Feasibility of Conventional Non-Destructive Testing Methods in Detecting Embedded FRP Reinforcements

: Fiber-Reinforced Polymer (FRP) bars/strands are the most promising alternative to their steel counterparts for reinforcing concrete elements due to their resistance to corrosion, lighter weight, higher strength and better durability. However, very limited research has been conducted in relation to non-destructive testing (NDT) methods that are applicable to damage detection in FRP bars or the detection of FRP reinforcements embedded in concrete. The ability to assess the condition of the relatively new and unique FRP reinforcements will increase the conﬁdence of the construction industry in their use as a reliable substitute for steel reinforcements. This paper investigates the ability of two of the most commonly used NDT methods, Ground Penetrating Radar (GPR) and Phased Array Ultrasonic (PAU), in detecting FRP bars/strands embedded in concrete elements. GPR and PAU tests were performed on two slab specimens reinforced with GFRP (Glass-FRP) bars, the most commonly used FRP bar, with variations in their depth, size and conﬁguration, and a slab specimen with different types of available FRP reinforcements. The results show that GPR devices can detect GFRP bars/strands and CFRP (Carbon-FRP) strands to some extent, and their detectability increases with the increase in their antenna center frequency. On the contrary, PAU is only capable of detecting GFRP and CFRP strands. The results of this paper also emphasize the need for further research and developments related to NDT applications to embedded FRP bars.


Introduction
Corrosion of steel reinforcement is one of the main problems in traditional concrete structures, which severely affect its safety and serviceability. Several corrosion protection measures, including but not limited to cathodic protection, epoxy-coated bars and galvanized steel reinforcements have been implemented in the past; however, they have only managed to delay corrosion rather than eradicating it [1]. The most promising alternatives available that could entirely stop the process of corrosion within concrete is the use of fiber-reinforced polymer (FRP) bars/strands [2][3][4][5][6]. Depending upon the types of reinforcing fibers used, the FRP bars/strands can be classified as GFRP (Glass-FRP), CFPR (Carbon-FRP), BFRP (Basalt-FRP) and AFRP (Aramid-FRP) bars, respectively for glass, carbon, basalt and aramid fibers. The FRP bars are resistant to all the elements that leads to corrosion in steel reinforced concrete (RC) structures, such as reduction of pH of the concrete from carbonation, chloride contamination of the concrete and diffusion of halides and chemicals [7,8]. The use of FRP bars for reinforcing new structures is even more preferable because they also have 1.5-2 times higher tensile strength than their steel counterparts [9][10][11]. In addition, although the initial cost of FRP bars is higher than that

Materials and Methods
In order to determine whether GPR and PAU can detect internal FRP reinforcement, small-scale concrete slabs were fabricated. Two slab specimens (i.e., labeled as J and L) were 30 in. wide by 30 in. long and 7 in. deep, and the third specimen (viz. Slab C) was 36 in. wide by 36 in. long and 5 in. deep. The concrete mix used to cast the slab specimens was the 'Class II 4500 Bridgedeck' concrete, as per the Florida Department of Transportation (FDOT). This class specified a guaranteed compressive strength of 4600 psi (31 MPa). Type II Cement was used with a water to cementitious material ratio (w/m) of 0.44, and #57 stone and silica sand was used as coarse and fine aggregate, respectively. To obtain the actual strength value; concrete cylinders were tested at 28 days according to ASTM C39 [94,95]. An average compressive strength of 31.70 MPa was obtained with a standard deviation of 0.69 MPa (coefficient of variation of 2.2%).
The slab specimens were fabricated, targeting different parameters such as FRP bars/strands type (GFRP, CFRP, BFRP), bar diameter, bar direction and bar depths. Table 1 shows the identification of the slab specimens by group, highlighting their main parameters. Because GFRP bars are the most commonly used FRP reinforcement in concrete elements, the first two slabs constructed (Slabs C and J) were only reinforced with GFRP bars of different sizes, at different depths and different configuration (bars in one direction and bars in two orthogonal directions, i.e., mesh). The concrete cover specified by ACI CODE-440.11-22 [96] for the concrete members reinforced with GFPR bars ranges from 0.75 in. to 3 in., hence the depth variations in the slab specimens were included to represent the layers of FRP reinforcement, which could be anywhere within the concrete cover range specified. Further, the third slab (Slab L) has different types of internal reinforcement (bars and strands) embedded into it, including one steel bar whose detectability acts as a control for this research. Having GFRP, CFRP, BFRP and steel bars/strands on the same slab specimen allows comparison of the detectability of different FRP bar/stands with the steel bar under the same test conditions. Table 1. Identification of small-scale concrete slab specimens.

