International Symposium (NDTCE 2003) NonDestructive Testing in Civil Engineering 2003  
Start > Contributions >Posters > Thermography: 
Active Infrared Thermography in Civil Engineering Quantitative Analysis by Numerical SimulationG. Wedler, A. Brink, Ch. Maierhofer, M. Röllig, F. Weritz and H. WiggenhauserFederal Institute for Materials Research and Testing (BAM), Berlin, Germany ABSTRACTThe coolingdown process of building structures after heatingup with an external radiation source was analysed to detect voids inside and below the surface. Quantitative results of concrete test specimen containing voids with different sizes at various depths will be presented here. The experimental results were compared to numerical simulations performed with a Finite Difference program developed at BAM. 1 IntroductionActive infrared thermography (IRT) has established as a fast and reliable tool in many areas of non destructive testing (NDT). It is well known for material testing in several industry branches for the detection of voids and delaminations [1]. Because of its small penetration depth and slow velocity of propagation "Thermal Waves" normally are used for thin materials with high thermal conductivity (e.g. metal plates). As concrete has a very small thermal conductivity the application of infrared thermography in civil engineering (CE) was mostly limited to passive investigations of the quality of thermal insulation of building envelopes. Further developments and applications in civil engineering are using the sun as a natural heat source. Examples are inspections of bridge decks [2] and of paving in general [3]. Nevertheless, in recent years a lot of investigations have been done to apply active thermography to civil engineering (CE). As shown recently there are certain CE applications where a customized active IRT is reasonable [4][5][6][7][8]. The application of impulse thermography (IT) as a quantitative nondestructive testing method in civil engineering [9][10][11] is investigated here. For locating delaminations at bridge decks with IT an ASTM standard already exists [12]. Furthermore, the technique is intended for the detection of manifold nearsurface inhomogeneities and common defects subsurface in typical building structures. Practical problems like locating voids and honeycombing in concrete, delaminations of plaster at concrete and masonry, delaminations and voids behind tiles on concrete embedded in mortar, interconnection of carbon fibre reinforced laminates glued on concrete and delaminations of bituminous sealing on steel are analysed. The measurements presented here were performed by external heating of the specimen surface and further investigation of the cooling down process. The heating up pulse causes a instationary heat flow, describable as propagation and attenuation of thermal waves. The propagation of the thermal wave strongly depends on the material properties like thermal conductivity, heat capacity and mass density. Due to anomalous thermal properties inhomogeneities in the structural element affect the uniform heat flow and thus the surface temperature distribution. While observing the temporal changes of the surface temperature distribution with the infrared camera near surface inhomogeneities will be detected. The differences between temperature transient curves at surface positions above nondefect regions and above inhomogeneities include information about defect parameters like depth, lateral size and the type of material. In this paper the results of investigations of concrete test specimen containing voids with different sizes at various depths will be presented. The experimental data are compared to results of simulations using a Finite Difference program based on the Fourier equation. The program was developed at BAM and allows a realistic simulation of thermal heat flow through different materials. By means of parameter variation (depth, size etc.) and comparison with experimental data the crucial question of the amount of concrete coverage above defects, well known as the inverse solution, can be solved. 2 ExperimentalThe experimental setup to perform impulse thermography measurements is shown in Figure 1. It consists of a thermal heating unit (1), a structural element to test (2), an infrared camera(3) and a computer system (4) which enables real time data recording.
The thermal heating unit contains three infrared radiators, each with a delivery rate of 2400 W. The heatingup procedure is usually done dynamically by moving the radiators across the specimen surface to obtain the best possible homogeneous heating. Therefore, radiators are mounted in a line array and are moved automatically parallel to the surface at a distance of about 15 cm. The coolingdown process of the surface is observed with a commercial infrared camera (Inframetrics SC1000). The camera contains a focal plane array of 256 x 256 PtSi semiconductor detectors and is able to detect radiation from the surface of the specimen in a wavelength range of 3  5 mm. The measured radiation intensity values can be converted into temperature values using lookup tables and can be presented as a grey or colour scaled picture (thermogram). During data acquisition the thermal image data are transferred to the computer at a maximum frame rate of 50 Hz and a storage depth of 12 bits per pixel. After receiving the thermal images the computer is used to analyse the data with dedicated software programs.
Typical testing problems in civil engineering are voids and honeycombing in concrete elements. Therefore a test specimen with a size of 150×150×50 cm^{3} was built containing voids of different sizes (10×10×10 cm^{3} and 20×20×10 cm^{3}) and different amounts of concrete coverage, Figure 2. Measurements were performed by heating the surface of the specimen with varied heating times up to 60 min at a fixed heating power of 1250 Wm^{2}. After switching off the heating source the coolingdown process was monitored with the infrared camera and thermal images were recorded with a frame rate of 2 Hz up to 120 min. To improve the signal to noise ratio the average of 10 images was determined. Thus a film of 1440 images with a time step of 5 s between single images is received. Figure 3a shows a thermogram selected from a series of measurements with 15 min heating up time. The image was taken 8 min after switching off the external heating unit. Shallow voids (up to 6 cm coverage) are clearly distinguishable with a good thermal contrast. Deeper voids cannot be detected so far but appear clearly after waiting for longer times (e.g. 50 min).
