Abstract
Guided waves have been used in the past for monitoring corrosion and other damages in plates, rods, pipes, and reinforced concrete structures. Past investigations tried to relate the recorded signal strength to the extent of corrosion or other damage. The main disadvantage of this approach is that the signal strength is also affected by the bonding condition between the sensors and the structure. Over time, this bonding condition is bound to deteriorate, and therefore, one cannot say for sure if the change in the signal strength is due to corrosion or because of the deterioration of the attachment of the sensors to the structure. A new guided wave–based technique is proposed here, which investigates the change in the time of flight of the propagating wave in loaded reinforced concrete structures at various levels of corrosion. Corrosion affects the bonding strength between concrete and reinforcing steel altering the stress level in the reinforcing steel bar in a loaded beam. Since the wave speed is affected by the internal stress, an increase in corrosion level should affect the wave speed in a steel bar and the wave’s time of flight through the bar. The main advantage of the proposed approach is that this result should not be affected by the bonding condition between the sensor and the structure. How delamination at the bar–concrete interface affects the signal strength and the effect of induced corrosion in free bars as well as in bars embedded in the concrete are also investigated.
Introduction
Reinforced concrete (RC) makes up a large part of the civil infrastructures in the United States and abroad. For instance, over half of the bridge inventory in the United States is made of RC. 1 Although a high alkaline environment of the concrete protects steel from corrosion, corrosion does occur, and it is currently one of the primary durability concerns for RC structures. 2 The corrosion of RC is a complex chemical process, and its occurrence is continuously increasing in aging infrastructure. This article presents a completely new method of monitoring corrosion in RC using guided elastic waves. The proposed method is different from the existing guided wave–based methods and should be able to cover longer range while monitoring corrosion in RC structures.
The guided wave–based techniques have been found to be very efficient for damage detection and condition assessments of various aerospace and civil structures.3 –33 One reason for the guided wave–based technique’s popularity is that it can detect damages from relatively large distances unlike conventional electromagnetic, optical, or chemical sensing techniques34,35 for which the probes must be placed relatively close to the inspection region. In spite of many applications of the guided wave–based techniques for damage detection, as some are referred above, in none of the earlier works the technique has been used in the same way as it is performed in this article. This distinction is highlighted in the “Basic principles of current guided wave–based techniques for corrosion monitoring,” “Difficulties associated with the current guided wave–based techniques,” and “Proposed new technique for corrosion monitoring using guided waves” sections.
Basic principles of current guided wave–based techniques for corrosion monitoring
The current guided wave–based techniques for sensing corrosion in steel bars are based on the following two simple principles:
Corrosion makes the surface of the reinforcing steel bars rough as shown in Figure 1 and therefore should affect the guided wave propagation characteristics. This dependence has been shown by Miller et al.30,31 in Figure 2. Miller et al. placed the transmitter and the receiver on two ends of the four steel bars of equal length (3 ft or 914.4 mm) that have varying degrees of corrosion as shown in Figure 1 and compared the received signal strength.
Corrosion eventually causes delamination or separation of steel rebar from the concrete affecting the strength of the propagating wave due to delamination. To investigate the effect of delamination on the propagating wave strength, Miller et al.30,31 fabricated test specimens with artificial separations of various lengths between concrete and reinforcing rod. Their specimen dimensions are shown in Figure 3 and the recorded signal strengths in Figure 4.
Four steel bars showing different degrees of corrosion—uncorroded case (bottom bar) to highly corroded case (top bar).
Dimensions of steel rod reinforced concrete specimens—length of separation or delamination between steel rebar and concrete varies from 0″ (no separation) to 18″ (75% separation).
Received guided wave strength after the wave propagates through four reinforced concrete specimens with different degrees of separation between steel rod and concrete. Specimen dimensions are shown in Figure 3. Higher level of separation increases the signal strength because less energy can leak into the surrounding concrete as the concrete is detached from the rod.30,31
Difficulties associated with the current guided wave–based techniques
From Figures 2 and 4, it is evident that increasing the corrosion level in a steel RC structure can either decrease the strength of the recorded guided waves (Figure 2) due to increasing surface roughness or increase the recorded signal strength (Figure 4) because of increasing separation between steel and concrete. Therefore, during the corrosion progression when both these phenomena can occur simultaneously at different stages of corrosion, the net signal strength may either increase or decrease depending on which effect is stronger at that time or may remain unchanged if the two opposing effects cancel each other.