Specimen Group Slab ID Bar Diameter No. of Bars
Slabs with GFRP bars in one direction Slab C #4 and #6 6 Slabs with GFRP bars in two directions (mesh) Slab J #6 10 Slab with different FRP bars in one direction Slab L #3, #5, #8 GFRP bars, #5 CFRP strand, #3 steel bar, #6 GFRP strand, #3, #5 BFRP bars 9 All specimens were fabricated and labeled following the same layout as shown in Figure 1. Every slab is identified with a letter (C, J and L), and every side has an identification number (from 1 to 4). The direction of the measurement will be determined by the number where the measurement is started to the end of the measurement. Appl

Materials and Methods
In order to determine whether GPR and PAU can detect internal FRP reinforcement, small-scale concrete slabs were fabricated. Two slab specimens (i.e., labeled as J and L) were 30 in. wide by 30 in. long and 7 in. deep, and the third specimen (viz. Slab C) was 36 in. wide by 36 in. long and 5 in. deep. The concrete mix used to cast the slab specimens was the 'Class II 4500 Bridgedeck' concrete, as per the Florida Department of Transportation (FDOT). This class specified a guaranteed compressive strength of 4600 psi (31 MPa). Type II Cement was used with a water to cementitious material ratio (w/m) of 0.44, and #57 stone and silica sand was used as coarse and fine aggregate, respectively. To obtain the actual strength value; concrete cylinders were tested at 28 days according to ASTM C39 [94,95]. An average compressive strength of 31.70 MPa was obtained with a standard deviation of 0.69 MPa (coefficient of variation of 2.2%).
The slab specimens were fabricated, targeting different parameters such as FRP bars/strands type (GFRP, CFRP, BFRP), bar diameter, bar direction and bar depths. Table  1 shows the identification of the slab specimens by group, highlighting their main parameters. Because GFRP bars are the most commonly used FRP reinforcement in concrete elements, the first two slabs constructed (Slabs C and J) were only reinforced with GFRP bars of different sizes, at different depths and different configuration (bars in one direction and bars in two orthogonal directions, i.e., mesh). The concrete cover specified by ACI CODE-440.11-22 [96] for the concrete members reinforced with GFPR bars ranges from 0.75 in. to 3 in., hence the depth variations in the slab specimens were included to represent the layers of FRP reinforcement, which could be anywhere within the concrete cover range specified. Further, the third slab (Slab L) has different types of internal reinforcement (bars and strands) embedded into it, including one steel bar whose detectability acts as a control for this research. Having GFRP, CFRP, BFRP and steel bars/strands on the same slab specimen allows comparison of the detectability of different FRP bar/stands with the steel bar under the same test conditions. All specimens were fabricated and labeled following the same layout as shown in Figure 1. Every slab is identified with a letter (C, J and L), and every side has an identification number (from 1 to 4). The direction of the measurement will be determined by the number where the measurement is started to the end of the measurement.   Table 2 shows the details and dimensions of the specimens. For each specimen, the distance to edge, depth to surface, bar diameter, bar material and slab depth are presented based on the convention shown in Figure 1. The formwork for each slab specimen constructed is shown in Figure 2.

