Experimental assessment of an RFID-based crack sensor for steel structures

The timely detection of fatigue crack formation in many engineering applications such as railways, bridges, aircraft, and pipelines is crucial for safety, extended service life, and reduced maintenance cost. Cyclic loading can produce fatigue, which occurs progressively and locally, leading to sudden failure below the yield stress limit. Traditionally, highway bridges use various types of steel-welded beams that are susceptible to these effects. For decades, cover plates were added to the flanges of such beams to increase their flexural capacity [1]. The use of cover plates in regions of high applied moment allows sections on primary beams to be lighter while providing more flexural capacity [2, 3]. However, the use of welded cover plates introduces a transverse weld periphery at the toe, forming a line of elevated tension where fatigue cracks can form prematurely and propagate into the main beam. The National Cooperative Highway Research Program [4] noted that fatigue crack propagation at nearly all other structural details occurred as cracks initiated from the toe of fillet or groove welds because of high stress concentration due to discontinuity and residual tension stress. Thus, welded cover plated beams are more susceptible to fatigue crack formation than rolled beams, and this requires closely monitored information of stress level by frequent visual inspection and field measurement.

The Federal Highway Administration has mandated that all highway bridges located on public roads be inspected every 2 years [5]. Most inspections rely exclusively on visual means. Crack monitoring activities using crack-detecting sensors can complement visual inspections and provide information about crack damage. As such, extensive research has been performed in flaw-detecting sensors and many non-destructive evaluation and testing methods have been developed.

A few methods to detect cracks in metallic materials have been developed. One of the most common and effective methods to detect flaws and cracks is using angle beam ultrasound [6]. The ultrasonic wave can be generated by wireless inductively-coupled piezoelectric transducers. This wave is reflected back to the transducer by some form of discontinuity, such as another surface or a crack. Typically, crack inspection with this method requires an experienced operator and it is manually performed by moving the transducer across the surface at different orientations. Another mature technology is the usage of eddy currents where the alternation of the electromagnetic field generates magnetic lines of force that reveal hidden defects. This method is very effective in determining subsurface cracks due to its great penetration depth [7, 8]. Other technologies that have been patented or are commercially available include ultrasonic flaw detectors [9], magnetorestive sensors [10], surface-mount piezoelectric paint sensors [11, 12], probe-pump-based Brillouin sensor systems [13], coaxial cable sensors [14], and fiber-optic sensors [15], among others. However, the use of these methods for long-term monitoring of crack patterns in larger scale civil infrastructure is time-consuming and expensive.

Passive radio frequency identification (RFID) antennas have been studied to provide a low-cost method for crack detection. Kalansuriya et al [16] introduced the concept of using RFID tag antennas to sense surface cracks and the usage of a grid of RFID tags to monitor crack patterns in civil infrastructure. The application of RFID technology for crack monitoring relies on the permanent changes in impedance and radiation efficiency caused by the presence of a crack. The RFID method uses electromagnetic transmission by means of a radio frequency-compatible integrated circuit (IC) to retrieve data from tags and send it to reader antennas. The communication protocols between the RFID tag and the reader are standardized and efficient, making RFID an ideal wireless communication infrastructure for dense arrays of sensors. The greatest advantage of RFID technology is the elimination of coaxial cables and the reduction of installation time and maintenance costs. The cost of a passive RFID tag in 2006 was $0.02–$0.05 and is still decreasing [17]. Passive RFID tags also possess the advantage of not depending on an external power source for operation, one of the most recurrent and concerning problems in non-destructive evaluation and health monitoring applications.