To study which effect dominates, some investigators have caused corrosion in steel rebars placing them inside the concrete and studied the strength variations of transmitted guided wave modes as the corrosion progressed and concluded that at different stages of corrosion, the signal strengths vary differently.32,36 Three difficulties associated with the corrosion detection by monitoring the strength of the recorded signals are listed as follows:
First, a change in the signal strength does not necessarily always imply increasing corrosion as explained earlier.
A second difficulty is illustrated in Figure 5. This figure shows two possible types of energy profiles inside the rod that can be observed in propagating guided wave modes as the wave energy travels through a steel rod from transmitter T to receiver R. These two types are denoted as modes A and B. For mode A, the energy profile is such that most of the energy propagates near the circumference of the rod, while for mode B, the energy is confined near the core or central axis of the rod. It should be noted that both longitudinal and flexural modes can be of type A or B. Also, the same mode can be converted from type A to B or vice versa by frequency tuning. Researchers have observed that the guided wave mode that predominantly excites the circumference of the rod (type A in Figure 5) is more sensitive to the corrosion damage at the rod–concrete interface, while the mode that has higher level of energy concentrated at the core or near the central axis of the rod (type B in Figure 5) is less sensitive to the concrete–steel interface corrosion but propagates longer distances through the rod. Therefore, although a mode of type A is more sensitive to the interface condition, the energy profile of a mode of type A causes more energy leaking into the surrounding concrete resulting in a higher attenuation of this mode; a less sensitive mode of type B, however, can propagate a longer distance through the rod due to low level of energy leaking into the concrete.
Besides the problem of energy leaking into the surrounding medium, another major difficulty associated with the propagation of the wave mode of type A is that when the plain steel bar is replaced by a rebar, this mode finds it even harder to propagate because of its sensitivity to the surface texture of the rebar. Therefore, the dilemma here is whether to choose a mode of type A or type B. Energy profile of a mode of type A makes it more attractive for sensing the interface condition, but it adversely affects its ability to propagate through a rebar because of its nonuniform surface texture. Surface corrugation scatters away the propagating wave energy increasing its attenuation significantly.
Two types of guided wave modes A and B can propagate from the transmitter T to the receiver R. Energy profile of mode type A not only makes it more sensitive to the steel–concrete interface condition (e.g. corrosion) but also allows more energy to leak into the surrounding concrete resulting higher attenuation. The situation is reverse for mode type B.
Proposed new technique for corrosion monitoring using guided waves
To let the guided wave travel a long distance through a corroded rebar, one needs to select wave modes that have the energy profile as shown in type B of Figure 5. This mode being less sensitive to the concrete–steel interface condition can propagate a longer distance but is less effective in its corrosion detection capability. This shortcoming of low interface sensitivity of mode type B can be overcome by loading the specimen as described here. It is suggested that the concrete beam that contains the steel bar whose corrosion we are monitoring should be loaded as shown in Figure 6. If the steel rod is placed away from the neutral axis of the beam, then it should be axially stressed due to the applied load P. If the bonding between concrete and steel is perfect (no slippage at the interface), then the steel rod will be subjected to the maximum axial stress. When this bonding deteriorates due to corrosion, then the axial stress in steel decreases due to slippage at the interface. Since the wave speed varies with the applied stress the time of flight (TOF) of the guided wave propagating from the transmitter T to the receiver R should depend on the stress level in the rod. Note that the velocity of all types of propagating wave modes—both types A and B of Figure 5—should be affected by the axial stress in the rod. Therefore, if the specimen is loaded, then although a mode of type A is affected by the corrosion, it is not necessary to select it for corrosion monitoring since it attenuates fast. Instead, it is advisable to select a mode of type B that can propagate a long distance through the rod and at the same time is sensitive to the corrosion when the beam is loaded.
Reinforced beam is loaded by a transverse load P. Guided wave is sent from one end of the rod to the other end through the steel rod of this loaded beam.
