental pollution and to contribute to the building's rising with lower costs.The use of crumbed rubber grains as an aggregates replacement in concrete has an important subject of research about 20 years ago [1].Many countries avoid storing or disposing of used tyres waste in deposits areas, providing an important stimulant to seek new manners to reuse [2].The impact of using recycled rubber as a partial replacement of the aggregate in concrete's structure has long been documented by compression, tension and bending tests [3,4,5,6].
The fire test involves the fire resistance of the structural components and the reaction to fire of flammable components [7].This could be important in the case of different ignition of stone aggregates and replacement of rubber aggregates (this paper shows).
Testing a new fire resistance material can be expensive (as a configuration for large-sized specimens).That's why validating a fire test model becomes important to be considered for future research projects.
Subsequently, have been casted new specimens, in which concrete's initial composition was modified by replacing 40%, 60%, respectively 80% of the total volume of fine aggregate 0-4 mm with rubber granules.The density of the rubber granules was similar to those of the replaced fine aggregate (sort 0-4 mm).For high temperature testing, flat-cylindrically specimens were taken from initial samples (by mechanical cutting), having 30 mm heigh and 100 mm diameter (Figure 1b).The study regarding the high-temperature behaviour of rubberized concrete specimens consisted of two main parts.In the first instance, a high-temperature exposure experiment was performed on all specimens.The next part consisted of simulations that were firstly focused on calibration (based on experimental data) and then used for comparing the behaviour of the specimens.

Experimental assessment
The experiment was performed using a metal stand, housing the heat source and the specimen, as sketched in figure 2. Heat was produced by an electrical air-heating gun, deemed capable of reaching temperatures of maximum 550ºC.The distance between the specimen and the gun was set to 1 cm.Temperature was measured by means of a FLIR BCam thermal imaging camera, located far enough as for it not to be affected by heat.The experiment was started once the temperature provided by the heat source was measured to be constant.This assessment was done using a K-type thermocouple positioned in the proximity of the air-gun's exit.
The temperature was measured on the side opposite to the heated one, in its center-point.Data was recorded every minute, for a total of 10 minutes.

Simulations assessment
For the purpose of finite element method (FEM) modelling, it was used the free version of LISA 8.0 software.This allows maximum 1300 nodes, which is enough for the types of specimens employed in the study.Each concrete cylinder was modelled beginning with a circular mesh of 107 nodes / 96 faces.On that, the extrusion was applied to achieve the specific thickness of each individual specimen.For this, a total of 10 subdivisions was used, thus obtaining 1177 nodes and 960 elements (figure 3).The heat source was modelled starting from a central surface on the face corresponding to the cylinder's exposed area, with a diameter of d = 2.5 cm.Through the extrusion process, an object was created out of this surface, and it was provided with a constant, desired, temperature (figure 4).The model also assumed the presence of convection on the lateral surface of the cylinder and on the face opposite to the heated one.For this, the following parameters were used: Tair = 22ºC and transfer coefficient (h) = 5 W/(m 2 •K).Values for density are quite predictable, since rubber is lighter than river coarse aggregates.The thermal properties measurements reveal that, although we would expect a similar linear variation with rubber concentration of both λ and cp, the heterogenous nature of the specimens results in a different behaviour.
Figure 5 hold the plots for thermal properties previously measured.The results do not, necessarily, highlight the properties of the recipe as a whole, but rather the local characteristics of the larger specimen.However, this is important from a practical point of view, since local properties are of great importance, for example, when an element experiences a thermal load.

Experimental data
Experimental data was plotted against the exposure time and designated as specimen_exp (figure 6).
Using rubber as aggregate replacement results in higher thermal inertia.This can clearly be seen for RC40 and particularly for the 80% specimen.The RC60 cylinder does not follow the same trend, which is consistent with the thermal properties determined beforehand.In this case, the temperature increase with exposure time is close to the PC variation up to 115ºC / 6min, when the curves start to diverge.After this point, the RC60 specimen's temperature tends to resemble the measurements for RC40.Having similar experimental data in the presence of the setup described above, as is the case of PC vs. RC60 (and, to a smaller extern, RC40), implies the need for corrections.Nevertheless, these results are important in order to test the viability of simulations, which will be further employed for comparing the specimens used in the study.

Simulations results
Models were designed for each specimen exposed to high-temperature.The input data needed for computing the transient thermal analysis consisted of: ρ, λ and cp.These are known quantities, precisely measured at the start of the investigation.The first aim of the simulations was validating the experimental work, which is regarded as accurate with the exception of the exposure temperature.Therefore, simulations were calibrated to fit the experimental data by changing the input temperature, to the point of obtaining satisfactory convergence of experiment and simulation.The following values in table 2 were found to provide suitable results.Experiment vs. simulation results when using calibrated temperatures Variation of temperatures with exposure time are presented in figure 7.Each graph holds the plots for the experimental data and the simulated results to match them.All cases demonstrate that the correlation between the two curves is acceptable, with the lower temperature regime somewhat more distorted.The simulation data converges towards the same curve as the experimentally determined one, but the initial ramp-up of the heating in the experiment might be prone to more errors due to more complicated convection phenomena than the ones introduced in the simulation.Nevertheless, results are proving the simulation to be effective, thus the final process was employed.
Using the same models as before, with only changing the input temperature to 500ºC for all specimens, the accurate comparison between PC, RC40, RC60 and RC80 was pursued.It can be observed that results are consistent to the differences in the calibration temperatures.The experimental errors caused RC60 high-temperature exposure to mimic firstly the PC's behaviour and tend to the RC40's one afterwards is not present here.Thus, by simulation, the proper curve was plotted.The temperature-time curves for PC, RC40 and RC60 are quite similar in values, in both experiment and simulation.From a measurement perspective, such results need very small-error experimental setups to be obtained.

Conclusions
The paper investigates the interplay between experiment and simulation in the case of high-temperature exposure of concrete and rubberized concrete.Since fire testing is extensive and expensive, alternative routes by studying smaller scale specimens with viable results is an important matter.If successfully implemented, such alternatives could result in great cost-cutting and a wider experimental reach in terms of mix recipes and new materials.

Figure 1 .
Figure 1.Cylinder-shape concrete specimens a) initial and b) cut samples

Figure 2 .
Figure 2. Set-up for experimental stand a) photo and b) sketch

Figure 3 .
Figure 3. Finite element model: a) circular cross-section and b) 3D

Figure 4 .
Figure 4. Specimen with heat source model

Figure 5 .
Figure 5. a) Thermal conductivity and b) specific heat variations of mortar specimens

Figure 8 .
Figure 8. Simulation results when using the same input temperature (500ºC)

Table 1 .
Parameters of specimens used for simulation.