Kalansuriya et al [16] presented a method where the tag antenna is sensitive to cracking for 50% of the radiating element's length. To counteract this limitation, they used 2D arrays. However, these grids contained gaps where crack detection would be missed and need to be optimized [16, 18]. In addition, Kalansuriya et al [16], Mohammad and Huang [19], and Cazeca et al [20] developed in-house RFID antenna-based sensors for crack detection using their knowledge as electrical engineers. From a civil engineering standpoint, the assembly of an RFID-based sensor is costly given the typical large deployments needed. Commercially available RFID technology is more viable due to the reduced unit cost in comparison to in-house developments. However, the performance of low-cost commercial RFID tags for crack detection has not been studied. In consequence, significant research is needed to determine the sensitivity to damage of commercial RFID technology and to establish a guideline for deployment regarding preferred orientation, placement, and configuration for reasonable results.

This paper presents a study on damage sensitivity of ultra-high frequency (UHF) passive RFID tags for crack detection on steel structures. Damage sensitivity has been determined to be the ability of the tag to return a different backscatter power signal upon changes to the underlying metallic surface and changes to its own integrity. A crack detection sensing system has been developed for two cases: using a single RFID tag and using a 2D array of RFID tags lain over a substrate material. Its performance has been validated by laboratory-scale experiments. These experiments explore the effects on backscatter power caused by: (1) read distance for tag antennas severed in difference locations, (2) damage stages as a crack propagates from an underlying metallic surface into the substrate and into a single tag, and (3) damage stages as a crack propagates from an underlying metallic surface into the substrate and into a tag in a 2D array. The system overcomes the adverse effect of the metallic structure on the backscatter signal of the RFID tag while maintaining a high tag density of the array for increased sensing pervasiveness.

2.1. Backscatter power

Kalansuriya et al [16] identify backscatter power as the most important measured value from a tag for crack detection and characterization. Backscatter power can be described by the radar range equation:

$$\begin{eqnarray}&&{P}_{{\rm{R}}}={P}_{{\rm{T}}}\tfrac{{G}_{{\rm{T}}}{G}_{{\rm{R}}}}{4\pi }{\left(\tfrac{\lambda }{4\pi {{\rm{R}}}_{{\rm{T}}}{{\rm{R}}}_{{\rm{R}}}}\right)}^{2}\sigma ,\end{eqnarray} \tag{ 1 }$$

where P_R is the backscatter power, P_T is the transmitted power, G_T is the transmitting antenna gain, G_R is the receiving antenna gain, λ is the signal wavelength, R_T is the distance between the target (the tag chip, in this application) and the transmitting antenna, R_R is the distance between the target and the receiving antenna, and σ is the target's radar cross section. In a monostatic scattering application, the antenna emitting the electromagnetic signal also receives the echo from the target tag chip so that R_R = R_T and G_R = G_T.

Equation (1) shows that backscatter power is attenuated as the reading distance increases by a quartic factor. P_R can be determined from the received signal strength indicator (RSSI) logged by the reader equipment from the following expression:

$$\begin{eqnarray}&&{\rm{R}}{\rm{S}}{\rm{S}}{\rm{I}}=10{\mathrm{log}}_{10}\left(\tfrac{{P}_{{\rm{R}}}}{1\,{\rm{mW}}}\right).\end{eqnarray} \tag{ 2 }$$

RSSI is then a decibel expression (dBm) of backscatter power.

Most reading equipment uses frequency hopping spread spectrum, so the value of RSSI depends on the transmit frequency channel. The signal frequency varies in the range of 902–928 MHz in North America. The Impnj Speedway MultiReader software used in this study performs a frequency hopping sequence by changing the transmitting channel during each inventory session. Table 1 shows the signal frequency assigned to each channel. This study uses average values of RSSI of data samples equally weighted across all frequency channels.

Table 1.  Impnj Speedway reader frequency plan for North America.

