Nondestructive Testing to Identify Delaminations Between HMA Layers, Volume 2 - Theoretical Models (2013)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

1Theoretical Models for Ground-Penetrating Radar The numerical simulations presented in this chapter were carried out by Infrasense with support from Dr. Kim Belli of Northeastern University in Boston, Professor Dennis Hiltunen of the University of Florida, and Professor Rajib Mallick of Worcester Polytechnic Institute. Numerical simulations of nondestructive testing (NDT) techniques were carried out to assess the NDT ability to detect delaminations and to evaluate the most promising configura- tions for implementing each NDT method. An NDT tech- nique and/or configuration that does not show promise in the numerical simulations would likely not succeed in the field evaluations; therefore, results obtained from the simu- lations would help the project team focus on the most prom- ising methods and/or configurations in the laboratory and field evaluations. The numerical simulations were conducted by using existing numerical models to simulate and compare simu- lation results of delaminated and intact pavements. The simulation information was used to define detectability and served to support recommendations of specific meth- ods and/or configurations for the laboratory and field evalu- ations. Modeling provided insight into results that could be expected from laboratory and field experiments, and it offered guidance into the effective configurations of the transmitter (T) and receiver (R) for each NDT technique. Modeling results were considerably cleaner than results obtained in the labo ratory and field evaluations, and the modeling results were based on certain idealizations of material characteristics and input signals. Simulation results showed (a) whether, under ideal conditions, the pavement defect can be detected and (b) under what conditions detect- ability can be enhanced. For electromagnetic and mechani- cal wave methods, the simulation results could show how the relative placement of the source and receiver could affect the NDT detectability. Simulation Overview A time-domain GPR antenna is typically used in highway and bridge evaluations. Therefore, Maxwell’s time-dependent curl equations were used to model wave propagation in this GPR simulation. The simulation was conducted via a two- dimensional (2-D) finite difference time domain (FDTD) algorithm. The FDTD method explicitly discretized Maxwell’s equations by using the finite difference approximation for the computation of differential equations. The electric field measurements from the 2-D FDTD simulation were analo- gous similar to the output from the GPR antenna. The FDTD method was also able to handle inhomogeneous materials of asphalt concrete. To simulate GPR investigation of a pavement in two dimen- sions, the transverse magnetic (TM) mode was chosen with wave propagation in the x-y plane (TMz). This mode had no magnetic field in the direction perpendicular to the cross- sectional geometry (z-direction). This perpendicular direction was assumed to be the polarization direction of the antenna. Simulation Parameters All GPR simulations were carried out by using a 2-D FDTD code. Discretization of the physical model was done at a reso- lution of 1 mm (0.04 in.) so that the thin delamination could be adequately modeled. The time step was selected to be 2 ps to meet stability requirements. The signals used to excite the models are shown in Figure 1.1. The modulated Gaussian pulse was centered at either 1.3 GHz or 2.6 GHz, and the –3dB bandwidth was chosen to be equal to the center frequency. GPR Simulations for Intact Pavement Typical 203-mm (8-in.) thick asphalt geometry (with no defects) is shown in Figure 1.2. Asphalt was assumed to have a dielectric C h a P t e r 1