Experimental investigation
The success of the proposed new approach depends on how reliably one can record the change in TOF due to the variation in applied load. The following two sets of experiments were carried out to investigate this issue. The first set of pilot experiments for which the bar was corroded outside the concrete was performed to investigate the feasibility of the proposed approach; then, a second set of experiments was carried out for which the bar was corroded while it was inside the concrete—we will call it in situ corrosion.
Experimental setup for the first set of pilot experiments
As shown in Figure 7, the experimental setup consisted of three equipments: a computer (shown on the left), a power amplifier (ENI 1040L, shown in the middle), and a Handy-scope HS-3 (shown on top of the power amplifier) provided by Analog Speed Instruments (ASI; Germany). The Handy-scope HS-3 is a computer-controlled measuring instrument that integrated the functions of the following instruments—oscilloscope, arbitrary waveform generator, and data logger. The waveform generated by the Handy-scope HS-3 was amplified by the ENI 1040L power amplifier (maximum output: 400 W, peak-to-peak voltage: ±1.5 V, it could operate between 10 and 500 kHz with 55-dB gain) and then sent to the transmitter. The sensor or receiver placed on the other end of the specimen picked up the propagated signal. Broadband Conrad transducers (EPZ-35MS29, item no. 712943-07) were used as transmitters and receivers (www.CONRAD.de). These thin circular transducers were fabricated by attaching a ceramic disk of 25 mm in diameter to a metal disk of 35 mm in diameter. A Chirp signal with the starting frequency of 10 kHz and final frequency of 100 kHz was used to excite the transmitter. The wave generator triggered the signal with 12-V amplitude and was recorded with a sampling rate of 50 MHz. The received signal, after being routed through an amplifier, was collected by the scope and displayed on the computer screen. The amplitude and time of arrival of the first wave peak were selected and analyzed. The changes of the signal arrival time in comparison to the baseline data, as the load was applied, were correlated and reported. These changes were recorded as the difference in the TOF (along the y-axis) versus the clock time of the computer (along the x-axis) during the duration of the experiment. The received signal data were acquired by the Handy-scope HS-3 and transmitted to the computer. In this experiment, a LabVIEW-based virtual-control platform was also adopted for the excitation signal setting and data recording. The concrete prisms shown on the right of the power amplifier were reinforced by the steel rod as illustrated in Figure 3. The schematic diagram of the experimental setup is shown in Figure 8.
Photograph of the instrument used in the experiment.
Schematic diagram of the experimental setup.
First set of experimental results
A 3-ft-long (0.91-m) free steel rod in the absence of any concrete (see Figure 1) was simply supported at its two ends and loaded at the midpoint as shown in Figure 9. A 1.54-lb (700-g) hanger was placed at the midpoint of the steel rod and then a load of 2.2 or 11 lb (1–5 kg) was applied and removed with an increment of 2.2 lb (1 kg). The load variation with time is shown in Table 1. The complete loading/unloading cycle took 17 min as shown in Table 1. The change in TOF as a function of time is plotted in Figure 10. The TOF variation was measured by the cross-correlation technique applied to the receiving signals for unloaded and loaded rods. Note that as soon as the hanger was placed on the rod, the TOF was reduced by 10–12 ns. As more weights were placed on the hanger, the TOF was reduced further, and when the load was removed, the TOF went back to its previous level. Clearly, the small variations of TOF due to the applied load could be experimentally detected for the free rod.
A steel rod is loaded by a transverse load P by placing weights on the 1.54-lb (700-g) hanger. Guided wave is sent through the rod from transmitter T to receiver R.
Applied load as a function of time
H: 1.54-lb (700-g) hanger.
TOF variation for a plain steel rod during loading–unloading cycles described in Table 1. Figure 9 shows how the load was applied on the steel rod. Horizontal axis shows the clock time, and the vertical axis shows the TOF (ns) relative to a reference time. Negative sign on the vertical axis implies earlier arrival than the reference time.