Channel number	1	2	3	4	...	49	50
Center frequency (MHz)	902.75	903.25	903.75	904.25	...	926.75	927.25

2.2. Substrate material: image theory behind RFID wave propagation

The usage of RFID technology for crack detection on metallic surfaces adds a layer of complexity to the received backscatter power signal related to image theory. A steel plate is a large conductive surface that behaves as a ground plane. The backscatter power signal received at the antenna is equal to:

$$\begin{eqnarray}&&{\rm{\exp }}(-{\rm{j}}{\beta }_{{\rm{a}}{\rm{i}}{\rm{r}}}z)({I}_{{\rm{t}}{\rm{a}}{\rm{g}}}+{I}_{{\rm{g}}{\rm{r}}{\rm{o}}{\rm{u}}{\rm{n}}{\rm{d}}}\,{\rm{\exp }}(-{\rm{j}}{\beta }_{{\rm{m}}{\rm{a}}{\rm{t}}{\rm{e}}{\rm{r}}{\rm{i}}{\rm{a}}{\rm{l}}}{\rm{\Delta }}z)),\end{eqnarray} \tag{ 3 }$$

where, ${\beta }_{{\rm{a}}{\rm{i}}{\rm{r}}}$ is the phase constant of air, ${\beta }_{{\rm{m}}{\rm{a}}{\rm{t}}{\rm{e}}{\rm{r}}{\rm{i}}{\rm{a}}{\rm{l}}}$ is the phase constant of the substrate material (a material separating the tag from the metallic surface), ${I}_{{\rm{t}}{\rm{a}}{\rm{g}}}$ is the primary backscatter of the tag, and ${I}_{{\rm{g}}{\rm{r}}{\rm{o}}{\rm{u}}{\rm{n}}{\rm{d}}}$ is the image backscatter. For a horizontal tag (i.e. a tag oriented parallel to the ground plane), ${I}_{{\rm{t}}{\rm{a}}{\rm{g}}}=-{I}_{{\rm{g}}{\rm{r}}{\rm{o}}{\rm{u}}{\rm{n}}{\rm{d}}}.$ If a horizontal metallic RFID tag is placed directly on the metallic surface acting as a ground plane $({\rm{\Delta }}z=0),$ the power signal received at the antenna will be zero.

The magnitude of the reflected wave can be maximized when the reflection of the RFID tag comes from an image source located half a wavelength below (see figure 1) since a half wavelength corresponds to a 180° phase change. In practical terms, this means that, ideally, the RFID tag should be placed at a quarter wavelength in front of the metallic surface in order to maximize the power received by the antenna.

Zoom In Zoom Out Reset image size

The antenna used in this investigation is a right hand circularly polarized antenna that transmits signals at frequencies ranging from 902 to 928 MHz. Since the wavelength of these signals is approximately 1 ft, a substrate material closely resembling the relative permittivity of air (such as polystyrene foam) should ideally be 3 in thick. However, such a thick substrate would raise concerns on the structural effect that it would sustain on the girder. If the material selected to separate the tag from the metallic surface has the adequate dielectric properties, it can be used to reduce the physical separation between the tag and the ground plane. The wavelength of an electromagnetic wave in a dielectric medium is given as:

$$\begin{eqnarray}&&\lambda =\tfrac{c}{f}\,* \,\tfrac{1}{\sqrt{{{\epsilon }}_{{\rm{r}}}}},\end{eqnarray} \tag{ 4 }$$

where, $f$ is the wave frequency, $c$ is the speed of light, and ${{\epsilon }}_{{\rm{r}}}$ is the relative permittivity of the material between the target tag and the ground plane. In vacuum conditions, ${{\epsilon }}_{{\rm{r}}}=1.$ The relative permittivity of air is approximately 1. Relative permittivity can be related to the phase constant by:

$$\begin{eqnarray}&&{{\epsilon }}_{{\rm{r}}}={\left(\tfrac{\beta c}{2\pi f}\right)}^{2}.\end{eqnarray} \tag{ 5 }$$

According to equation (4), a material with higher permittivity also reduces the wavelength in a dielectric medium, thus reducing the required distance ${\rm{\Delta }}z$ to maximize the received backscatter power. Therefore, a thinner substrate made of a material with a relative permittivity greater than one is preferred to increase the received backscatter power signal. This substrate must also be sufficiently elastic in order to transfer the strains at the extreme tension fiber of the girder into the RFID tag and it should also be able to adhere to a metallic surface with weatherproof adhesive. From the available materials that fulfill all of these characteristics, ethyl-vinyl acetate (EVA) rubber, which has a relative permittivity of approximately 2.8, has been found to increase RSSI satisfactorily while also being a very flexible and durable material.