GPR Tests
Ground penetrating radar (GPR) is a non-destructive testing (NDT) method that is used to the analyze internal characteristics of structures or elements in real-time. It works on the principle that electromagnetic waves reflect back when they encounter an interface between two materials with different dielectric constants [31,39]. GPR functions by emitting electromagnetic waves (in the form of radio waves) through the test material and then detecting the waves that bounce back off any discontinuities within the material. These discontinuities can take various forms, including interfaces between different materials, such as the concrete-bar interface shown in Figure 3 or the concrete-air/water interfaces (subsurface defects like voids, cracks, debonding and delamination) [97]. Due to the limited research on the use of GPR for FRP-reinforced structures, this method has not yet been able to establish itself as a reliable NDT method. This paper will explore the possibility of using GPR for FRP bar detection.
In this study, the GPR tests were conducted both as individual line scans along a straight line and as grid scans over a test area. The line scans were conducted as a reconnaissance survey or preliminary inspection prior to the detailed grid/area scan to form a quick idea of what to expect inside the structure, the orientation of reinforcement or other subsurface features, and the depth of exploration. The GPR line scans conducted along a straight line on the surface of the test specimen were used to obtain its cross-section image through a plane normal to the surface along the direction of the scan. However, going back and forth between several line scans and interpretating each line scan was time consuming and labor intensive. Thus, in order to simplify the interpretation of results, grid scans were also conducted over the surface of the test specimens. Grid scans were performed simply by taking GPR data along the straight lines of a grid over an area covering the test specimen, as shown in Figure 4. The grid required for a grid scan was arranged using a grid mat secured to the surface so that it did not move while taking the GPR data.  (c) (d)

GPR Tests
Ground penetrating radar (GPR) is a non-destructive testing (NDT) method that is used to the analyze internal characteristics of structures or elements in real-time. It works on the principle that electromagnetic waves reflect back when they encounter an interface between two materials with different dielectric constants [31,39]. GPR functions by emitting electromagnetic waves (in the form of radio waves) through the test material and then detecting the waves that bounce back off any discontinuities within the material. These discontinuities can take various forms, including interfaces between different materials, such as the concrete-bar interface shown in Figure 3 or the concrete-air/water interfaces (subsurface defects like voids, cracks, debonding and delamination) [97]. Due to the limited research on the use of GPR for FRP-reinforced structures, this method has not yet been able to establish itself as a reliable NDT method. This paper will explore the possibility of using GPR for FRP bar detection. Figure 3. Reflection of incident wave at concrete-bar interface [98]. In this study, the GPR tests were conducted both as individual line scans along a straight line and as grid scans over a test area. The line scans were conducted as a reconnaissance survey or preliminary inspection prior to the detailed grid/area