The corroded and noncorroded steel rods were then placed inside the concrete (see Figure 3). The rebar was corroded in open air and then the concrete was poured around the rebar. The specimens thus fabricated were then loaded at the midpoint, as shown in Figure 6, up to 125 lb (56.8 kg) with an increment of 25 lb (11.4 kg) and then unloaded to 0 at 25 lb steps, as illustrated in Table 2. The loading–unloading cycles took 11 min. The TOF variations for corroded and noncorroded rebars placed in the concrete beams are shown in Figure 11(a) and (b), respectively. Note that the TOF increases by almost 35 ns for the corroded rebar and 22 ns for the noncorroded rebar. The vertical axis scales in Figure 11(a) and (b) are different. Since the vertical scale has more magnification in Figure 11(b), the experimental noise is also magnified in this plot. The rough surface of the corroded bar produced a good bonding between concrete and bar, and thus, a better stress transfer occurred from the concrete to the rebar for the corroded case causing relatively larger variations in the TOF measurement.
TOF: time of flight.
Variation of TOF of the guided wave propagating through the steel rebar placed in the reinforced concrete beam for the load variation shown in Table 2: (a) corroded rebar and (b) noncorroded rebar.
Experimental setup and specimen preparation for the second set of experiments for monitoring in situ corrosion
Next, it was investigated how the TOF would change with loading at different corrosion levels if the reinforcing bar was corroded inside the concrete. To this aim, 5″ × 5″ × 24″ (127 mm × 127 mm × 610 mm) concrete beam was cast with one reinforcing steel bar placed eccentrically, away from the neutral axis of the concrete beam. The reinforcement was a number 6 rebar that has a diameter of 0.75″ (19 mm) and was placed with 1″ (25.4 mm) of the concrete cover. In practice, the concrete structural members are required to have at least 2″–3″ (51–76 mm) minimum cover of the reinforcement. In this experiment, the concrete member was relatively smaller and the required 2″ cover would have placed the rebar very close to the neutral axis of the beam, and therefore, only 1″ concrete cover was chosen. The dimensions of the RC specimen are shown in Figure 12. The reinforcing steel bar was 36″ (91.4 cm) long and had a perpendicular rib pattern meaning ribs ran perpendicular to the axis of the bar. The bar was grade 60 with the expected minimum yield stress of 60,000 lbf/in2 (414 MPa). The rebar had 6″ (15.2 cm) exposed length at both ends of the concrete beam for the ease of transmitter/receiver attachment.
Specimen geometry for in situ corrosion monitoring.
The concrete used for casting the specimen was a product from Sakrete, High Strength Concrete Mix in the 80-lb (36.4-kg) bag. The product was a mixture of Portland cement, washed and graded sand, and gravel, which met the ASTM C 387 requirement with the expected compressive strength of 4000 lbf/in 2 (27.6 MPa) after 28 days. The specimens were cast in the wooden mold and then they were placed in the concrete laboratory moisture room for 28 days for curing.
The concrete specimen was then immersed in the corrosion-inducing solution for producing accelerated corrosion as shown in Figure 13. After the corrosion-free RC members were placed in the corrosive environment, corrosion attacks took place at the reinforcing steel. The corrosion process developed locally and then globally on the surface of the reinforcing steel. The corrosion rates depended on how aggressive the environment was. With an attempt to mimic the natural corrosion process, “healthy” RC specimens, prepared for this study, were placed into the corrosion-inducing system. It should be noted that the corrosion process in real structures occurs at a much slower rate than the forced corrosion produced in the laboratory environment. As shown in Figure 13, two plastic containers (44″ L × 20″ W × 6″ D and 34″ L × 16″ W × 6″ D) were used. The inner (smaller) container held the corrosive solution and the specimen, while the outer (larger) container captured any leakage or spillage of the solution. Slots were cut on the end wall of the small container to make room for the rebar and were sealed with Play-Doh. The corrosion-inducing solution was a mixture of 10 lb of Morton pool salt, 30 lb of desert soil, 0.5 gallon of Kem-Tek Chlorinating Liquid, and 8 gallons of tap water. The corrosive solution was filled and maintained at the same level (at the top of the rebar) as closely as possible throughout the experiment. The concrete beam specimen was placed on two aluminum plates of 1/4″ (6.4 mm) in thickness, 3″ (76.2 mm) in width, and 6″ (152.4 mm) in length (one near each edge of the concrete, as shown in Figure 13) for the in situ corrosion load tests.