3.1. Single RFID-based crack sensor configuration

The components of a single crack detection sensor include a commercial RFID tag, a layer of a substrate material, and adhesives (see figure 2). Multiple adhesives including epoxy, double-sided tape, and a cyanoacrylate-based glue were used to test their effect on RSSI and no significant difference was found. Epoxy was finally chosen as an adhesive for bonding at both interfaces in field deployment because of its proven effectiveness in strain transferring applications, such as with optical fiber sensors.

Zoom In Zoom Out Reset image size

The commercial UHF RFID tag used in this paper is the Alien Technology ALN-9662 Short Inlay tag. This tag is EPG Gen 2 and ISO/IEC 18000-6C compliant and it uses a Higgs 3 EPC Class 1 Gen 2 RFID tag IC. The tag antenna is made of a flexible metallic material, which is adhered to a wet inlay.

The substrates used in this paper are EVA rubber or polystyrene foam. Polystyrene foam is used for experiments where a material with relative permittivity similar to that of air is desired for basic evaluations of backscatter power behavior. Since EVA rubber has a higher relative permittivity and is flexible and durable, it is used in experiments for sensing system performance evaluation where higher RSSI readings are desired.

3.2. Performance evaluation experiments

There are two directions of fatigue crack propagation that are possible in beams with welded cover plates: (1) crack propagation from the beam into the substrate material and then into the tag, causing the tag antenna to be severed and (2) propagation along the beam surface beneath the sensing system. The first case implies that the substrate material and eventually the tag antenna will be severed. The second case assumes that the separation of the metallic material will induce strain into the substrate material and thus into the tag antenna. The direction of crack propagation assumed in this study is the first case.

3.2.1. Read distance and loss of the effective area of a tag antenna

As implied by equation (1), greater read distances cause backscatter power from the tag in question to be reduced by a quartic factor. The effects of read distance on RSSI along with the effect of severing a portion of the tag antenna have been studied. RSSI was measured for read distances at 3 foot intervals up to 15 ft, as shown in figure 3. The RFID tag was attached to a polystyrene foam block. Polystyrene foam is known to have low reflective and absorptive properties, closely simulating air.

Zoom In Zoom Out Reset image size

The subsequent measurements of RSSI were made for tags that were severed at different locations, simulating different scenarios of an underlying crack that has propagated into the antenna. The two cutting sequences shown in figure 4 were examined. Each cut number represents a different location in which a crack disconnects an additional portion of the antenna from the IC. For example, the damage scenario #2 in sequence #1 represents the situation in which a crack along the dotted line #1 and another crack along the dotted line #2 disconnected the portions to the right of the dotted line #1 and to the left of the dotted line #2. Scenario #3 represents two cracks that disconnected the portions to the right of dotted line #3 and to the left of dotted line #2. The RSSI at each stage in the cutting sequence was measured at the distances shown in figure 3.

Zoom In Zoom Out Reset image size

The trends of RSSI versus damage scenario and read distance for both sequences are shown in figure 5. As greater areas of the tag antenna were disconnected from the tag IC, RSSI exponentially decayed. Cuts made between the IC and the square patches do not yield as great a gradient in RSSI as cuts made within the square patches. When greater antenna surface area was disconnected from the IC loop, a larger drop in RSSI was detected. Therefore, damage within the tag can be more precisely located when the damage is within the square patch than when the damage is between the patch and the IC.