2constant of 5, and it was not considered to be conductive. The geometry included a layer of air under the asphalt. This layer was expected to produce a higher reflection from the bottom of the asphalt layer than if it was over another medium (such as concrete or granular base). However, because the purpose of this simulation was to determine the reflection from the defects within the asphalt layer, it should not affect simulation results. Note that the T and R locations were assumed to be at the surface of the asphalt. It was understood that configurations to be considered in this program would include air-launched antennae, but that the ground-coupled configuration would adequately represent the response for modeling purposes. In the field, data would not be collected at every point in the structure. However, this information was available in a computational model and could be very helpful in determin- ing what was going on in the responses recorded at the receiv- ers. For the intact pavement, images of the electric field in the model at four different times are presented in Figure 1.3. The black horizontal lines indicate the boundaries of the asphalt pavement. Note that the 500-mm axis location in Figure 1.3 corresponds to the 0-cm transmitter location in Figure 1.2. Move-Out Simulations The move-out simulations were conducted to provide insight into (a) whether a defect was expected to be seen under ideal conditions and (b) how the reflection from the defect was expected to be observed at different transmitter and receiver configurations. For modeling purposes, a ground-coupled antenna configuration was represented, with two alternative vertical positions of the transmitter and receiver antennae: one directly on the asphalt surface, and the other elevated 3 mm (0.12 in.) above the asphalt surface. The simulations showed lit- tle difference between these two cases. Since the 3-mm (0.12-in.) elevation was more representative of an actual ground-coupled antenna, the results presented later were for that case. Move-out simulations were conducted by assuming a fixed transmitter location and a variety of receiver locations mov- ing out from the transmitter at 10-mm (0.4-in.) intervals. The defect condition geometries for the move-out simula- tions are shown in Figure 1.4. While the transmitter (+) and each of the receiver (o) locations were shown on the surface of the asphalt, the results presented were for the case where the antenna was 3 mm (0.12 in.) above the surface, as pre- viously discussed. As shown in Figure 1.4, three delamina- tion cases were simulated, including (1) 2-in (50-mm) deep, 1-mm (0.04-in.) thick delamination; (2) 4-in (100-mm) deep, 1-mm (0.04-in.) thick delamination; and (3) stripped asphalt layer between 3-in (75-mm) and 4-in (100-mm) depths. In Figure 1.1. Simulation excitation signals (3dB bandwidth equal to center frequency). Figure 1.2. Simulation geometry for intact pavement: 3-mm-raised antenna is shown on the surface.

3 Figure 1.3. Electric field (E-field) throughout the modeled intact pavement at four given times for the surface 2.6 GHz excitation. Figure 1.4. Move-out simulation geometries for (a) 2-in.-deep, 1-mm (0.04-in.) thick delamination; (b) 4-in.-deep, 1-mm (0.04-in.) thick delamination; and (c) stripped asphalt between depths of 3 in. and 4 in. Note: The defect runs through the entire cross section of the model. (a) (b) (c) those cases, the dielectric constant of the stripped layer either was 3.5 (containing more air) or 9.0 (containing more water). B-Scan Simulations Instead of considering the fixed transmitter and moving receiv- ers as in the move-out simulations, for the B-scan simulations, the distance between the transmitter (T) and receiver (R) was fixed, either 125 mm (5 in.) or 40 mm (1.6 in.), and they were moved across the pavement. This is the typical configuration for a single ground-coupled antenna that is moved across the surface of the pavement, and the resulting display is gener- ally referred to as a B-scan. This configuration resulted in a single simulation run for each T/R location, and one run

4was carried out every 10 mm (0.4 in.) across the pavement. The T/R pair was located 3 mm (0.12 in.) above the surface. The B-scan geometries for intact and stripped pavements are shown in Figure 1.5. The initial and final T/R midpoints are shown in Figure 1.5a. Note that the T or R locations would change depending on whether the separation was 125 mm (5 in.) or 40 mm (1.6 in.), but all of the discussions would consider the midpoint. Since the geometry was symmetric, simulations were only run until the midpoint of the defect (100 cm or 39.4 in.) and then mirrored to construct the final B-scan. This procedure resulted in 101 simulations per case. Simulation Results Move-Out Simulations Figure 1.6 shows the response recorded at each R for the intact and 2-in. delamination cases for the 2.6-GHz antenna model. The horizontal axis was the distance across the pavement from R to T, which was fixed and never moved. The response at approximately 1 ns, for the R located zero mm from the trans- mitter, was due to a direct signal and the reflection from the surface of the asphalt. The response at the same horizontal location and approximately 4 ns was the reflection from the bottom of the asphalt. As previously discussed, the asphalt was floating in free space so this bottom reflection was stronger than would be observed in the field. Since the most promising results were obtained by using the 2.6-GHz antenna model, these results are presented later in this report. Figure 1.7a shows the 2.6-GHz antenna simulation for the case of the delamination at 4 in. The effect of the delami- nation at 4-in. depth shows up deeper (later in time) than does the effect for the 2-in.-deep delamination (Figure 1.6b). Figure 1.7b shows the Figure 1.7a results with the intact data of Figure 1.6a subtracted. The subtraction of the intact data highlights the effect of the delamination. Figure 1.7b shows that the amplitude of the reflection from the delamination increases as the receiving antenna is moved from Figure 1.6. Move-out simulation results for (a) intact case and (b) delaminated at 2 in. (a) (b) reflection from delamination AC bottom reflection AC surface wave air wave Figure 1.5. Example of B-scan simulation geometries for (a) intact case and (b) stripped asphalt 3–4 in. deep. Note: Antenna is shown on the surface and geometries are similar for 3-mm raised antenna. Delamination geometries are not shown because at this scale, the representation of the delamination is difficult to see. Vertically, geometries are similar to those in Figure 1.4a and b, and, horizontally, to that in (b). (a) (b)