scan to form a quick idea of what to expect inside the structure, the orientation of reinforcement or other subsurface features, and the depth of exploration. The GPR line scans conducted along a straight line on the surface of the test specimen were used to obtain its cross-section image through a plane normal to the surface along the direction of the scan. However, going back and forth between several line scans and interpretating each line scan was time consuming and labor intensive. Thus, in order to simplify the interpretation of results, grid scans were also conducted over the surface of the test specimens. Grid scans were performed simply by taking GPR data along the straight lines of a grid over an area covering the test specimen, as shown in Figure 4. The grid required for a grid scan was arranged using a grid mat secured to the surface so that it did not move while taking the GPR data. The spacing between each scan in the grid were set as 2 in. for this study. The spacing between the individual lines in the grid determines the resolution of the data collected and it is termed as grid resolution. The lower the spacing between each line scans in a grid, the higher the resolution of the data collected. Hence, the images will be clearer, and it would be easier to interpret the data. Apart from using closely spaced grid lines, a higher GPR frequency can also be used to increase the resolution. The depth of penetration and the resolution of a GPR depends upon the frequency of the pulse transmitted into the material. Lower frequencies allow deeper penetration with lower resolution, whereas higher frequencies allow detection of small defects (higher resolution) but localized within a shallower depth [31,93,99,100]. Four different GPR systems with different center frequency ranges were used in this experiment to determine the effect of GPR frequency on the bar detectability, as shown in Table 3 and Figure 5. The data acquired from grid scans were used to give the cross-section image of the test specimen through the plane parallel to the surface of the specimen along its depth. The cross-sectional image along the depth of the specimen is termed as a depth slice or time slice image.  The spacing between each scan in the grid were set as 2 in. for this study. The spacing between the individual lines in the grid determines the resolution of the data collected and it is termed as grid resolution. The lower the spacing between each line scans in a grid, the higher the resolution of the data collected. Hence, the images will be clearer, and it would be easier to interpret the data. Apart from using closely spaced grid lines, a higher GPR frequency can also be used to increase the resolution. The depth of penetration and the resolution of a GPR depends upon the frequency of the pulse transmitted into the material. Lower frequencies allow deeper penetration with lower resolution, whereas higher frequencies allow detection of small defects (higher resolution) but localized within a shallower depth [31,93,99,100]. Four different GPR systems with different center frequency ranges were used in this experiment to determine the effect of GPR frequency on the bar detectability, as shown in Table 3 and Figure 5. The data acquired from grid scans were used to give the cross-section image of the test specimen through the plane parallel to the surface of the specimen along its depth. The cross-sectional image along the depth of the specimen is termed as a depth slice or time slice image.