The main approach in analysing the thermal data was to interpret the function of surface temperature versus cooling time for selected areas with and without inhomogeneities. These selected transient curves were compared and difference curves (between transient above a void and transient above a sound area) were calculated. In Figure 3b the results of a defect with an approximate concrete coverage of 3.5 cm are shown. The temperature difference curve usually has a maximum DT_{max} at a distinct time t_{max}, indicating the point of maximum temperature contrast. The values of DT_{max} t_{max} depend mainly on the depth of the void and the amount of heating up time. 3 Numerical SimulationNumerical calculations based on the Finite Differences method were carried out for further data interpretation and for solving the inverse problem. Based on the 3 dimensional (3D) differential equation of Fourier, the temperature T_{x,y,z,t }of each node of a 3D grid for the next time step Dt is given by the following equation:
with the mass density r, the thermal conductivity l and the heat capacity c_{p}In the extensive simulation program a thermal heating pulse hits the surface of the test specimen and the generated instationary thermal heat flux inside the specimen is calculated, whereby also surface radiation and convection losses have been considered. The result of the calculation is given by a 2D matrix containing the temperature distribution for each surface element at each time step during heating up and cooling down processes. This 2D matrix can be analysed similar to the experimental data. To compare the numerical results with the experimental data afterwards several voids in concrete with a sizes of 10×10×10 cm^{3} and 20×20×10 cm^{3} were simulated. Therewith the resulting temperature contrast for several heating up times at different defect depths up to 10 cm was calculated. Figure 4a shows an example of a temperature decay versus time above a simulated defect with 3 cm coverage. By means of an undisturbed reference point the temperature difference curve was calculated, containing a maximum difference temperature DT_{max} at a distinct time t_{max} similar to the experimental results in Figure 3b. To determine the dependence of t_{max} on defect depth and heating up time numerous calculations were done. In Figure 4b t_{max} versus defect depths for different heating up times are shown. Obviously, t_{max} depends on both the heating up time and the defect depth (t_{max} increases with increasing defect depth and decreasing heating up time). Noteworthy, at a defect depth of 1 cm and a heating up time of 60 min t_{max} is less than the switch off time. Here the temperature maximum occurs already during the heating up time, causal for this result.
To determine the influence of environmental conditions and measurement parameters on the results of impulse thermography a study of individual parameters were done. As a first result we could show that periodically heating with repetition rates of 30 s has nearly no influence on DT_{max} and t_{max}. Furthermore, varying of the ambient temperature does not cause any changes on DT_{max} and t_{max}. The amount of heat has a linear influence on DT_{max} but does not affect t_{max}. However, surface losses and material parameters should be known very well,because their effect cannot be calculated easily. Eventually, the main impact on DT_{max} and t_{max} refers to heating time and defect depth, whereas the variation of the heating time is significant less important than the variation of the defect depth.
For solving the inverse problem a comparison of numerically simulated and experimentally measured results were carried out. In doing so, the different values of t_{max} for several defects obtained experimentally were compared to numerically simulations by means of the time of maximum t_{max}, see Figure 4b. The result of this comparison is depicted in Table 1. Shown are the depth of this kind of inverse solution (column 5), compared to the nominal defect depth (column 3). However, the experience of producing concrete specimen shows that the real depth may be deviate significant from the nominal one. Thus additionally reliable RADAR measurements were done to determine the real depth [13]. The results are also shown in Table 1, column 4. Within the scope of measurement/calculation error the values of the computed depths match the RADAR values. Nevertheless, sporadic deviations of more than 20% occur. Systematically investigations were done also for the dependence of the amount of temperature difference DT_{max} on defect depth. The analysis accomplished similar to t_{max}, as mentioned above. Simulated values for DT_{max} were compared to measured ones and in conclusion a series of defect depth values can be presented. The results are shown in Table 1, column 6. Obviously, the results via DT_{max} match the RADAR values better than those via t_{max}. The deviation of the results by inverse solution may be caused by a series of circumstances. On one hand, a numerical model in general is simplified and restricted to finite elements. Further on, a realistic simulation of several material parameters is distinct difficult, e.g. the thermal conductivity of concrete, l_{concrete} Broadly, because of the multitude of different kinds of concrete a declaration of a single standard value of l_{concrete} is unrealistic. However, in particular t_{max} depends significant on l_{concrete}, in contrast to DT_{max}. Without much doubt that's one reason that the inverse solution via t_{max} fits the RADAR values much better then those via DT_{max}, as further investigations show. Beside the displacement of the defect compound during concreting, also the orientation of the cuboids relative to the test specimen has been changed distinct. Thus an undefined situation is available to obtain quantitative depth values for both, RADAR as well as IT. Therefore, shortly the test specimen will be destroyed to determine the real depths. 4 Conclusion and OutlookImpulse thermography can be used as a good tool for nondestructive testing in civil engineering. Locating defects in concrete with concrete covers up to 10 cm has been performed successfully. It is a very fast and efficient nondestructive method, which can be used contactless giving direct images of the surface. A quantitative analysis of the experimental data can be carried out. Numerical simulations are useful to investigate the influence of environmental conditions and material parameters and to determine the depth of real voids by comparing simulated and measured values (inverse problem). Further investigations with impulse thermography have shown that the method can also be applied for the detection of delaminations of surface layers (plaster on concrete and masonry, carbon reinforced laminates on concrete, cleaving tiles embedded in mortar on concrete) and for locating enhanced moisture in the near surface region [14]. A crucial improvement of active IRT is achieved by analysing the experimental data in the frequency domain instead the time domain by means of Fast Fourier Transform. This method is well known as Pulse Phase Thermography and is presented as the oral contribution "Investigation of Concrete Structures with Pulse Phase Thermography" at NDT/CE 2003. AcknowledgementsGratefully acknowledged is the fruitful cooperation with the group of Prof. Hillemeier at Technical University Berlin and the continuous support by the German Science Foundation (DFG, Projekt MA 2512). References