Zoom In Zoom Out Reset image size

The reader was incapable of detecting the tag backscatter power at further distances. The dependence of RSSI on read distance establishes a requirement for read distance standardization. Since RSSI decreases with distance, the chosen parameter for crack detection will depend on the final choice of read distance.

Finally, it was noted that RFID tags could not reflect power when receiving direct sunlight. This limits the usage of RFID tags for crack detection to shaded areas. As the bridge deck covers bridge girders throughout most of the day, it is anticipated that proper scheduling will suffice for proper damage identification in the field.

3.2.2. Damage detection of crack propagating into a tag antenna

A second set of tests was conducted to demonstrate the crack propagation detection capability of commercial RFID tags on a metallic surface. Figure 6 shows the overall arrangement for these experiments. A PC with Impnj MultiReader software was connected to an Impnj Speedway Revolution R420 UHF RFID Reader. A high gain circular right hand polarized patch antenna was connected to the reader. A 1/8 in thick rectangular aluminum plate was used as a test specimen. A 1 in long incision was made into one side with a 1/16 in vertical band saw blade. The opposing side was left unaltered as a control surface.

Zoom In Zoom Out Reset image size

As expected, no radiation was detected from the tag when it was directly attached to the metallic surface. A polystyrene foam plate was used as a substrate material to simulate air as closely as possible. In a similar setup, an EVA foam plate was also used as a substrate material for comparison. The aluminum plate, polystyrene foam or EVA foam plate, and the ALN-9662 RFID tag were raised onto a stack of polystyrene foam so that the tag IC would be elevated to the same height as the center of the reader antenna while avoiding excessive radiation interference. The read distance was fixed at 3 ft.

Four damage scenarios were tested for detection: (1) undamaged surface with uncut tag, (2) cracked surface with uncut tag, (3) cracked surface and cracked substrate material with uncut tag, and (4) cracked surface with cut tag (see table 2). Each scenario represents a stage in crack formation and propagation, the first being no damage at all and the last being the ultimate damage case. Damage stage #2 shows the initial stage of crack formation on the metallic surface while the sensing system is untouched. The inclusion of damage stage #3 ensured to account for the changes in system impedance that are to be expected upon the comprisal of a substrate material with a higher permittivity, as in the case of EVA foam. Since polystyrene foam has a permittivity close to that of air, the damage scenarios that were tested when this material was the substrate were damage scenarios #1, #2, and #4. Figure 7 shows the tag placement on the substrate materials and the metallic surface.

Table 2.  Conditions in each damage scenario.

	Sensing component
Damage scenario	Tag	Substrate	Metallic surface
1	Intact	Intact	Intact
2	Intact	Intact	Crack damage
3	Intact	Crack damage	Crack damage
4	Crack damage	Crack damage	Crack damage

Zoom In Zoom Out Reset image size

It is not anticipated that EVA foam will break under field conditions, particularly with minor cracking. However, it is expected that damage stage #2 will occur in the field and that cracks will propagate parallel to the monitored surface. The inclusion of damage stages #3 and #4 in the present study serves the purpose of providing a complete picture of tag sensitivity to damage propagating perpendicularly to the monitored surface.

Figure 8 shows the resulting measurements of RSSI for all damage scenarios using intermediate polystyrene foam and EVA foam. The overall performance of the tag was improved (i.e. RSSI was increased) when EVA foam was used, as raw RSSI ranged between −52.87 and −40.46 dBm compared to the range of −56.03 and −45.38 dBm when polystyrene foam was used. As a crack formed on the metallic surface and propagated up to the substrate, RSSI consistently increased regardless of the substrate material. A drop in RSSI occurred upon the cracking of the tag antenna (damage scenario #4) compared to the control (damage scenario #1) when either EVA or polystyrene foams were used. It is clear then that the underlying metallic surface increases the radiation efficiency of the system when a crack is present on the left side of the tag IC with respect to the direction of the incident electromagnetic wave. It is therefore possible to detect an underlying crack that has not propagated into the RFID sensor and that has opened a gap across the depth of the metallic surface.