In situ corrosion setup.
A current of magnitude 1.5 A was induced and maintained across the solution and the specimen for generating accelerated corrosion. The anode was a galvanized wire mesh with no. 2 wire at 1/2″ on center that wrapped around the concrete specimen, and the cathode was a copper pipe (Figure 13). The power source (a low-voltage direct current regulated power supply) was used to generate the electric current. The power source was a product of HQ Power, model PS1502AU, which generated a constant current of 1.5 A, while the voltage varied from 0 to 15 V.
The loads were placed directly on top of the beam (resting on two aluminum plates) at the mid-span location. Five different weights, varying from 50 to 250 lb (22.7–113.6 kg) with increments of 50 lb (22.7 kg), were used to load the specimen.
In order to minimize the disturbance to the specimen, the specimen was loaded while still in the corrosion-inducing solution, and the TOF variations were recorded as the specimen was loaded. Each desired load was placed on the specimen for a duration of 30 s. The incremental load was then applied for the same duration. After reaching the maximum load, the weights were removed from the beam, at the same rate—at 30-s intervals. Table 3 shows the applied load versus time for the loading–unloading cycle of the RC specimen.
Applied load versus time for the concrete beam with reinforcing steel bar for in situ corrosion monitoring
The applied load of 250 lb (113.6 kg) is well below the failure load for the concrete; therefore, the specimen does not fail at these loads. First, this load meets the nondestructive testing (NDT) criterion that the inspection technique should not cause any damage to the structure. Second, the required load, or in other words, the required stresses in order to observe the changes in TOF, is relatively small, which makes it more manageable—no bulky load application machine is needed. The changes in the signal arrival time (or TOF) can be observed for the reinforced specimen as described in the next section.
Experimental results for the second set of experiments for monitoring in situ corrosion
Figure 14 shows how the received signal strength decays as the corrosion progresses. Four plots of Figure 14 show received signals (white signals in the black screen on the right side) after 19, 35, 49, and 73 days of induced corrosion. At regular intervals, the specimen was loaded and unloaded as illustrated in Table 3, while it was still in the corrosion-inducing system. The signals as shown in Figure 14 were recorded daily but only four signals are displayed here. However, the load tests (loading and unloading the specimen following Table 3) were carried out less frequently with 1–2 weeks of interval between two consecutive load tests. Figure 15 shows a typical load test result for the corroded specimen. This plot was generated after corroding the specimen for 49 days. Note that the TOF changed by about 40 ns (from −5 ns at the start of the load test to +35 ns when a load of 250 lb (113.6 kg) was applied). This gives an average jump of 8 ns (40/5) in TOF for every load increment of 50 lb (22.7 kg). Clear jumps can be seen in this figure every time a load of 50 lb (22.7 kg) was placed on the beam. However, the jumps in the TOF during the unloading process were not much clear. Although the jumps were not clear, the curve gradually came down approximately to the initial level (before the loading–unloading cycle was applied).
Received signals after the specimen was submerged in the corrosive solution for (a) 19, (b) 35, (c) 49, and (d) 73 days.
Load test result for the reinforced concrete specimen after it was corroded for 49 days. Time of arrival changed 40 ns (from −5 to +35 ns) for a total of applied load of 250 lb (113.6 kg); a sharp increase of TOF every time a load of 50 lb was applied can be seen on the left side of this plot.
Figure 15 showed a continuous change in TOF even during the time durations when the applied load was held constant. Nonequilibrium nonlinear dynamics causes this variation. Other investigators37 –39 have shown that when rocks, concrete, cement, damaged materials, sintered metals, and granular materials are disturbed specifically by waves, it takes them tens of minutes to hours to return to their fully equilibrium states. This is called nonequilibrium nonlinear dynamics or slow dynamics. The nonequilibrium dynamics is observed in materials having mechanically “soft” inclusions in a “hard” matrix. 40 The cracks in a solid can introduce nonequilibrium dynamics. To achieve a full equilibrium state, the setup was required to be left in the room for hours after reducing the corrosive solution level and before recording the experimental data since fluids act to modify the internal forces in porous media and thereby significantly influence the nonlinear behavior. As a result, the TOF of the propagated signal had the tendency to continuously adjust itself to the stable or full equilibrium state. However, several hours of setting time was not provided before each reading to complete the experiment within a reasonable amount of time. As a consequence, the signal was in the self-adjusting mode while recording the experimental data, and as a result, an inclined curve was displayed. However, even from this inclined curve, the TOF change due to the load increment could be easily calculated, about 8 ns for every load of 50 lb (22.7 kg), as discussed above. It was then investigated how this TOF variation with loads was affected as the corrosion progressed. The results are shown in Table 4 and plotted in Figure 16. The recorded data (Table 4 and Figure 16) for the lateral load tests (Figure 6) on the corroded specimen show a consistent trend in the relationship between the degree of corrosion, signal loss, and the variation in the TOF. Both signal loss and TOF variations increased with the degree of corrosion.