5 100 to 200 mm (3.9 to 7.9 in.) from the T and then decreases. This pattern suggested that there is an antenna spacing that optimizes the measured delamination response. In order to examine this effect in further detail, the reflection amplitude versus spacing is plotted in Figure 1.8. The figure shows that the optimum antenna spacing ranged from 100 to 170 mm (3.9 to 6.7 in.) for delamination depths ranging from 2 to 4 in. Figure 1.9 shows the simulation results for the stripped asphalt layer between 3 and 4 in. The reflection from the stripped layer had two elements: the reflection from the top and the reflection from the bottom of the stripped layer. B-Scan Simulations Sample B-scan simulation results are shown in Figure 1.10 for the intact, delamination, and stripped conditions. The results show that the delamination and stripping condi- tions are detectable in principle in the B-scan simulation model. Figures 1.10e and 1.10f show how the presence of moisture in the delamination significantly increased the response. Conclusions from Electromagnetic Simulations In all cases, the 2.6-GHz antenna provided better results than those of the 1.3-GHz antenna. Both frequencies pro- vided adequate resolution down to the bottom of the deck (203 mm, or 8 in.). Changing the height of the T/R pair from zero to 3 mm (0.12 in.) above the surface seemed to have little impact on the results. As expected, water-filled delaminations were much more evident in the simulation results than were air-filled delaminations of the same size. The move-out simulations showed that a T/R separa- tion (as opposed to a monostatic configuration) was often desirable. However, the ideal separation was governed by a variety of factors, including depth of damage and the electromagnetic properties of the materials, both of which affected the angle of reflection. For this reason, an array con- figuration with a single T and multiple Rs may be beneficial. From the move-out results, it appears that spacing between 100 and 170 mm (3.9 and 6.7 in.) is optimal for detecting delaminations. Figure 1.7. Move-out simulation results for (a) delamination at 4 in. and (b) delamination at 4 in. with intact results subtracted. reflection from delamination reflection from delamination (enhanced) (a) (b) Figure 1.8. Amplitude versus move-out distance.

6reflection from stripped layer reflection from stripped layer (enhanced) (a) (b) Figure 1.9. Move-out simulation results for (a) stripping (dry) between 3 and 4 in. and (b) stripping (dry) between 3 and 4 in. with intact results subtracted, (continued on next page) delamination (b) bottom of AC direct T-R (a) Figure 1.10. Sample B-scan simulation results for 2.6 GHz antenna model: (a) stripping (dry) at 3 to 4 in., (b) intact case.

7 bottom of AC direct T-R delamination delamination (c) (d) (f)(e) Figure 1.10. Sample B-scan simulation results for 2.6 GHz antenna model: (c) dry delamination at 4 in., (d) dry delamination at 2 in., (e) wet delamination at 4 in., and (f) wet delamination at 2 in. (continued).