Zoom In Zoom Out Reset image size

4.1. Multiple sensor array configurations

To increase the pervasiveness of the crack propagation monitoring system, 2D arrays of tags were considered. It is known that the proximity of RFID tags has an effect in their sensitivity, causing some tags to report a gain or a reduction in backscatter power depending on the layout of the surrounding tags. This change in sensitivity is caused by tag detuning, tag shadowing, and re-radiation cancelation, collectively known as coupling or proximity effects [21].

A 2D array of tags should behave in a way analogous to the parasitic elements in a Yagi–Uda antenna [22]. The principal tag of interest in an array (hereafter referred to as the control tag) would be the driven element. The RSSI of this tag will be the principal indicator of crack formation and propagation. The strength of the RSSI of the control tag will be influenced by the surrounding tags in the array, similar to how director parasitic elements work together in the Yagi–Uda antenna to increase the antenna's gain. Therefore, a 2D array must be selected such to improve pervasiveness with close spacing and to enhance sensitivity to damage with placements that increase RSSI in the control tag. Different tag array configurations were studied based on their effects on the RSSI of a control tag. There are two features that are primarily pertinent to the development of the tag array to be used in this sensing system: spacing between tags and configuration.

4.1.1. Spacing between tags in 2D arrangement

Coupling effects between tags can either increase or decrease the backscatter power of the tags involved. In order to optimize the backscatter power of a control tag (T-1) the configurations shown in table 3 were tested for the following separations: 1/8, 1/4, 1/2, 1, and 2 in. Distance 1–2 refers to the separation between tags T-1 and T-2 and distance 1–3 refers to the separation between tags T-1 and T-3. Tag separation was measured as a clear distance from the edge of a patch of one tag to the nearest edge of the patch of the other tag. Figure 9 shows the setup used for all tag array experiments. The substrate material used in this experiment to separate the array from the 1/8 in thick aluminum plate was polystyrene foam and the read distance was kept constant at 3 ft in reference to the control tag, T-1.

Table 3.  RFID tag configurations used for spacing optimization.

Zoom In Zoom Out Reset image size

Figures 10 and 11 show the variation in RSSI on tag T-1 for different separations. The blue filled-in circle indicates the original received power of T-1 in standalone configuration (C1). Figure 12 shows the values of RSSI for all spacing combinations in configuration C4A. The spacing in horizontal configuration (C2) improved RSSI the most in tag T-1 at 1/4 in separation. The spacing in vertical configuration (C3) improved RSSI the most in tag T-1 at 1/8 in separation. Figure 12 shows that the same spacing increases RSSI the most. Therefore, the spacing combination that best increases RSSI is 1/4 in for horizontal spacing and 1/8 in for vertical spacing. This spacing will be used in all experimentation involving 2D arrays.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

4.1.2. Array placement

The position of a tag in relation to others can also have a significant impact on its backscatter power. Since crack propagation monitoring requires an increased number of sensing units for greater pervasiveness, a basic array of 3 rows by 2 columns of RFID tags was chosen to determine the best configuration for increased received power. Using polystyrene foam as the substrate material, the configurations shown in table 4 were tested for RSSI in the control tag, T-1. The array was placed on a 1/8 in thick aluminum plate. These configurations have been ordered from the one yielding the lowest RSSI in T-1 to the one yielding the highest RSSI in T-1. Figure 13 shows a bar graph of the ordered configurations.

Table 4.  2D array configurations sorted by RSSI on T-1.