PAU Tests
Ultrasonic testing (UT) operates on the principle that the incident ultrasonic waves generated by ultrasonic transducers (which transform electrical or optical signals into ultrasonic waves and vice versa) will be reflected back when they encounter an interface between two materials with different acoustic impedances. A phased array ultrasonic (PAU) is an advancement over the UT technology that can be achieved when several of these transducers are arranged together into an array and operated at slightly different times (either electrically or physically) so that the individual waves interact both positively and negatively, allowing beam focusing and beam steering, as shown in Figure 6 [108]. The advantages of PAU in comparison to conventional UT include a 5 to 10 times faster scanning rate, better images due to multiple angles and frequencies, which require less interpretation, higher resolution, capability of beam focusing, reliability, portability and mobility [109,110]. Unlike conventional UT, PAU testing allows signal focusing at desired locations and angles, which is advantageous for the testing of composite materials that have an anisotropic structure, creating challenges in signal evaluation [53,111]. However, PAU is a relatively new technique compared to the traditional NDTs, and hence the limitations of portable units of PAU include the uncertainty in its application as it has yet to be completely proven.

PAU Tests
Ultrasonic testing (UT) operates on the principle that the incident ultrasonic waves generated by ultrasonic transducers (which transform electrical or optical signals into ultrasonic waves and vice versa) will be reflected back when they encounter an interface between two materials with different acoustic impedances. A phased array ultrasonic (PAU) is an advancement over the UT technology that can be achieved when several of these transducers are arranged together into an array and operated at slightly different times (either electrically or physically) so that the individual waves interact both positively and negatively, allowing beam focusing and beam steering, as shown in Figure 6 [108]. The advantages of PAU in comparison to conventional UT include a 5 to 10 times faster scanning rate, better images due to multiple angles and frequencies, which require less interpretation, higher resolution, capability of beam focusing, reliability, portability and mobility [109,110]. Unlike conventional UT, PAU testing allows signal focusing at desired locations and angles, which is advantageous for the testing of composite materials that have an anisotropic structure, creating challenges in signal evaluation [53,111]. However, PAU is a relatively new technique compared to the traditional NDTs, and hence the limitations of portable units of PAU include the uncertainty in its application as it has yet to be completely proven.
Line scan using PAU was performed by moving the array of ultrasonic transducers in a sideways direction along the desired line of inspection, as shown in Figure 7. Each scan at individual positions is stitched together to give one continuous line scan, which is the cross-sectional image through a plane normal to the surface along the direction of the scan. Similarly, the area scan using PAU was conducted as a stripe scan. The stripe scan was performed by positioning the array of ultrasonic transducers perpendicular to the desired line of inspection and then taking scans while moving forward with desired spacing between the consecutive scans, as shown in Figure 7. Each individual scan collected during stripe scan is simply a line scan having a width of measurement equal to that of the PAU device or the array of transducers. These individual scans along the line of inspection are stitched together to give a time slice or depth slice view, which can be further processed into a full 3D iso-surface representation. The depth slice view is simply the cross-section image parallel to the surface scanned.
Line scan using PAU was performed by moving the array of ultrasoni in a sideways direction along the desired line of inspection, as shown in F scan at individual positions is stitched together to give one continuous line s the cross-sectional image through a plane normal to the surface along the di scan. Similarly, the area scan using PAU was conducted as a stripe scan. T was performed by positioning the array of ultrasonic transducers perpend desired line of inspection and then taking scans while moving forward with ing between the consecutive scans, as shown in Figure 7. Each individual s during stripe scan is simply a line scan having a width of measurement eq the PAU device or the array of transducers. These individual scans along spection are stitched together to give a time slice or depth slice view, which c processed into a full 3D iso-surface representation. The depth slice view cross-section image parallel to the surface scanned. Two different PAU systems were used in this experiment to determine  Line scan using PAU was performed by moving the array of ultrasonic transducers in a sideways direction along the desired line of inspection, as shown in Figure 7. Each scan at individual positions is stitched together to give one continuous line scan, which is the cross-sectional image through a plane normal to the surface along the direction of the scan. Similarly, the area scan using PAU was conducted as a stripe scan. The stripe scan was performed by positioning the array of ultrasonic transducers perpendicular to the desired line of inspection and then taking scans while moving forward with desired spacing between the consecutive scans, as shown in Figure 7. Each individual scan collected during stripe scan is simply a line scan having a width of measurement equal to that of the PAU device or the array of transducers. These individual scans along the line of inspection are stitched together to give a time slice or depth slice view, which can be further processed into a full 3D iso-surface representation. The depth slice view is simply the cross-section image parallel to the surface scanned. Two different PAU systems were used in this experiment to determine their FRP bar detection capability, as shown in Table 4 and Figure 8. The MIRA 3D device had 64 ultrasonic transducers located in a 16 × 4 grid at 3 cm spacing (extended 16 rows of 4 transducers each) where the Pundit Live Array Pro had 24 ultrasonic transducers located in an 8 × 3 grid at 3 cm spacing (8 rows of 3 transducers each).  Two different PAU systems were used in this experiment to determine their FRP bar detection capability, as shown in Table 4

GPR Tests Results
GPR response or line views traversing perpendicular to bars embedded in each slab specimens obtained using a 1 GHz GPR device are shown in Figure 9, where the top of