Zoom In Zoom Out Reset image size

The configuration that most reduced backscatter power was a vertical array (C3) and the configuration that most increased backscatter power was a purely horizontal one (C2). The RSSI of T-1 was generally improved (in reference to itself in standalone configuration–C1) when there was at least one other tag somewhere on the column next to it. This implies that the array should include tags side by side to the control tag. This also suggests that a combination of configurations C4A, C4B, C4C, C5A, C5B, and C2 will be best to maintain a high RSSI while increasing pervasiveness. To maintain consistency and cover as much surface area as possible, configurations C5A and C5B, i.e. a 2 × 2 configuration with 4 tags, were preferred for the performance evaluation of the array.

4.2. Performance evaluation experiment: 2D array

Using the same setup as in previous 2D array experiments (see figure 9), a 2D array placed on a 0.5 in thick EVA foam sheet conforming to configuration C5A was used to verify the sensitivity of the system (see figure 14). The same four damage scenarios explained in table 2 were examined: (1) undamaged surface with uncut tag, (2) cracked surface with uncut tag, (3) cracked surface and substrate with uncut tag, and (4) cracked surface, substrate, and tag. Figure 15 shows the final damage stage (damage scenario #4). The crack made to the aluminum plate was a 1 in long incision into one side with a 1/16 in vertical band saw blade. The average RSSI of the control tag (upper left tag in the array) was used to compare changes in backscatter power at each damage stage.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The pattern of RSSI changes in the control tag as a crack propagated into the system is shown in figure 16. The selection of the 2D array configuration C5A lain on EVA foam and oriented perpendicular to the reader antenna showed an improvement in performance, yielding RSSI values ranging between −42.91 and −39.25 dBm. In this 2D array, damage to the metallic surface underneath the right side of the control tag IC caused a small drop in RSSI. Further damage in the substrate decreased RSSI slightly more. Finally, the ultimate damage state (damage scenario #4) increased RSSI significantly, providing a notable change to indicate damage propagation onto the tag antenna. This trend is mirrored to the pattern observed in the single tag system because the location of the crack in the 2D array performance evaluation experiment was on the right side of the IC with respect to the incident electromagnetic wave instead of on the left side of the IC as was the case in the single tag performance evaluation experiment. This behavior has been observed in in-house developments of linearly polarized RFID-based crack sensors where the direction of the change in backscatter power depends on the location of the crack with respect to the IC [16].

Zoom In Zoom Out Reset image size

Performance evaluation studies have revealed that raw RSSI varies from one commercial RFID tag to another when all other experimental parameters remain unaltered. For further automation of crack detection in bridges, establishing a damage index based on the change of RSSI instead of on raw RSSI values is desirable. Because the change in RSSI is proportional to the severity of crack propagation, percentage change in RSSI before and after crack damage can be used effectively to this end. The RSSI percentage change is:

$$\begin{eqnarray}&&{\rm{\Delta }}\mathrm{RSSI}\,( \% )=\left(\tfrac{{\rm{R}}{\rm{S}}{\rm{S}}{{\rm{I}}}_{{\rm{i}}{\rm{n}}{\rm{t}}{\rm{a}}{\rm{c}}{\rm{t}}}-{\rm{R}}{\rm{S}}{\rm{S}}{{\rm{I}}}_{{\rm{d}}{\rm{a}}{\rm{m}}{\rm{a}}{\rm{g}}{\rm{e}}{\rm{d}}}}{{\rm{R}}{\rm{S}}{\rm{S}}{{\rm{I}}}_{{\rm{i}}{\rm{n}}{\rm{t}}{\rm{a}}{\rm{c}}{\rm{t}}}}\right)\times 100,\end{eqnarray} \tag{ 6 }$$

where ${\rm{R}}{\rm{S}}{\rm{S}}{{\rm{I}}}_{{\rm{i}}{\rm{n}}{\rm{t}}{\rm{a}}{\rm{c}}{\rm{t}}}$ is the RSSI of the unaltered state and ${\rm{R}}{\rm{S}}{\rm{S}}{{\rm{I}}}_{{\rm{d}}{\rm{a}}{\rm{m}}{\rm{a}}{\rm{g}}{\rm{e}}{\rm{d}}}$ is the RSSI of the damage stage in question.