GPR Tests Results
GPR response or line views traversing perpendicular to bars embedded in each slab specimens obtained using a 1 GHz GPR device are shown in Figure 9, where the top of the hyperbolic shape, i.e., inverted U shape, indicates the location of the bar and the shape of the tails gives a measure of velocity and depth [116]. The background subtraction filter was used to enhance the hyperbolas from embedded FRP bars. Without the application of this filter, the top of the hyperbola of the shallow targets (FRP bars close to the top surface) would have been obscured by the direct wave band that appears at the top surface. However, using the filter also removed the horizontal band indicating the bottom surface of the slab, but because the main objective of this research was to detect the FRP bars, the detection of the bottom surface was deemed not important. Only the line scans conducted at the centerline of the slabs using the lowest frequency GPR device has been presented in this paper for the brevity of the results as it would represent the lowest resolution among the GPR devices used in this study. Further, it should be noted that in order to confirm that the hyperbolas detected were indeed the FRP bars, but not other internal features (such as voids), several line scans were performed and checked for repetition of the same pattern of hyperbolas in each line scan (which indicates the presence of a continuous internal target such as a reinforcement bar). For Slab C with GFRP bars in one direction, bars at up to a depth of 1.5 inch were detectable. For Slab J with GFRP bars in two directions (mesh), the hyperbolas for the bars on both the top and bottom mesh were detectable. The bars near the edge were shadowed by the hyperbola due to the edge of the slab, which could have been avoided by leaving an offset from the edge for taking the measurements. The GPR line scan of Slab L was able to detect carbon strand (Bar 5), steel bar (Bar 6) and GFRP strand (Bar 7), along with the larger diameter #8 GFRP bar (Bar 2). However, it was not able to detect other glass (Bars 1, 3, 4) and basalt ( Bars 8,9) bars. The bar detectability of the line scans conducted using lower frequency device is further summarized in Table 5.
The depth slices obtained for each slab specimen using each GPR device are further illustrated in Figure 10, where it can be seen that the bar detectability improves as the device central frequency increases. Although the line scan in Figure 9a shows four hyperbolas for Bars 1-4, the depth slice in Figure 10a only shows the presence of Bar 2, which is the larger GFRP bar (#6 GFRP bar) at shallower depth in Slab C. This could be because of the limitations in vertical resolution due to a lower frequency of the GPR device. Larger GFRP bars are detectable using lower frequency GPR device at shallower depth, but they return weaker hyperbolas in comparison to metallic objects. Moreover, the relatively higher amplitude of shallower GFRP bars may have dominated and masked the lower amplitude returns from the GFRP bar that are deeper and smaller in diameter, which could be the reason only Bar 2 was visible in Figure 10a. However, the depth slices in Figure 10b,c, obtained using higher frequency GPR devices, are capable of detecting a smaller diameter bar, even at deeper depths. that the hyperbolas detected were indeed the FRP bars, but not other internal features (such as voids), several line scans were performed and checked for repetition of the same pattern of hyperbolas in each line scan (which indicates the presence of a continuous internal target such as a reinforcement bar). For Slab C with GFRP bars in one direction, bars at up to a depth of 1.5 inch were detectable. For Slab J with GFRP bars in two directions (mesh), the hyperbolas for the bars on both the top and bottom mesh were detectable. The bars near the edge were shadowed by the hyperbola due to the edge of the slab, which could have been avoided by leaving an offset from the edge for taking the measurements. The GPR line scan of Slab L was able to detect carbon strand (Bar 5), steel bar (Bar 6) and GFRP strand (Bar 7), along with the larger diameter #8 GFRP bar (Bar 2). However, it was not able to detect other glass (Bars 1, 3, 4) and basalt ( Bars 8,9) bars. The bar detectability of the line scans conducted using lower frequency device is further summarized in Table  5.