The algebraic sign of the percentage change is an indicator of the location of the crack formation with respect to the tag IC. Figure 17 shows the location of the crack formation with respect to the tag IC viewed from the angle of the incident electromagnetic wave in all performance evaluation experiments. In single tag configuration experiments (figures 17(a) and (b)), the crack was to the left of the IC while in the 2D tag array experiment (figure 17(c)), the crack was to the right of the IC. Figures 18 and 19 show the percentage changes in RSSI for the single tag configuration and the 2D array configuration, respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Damage occurring to the left of the tag IC presented small increases in RSSI during initial damage stages. When damage was to the right of the tag IC, small decreases in RSSI were observed during initial damage stages. In the single tag performance evaluation experiment involving EVA, initial damage stages (scenarios #2 and #3) induced percentage changes in RSSI ranging from 3.097% to 3.586%. The same damage stages yielded an RSSI percentage change of 2.665% on a single tag configuration on polystyrene foam and a range of 1.586%–2.52% on the control tag of a 2D array on EVA foam. Therefore, initial stages of damage can be detected when changes in RSSI are between 1.5% and 3.6%.

Furthermore, damage on the left of the tag IC yielded large drops in RSSI at the ultimate damage state (scenario #4) while ultimate damage on the right of the tag IC exhibited the opposite behavior. Ultimate damage on a single tag configuration on EVA foam presented an RSSI percentage change of 26.00%. The same configuration and damage state on polystyrene foam yielded a change of 20.19%. Ultimate damage on the control tag of the 2D array on EVA foam caused a 6.215% reduction in RSSI. Thus, advanced damage stages will cause the direction of RSSI gradient to change with values greater than 6%.

In summary, single tag and multiple tag configurations can both be used for specific crack monitoring situations. As percentage changes in RSSI in single tag configurations are larger, small cracks can be monitored more accurately. On the other hand, multiple tag arrangements reveal lower percentage changes in RSSI due to coupling effects. However, these gradients are sufficiently large to detect changes in longer cracks, which would produce larger percentage changes in RSSI. 2D arrays can also cover larger areas for expanded pervasiveness. The behavior of the sensing system makes percentage change in RSSI a competent damage index for crack monitoring.

An RFID-based crack sensor was successfully developed using low-cost commercial tags and its performance was validated with comprehensive laboratory-scale experiments. The percentage change in backscatter power from the RFID tags is a relevant damage index, as it is sensitive to crack propagation. The magnitude of the damage index consistently increases as crack damage gradually propagates from the metallic specimen to the substrate and the tag; this demonstrates the potential of the developed sensors for crack detection. Furthermore, the algebraic sign of the damage index indicates the location of damage within a tag with respect to its IC. A crack sensor with a single tag configuration displayed the best performance in crack detection with a high damage index change. Yet, a 2D array of tags is preferred to increase the sensing range. Guidelines on optimal spacing and configuration of multiple tags for the development and operation of RFID-based crack sensor arrays were defined. RFID crack sensor arrays are also effective at crack propagation monitoring, as indicated by the significant damage index changes measured in laboratory experimentation. Therefore, the developed system significantly propels the advancement of crack propagation monitoring on steel structures at a low cost, enabling extensive usage on welded cover plates within a restricted budget. Further work to validate the performance of the developed system under environmental uncertainties for field deployment is underway.

This research was sponsored by the Joint Highway Research Advisory Council of the University of Connecticut and the Connecticut Department of Transportation through Project 16-1 of the Connecticut Cooperative Transportation Research Program. This material is also based upon work supported by the Department of Education under the Graduate Assistance in Areas of National Need program (Award Number P200A140212). The contents reflect the views of the authors who are responsible for the accuracy of the information presented herein. The contents do not necessarily reflect the official views or policies of the US Department of Education, the University of Connecticut or the Connecticut Department of Transportation.