Slab ID Bar 1 Bar 2 Bar 3 Bar 4 Bar 5 Bar 6 Bar 7 Bar 8 Bar 9 Bar 10
The depth slices obtained for each slab specimen using each GPR device are further illustrated in Figure 10, where it can be seen that the bar detectability improves as the device central frequency increases. Although the line scan in Figure 9a shows four hyperbolas for Bars 1-4, the depth slice in Figure 10a only shows the presence of Bar 2, which is the larger GFRP bar (#6 GFRP bar) at shallower depth in Slab C. This could be because of the limitations in vertical resolution due to a lower frequency of the GPR device. Larger GFRP bars are detectable using lower frequency GPR device at shallower depth, but they return weaker hyperbolas in comparison to metallic objects. Moreover, the relatively higher amplitude of shallower GFRP bars may have dominated and masked the lower In the case of depth slices for Slab J with a top and bottom mesh of GFRP bars, all the GPR devices could clearly detect the top mesh of #6 GFRP bars at a shallower depth, while the bottom mesh was only visible for the higher frequency GPRs, as shown in Figure 10d-f. The inability of lower frequency GPR to show the bottom mesh could be due to the same reason for not showing deeper GFRP bars in Slab C, as explained previously. In addition to the higher center frequency of the 2 GHz GPR device, its dual polarization feature further permits detection on both first and second levels of bars, whereas for the GPR device with maximum center frequency of 4 GHz and 6 GHz, the ability to detect the bottom mesh is solely due to the higher resolution, which it can afford because of its higher frequency.
The line scan of Slab L shows that Bar 2 (#8 GFRP bar) is visible as a faded hyperbola in addition to distinctive hyperbolas of Bars 5, 6 and 7. But again, Bar 2 is not visible in the depth slice in Figure 10g because GFRP bars return weaker hyperbolas in comparison to steel bars (Bar 6) and the high amplitude steel bar dominates and hides the lower amplitude returns from the GFRP bar. In Figure 10g-i, it can be seen that the strongest detection is that of the steel bar (Bar 6), followed by the carbon strand (Bar 5) and GFRP strand (Bar 7). Carbon strands, in addition to having cavities from the twisting shape of exterior wires that are filled (or not filled) with concrete, they are also electrically conductive, which could be the reason they are distinctively visible, and the GFRP strands are visible only because of the cavities within the GFRP twisting wires. Apart from Bars 2, 5, 6 and 7, the higher frequency GPR was also able to capture other bars, as shown in Figure 10i. Appl

PAU Tests Results
The results of line scans and the area scans obtained from the PAU testing on each slab specimen using MIRA 3D and Pundit Live Array Pro devices are shown in Figure 11. It can be seen that PAU could not detect bars in Slab C and Slab J. However, it was able to detect carbon strands, steel bars and GFRP strands in Slab L. As ultrasonic testing and PAU is very sensitive in detecting the presence of (air) voids, their ability for detecting CFRP and GFRP strands could be attributed to the presence of (air) voids within the twisted FRP cables of these strands.
is that of the steel bar (Bar 6), followed by the carbon strand (Bar 5) and GFRP strand (Bar 7). Carbon strands, in addition to having cavities from the twisting shape of exterior wires that are filled (or not filled) with concrete, they are also electrically conductive, which could be the reason they are distinctively visible, and the GFRP strands are visible only because of the cavities within the GFRP twisting wires. Apart from Bars 2, 5, 6 and 7, the higher frequency GPR was also able to capture other bars, as shown in Figure 10i.

PAU Tests Results
The results of line scans and the area scans obtained from the PAU testing on each slab specimen using MIRA 3D and Pundit Live Array Pro devices are shown in Figure 11. It can be seen that PAU could not detect bars in Slab C and Slab J. However, it was able to detect carbon strands, steel bars and GFRP strands in Slab L. As ultrasonic testing and PAU is very sensitive in detecting the presence of (air) voids, their ability for detecting CFRP and GFRP strands could be attributed to the presence of (air) voids within the twisted FRP cables of these strands. (e) (f)

Discussion
The GFRP bars that dominate the embedded concrete reinforcement application are not conductive and have a density similar to concrete. Therefore, the conventional NDT methods that use electromagnetic and stress waves for the detection of steel bars fail to detect GFRP bars with the same clarity as steel. As explained earlier in sections 2.1 and 2.2, the GPR and PAU devices work on the basis of differences in dielectric constant and

Discussion
The GFRP bars that dominate the embedded concrete reinforcement application are not conductive and have a density similar to concrete. Therefore, the conventional NDT methods that use electromagnetic and stress waves for the detection of steel bars fail to detect GFRP bars with the same clarity as steel. As explained earlier in Sections 2.1 and 2.2, the GPR and PAU devices work on the basis of differences in dielectric constant and acoustic impedance encountered at an interface. The reflection coefficient of the electromagnetic wave (for GPR tests) and ultrasonic waves (for PAU tests) passing through concrete when they encounter air (bottom surface), steel bar and GFRP bars are shown in Table 6. It can be seen that the concrete-GFRP interface reflection coefficient, i.e., the amount of reflected energy, is less than 8% for GPR tests and has a very minimal value of 0.04% for PAU tests, which is the reason why GPR had limited detectability of GFRP bars and PAU was not able to detect them at all. The dielectric constants and acoustic impedances for average concrete, air and steel used in Table 6 were obtained from the literature [116][117][118]. From the specifications chart provided by the manufacturer of the GFRP bar, the dielectric constant was determined to be a value of less than 5, and the acoustic impedance was estimated to be 10.7 × 10 6 kg/m 2 s (from the equation Z = E.ρ, where E = 54.5 GPa is the modulus of elasticity and ρ = 2.1 g/cm 3 is the density of the GFRP bar).  However, the depth slices of GPR tests obtained using four different devices with different antenna center frequency demonstrated that the detection of FRP bars becomes better with the increase in frequency, as shown in Table 7. Apart from the frequencies playing the major role in the detection of FRP bars, different detectability levels could also be due to several other factors related to the radar technologies being used in each of the GPR devices. For example, the 2 GHz GPR device also had a dual polarization feature that enhances the detection of multilevel bars. The maximum frequency GPR devices used in this study also had a Stepped Frequency Continuous Wave (SFCW) system. This provides both the benefits of higher resolution in the detection of shallow targets and the increase in penetration depth. Moreover, there were some disparities in the results of the line scans and the grid scans for the lower frequency GPR device as the bars visible in the line scan were not visible in the depth slices. This could be due to the fact that the depth slices are produced through the interpolation of a series of line scans, and it would show the bars with stronger signals, i.e., stronger hyperbola more dominantly in comparison to the bars with relatively weaker hyperbola. Thus, although it is time consuming, it is recommended to go through each line scan of a grid while analyzing the results for GPR tests. On the other hand, it was found that PAU can only detect FRP strands, which could be due to the presence of (air) voids within the twisted wires of FRP or irregularity of the shape of their surface in comparison with smooth bars.

Summary and Conclusions
The objective of this research was to determine the feasibility of using commercially available NDT methods such as GPR and PAU in detecting FRP bars embedded in concrete.
These two NDT methods were tested over several parameters, such as FRP type, bar diameter, bar direction, and bar depths to determine their limitations and detection capabilities. Detecting the FRP bars embedded in concrete is the first obstacle to be overcome before being able to detect damages for ensuring their structural safety. Thus, the long-term objective of this study is to promote the research on NDT methods applicable to embedded FRP bars, which could help in proliferating their use in the construction industry.
In conclusion, the results of this study show that NDT methods that rely on electromagnetic waves such as Ground Penetrating Radar (GPR) become less effective, if not obsolete for non-metallic/non-conductive embedded bars. However, with the increase in center frequency of the GPR device, the detectability of FRP bars can be fairly improved. Similarly, other NDT methods such as Ultrasonic Testing (UT) or Phased Array Ultrasonic (PAU) that are based on stress waves have some capability for detecting steel reinforcement, but they perform poorly for the detection of the most commonly used FRP embedded bars (GFRP bars) but are good for detecting FRP strands.
However, there are some limitations related to the research conducted in this paper, which can be further explored and incorporated in the future studies related to NDT methods for the inspection of FRP-reinforced concrete elements. The scope of this study was limited to the real time test results obtained from post processing tools embedded within the devices using GPR and PAU, which were originally tuned for the detection of steel bars in traditional reinforced concrete elements. The detectability of the FRP bars embedded in concrete can be further verified by the use of advanced post processing algorithms such as SAFT (synthetic aperture focusing technique) or FMC/TFM (full matrix capture/total focusing method) in a follow up future study. Similarly, this study is limited to only three specimens with limited variation in depth of bars up to 3 in., which can be overcome in future studies by either conducting experimental verification on several specimens or by conducting numerical simulation analysis with adequate test parameters to collect more data for a statistically sound validation.

Data Availability Statement:
The data presented in this study are available by request from the corresponding author.