TOF variation caused by the applied lateral load at the mid-span
TOF: time of flight.
Percent changes in the signal loss and the TOF variations (relative to the noncorroded case) for the reinforced concrete specimen are shown as a function of the corrosion level. The number of days the specimen was submerged in the corrosive solution is shown along the horizontal axis, while the percent changes are plotted along the vertical axis.
The signal amplitude listed in Table 4 is the peak value recorded during the load test experiment. The signal loss was calculated relative to the received signal peak amplitude for the noncorroded specimen. The strength of the signal amplitude from the corroded specimen as a percent of signal strength from the noncorroded specimen was first determined. The signal loss percent was then calculated from the difference between the signal strength from the noncorroded specimen (100%) and the corroded specimen (<100%). The total difference in TOF, as a load of 250 lb (113.6 kg) was applied, was calculated by subtracting the starting value (which is −5 ns in Figure 15) before the load application from the final value (35 ns in Figure 15) after a load of 250 lb (113.6 kg) was applied. Alternately, from every sharp jump, the instantaneous change in TOF for a load increment of 50 lb (22.7 kg) can be calculated. The values of the five sharp jumps seen in Figure 15 are 15, 8, 15, 15, and 13 ns. Therefore, the total jump is about 66 ns (= 15 + 8 + 15 + 15 + 13) for a load increment of 250 lb (113.6 kg). About 40% of this value is lost due to slow dynamics effect in the specimen in 2 min, and only 40-ns difference between the received signals before and after the loading process is noted. For all recordings, similar slow dynamics effect was observed during the 2 min of loading time. Therefore, for relative comparisons, the slow dynamics effect should not make any difference as long as we are consistent—that is, if we either consider or ignore the slow dynamics effect for all cases.
Figure 16 shows how the time of arrival (or TOF) changes as the RC specimen is subjected to bending loads at different corrosion levels. This graph was generated from the data shown in Table 4. Note that both the signal loss and the TOF variation increase with the corrosion level in a very similar manner. The slopes of both curves continuously increase, and after 73 days, the rate of increase (or slope) has the highest value indicating fast progression of corrosion.
Although the corrosion can be predicted from both curves of Figure 16, the TOF curve has a significant advantage over the signal strength loss curve—the bonding condition between the sensor and the specimen does not affect the TOF curve. Therefore, if this bonding deteriorates over time, the TOF curve remains unchanged while the signal strength shows some variations creating a false alarm for corrosion. Only when the bonding glue completely fails and the sensor is fully detached from the specimen, then the TOF curve disappears and so does the signal loss curve.
Concluding remarks
Although some investigators in the past tried to detect corrosion in the RC structures by measuring the change in amplitude of propagating guided waves due to corrosion, in this article, for the first time, the degree of corrosion is related to the change in the TOF of the propagating waves as the RC beam is loaded laterally. The main advantage of the TOF-based monitoring system is that it is not sensitive to the bonding condition between the sensors and the specimen. The experimental results presented here verify that both material attenuation or recorded signal strength and TOF variation are affected by the degree of corrosion. Since the proposed TOF-based technique has the added advantage of not being sensitive to the bonding condition between the sensor and the structure, this technique is more desirable. The experimental results presented here show that delamination at the bar–concrete interface increases the recorded signal strength, while the induced corrosion in free bars as well as bars embedded in the concrete reduces the signal strength.
For in situ applications, the transducers can be mounted inside the concrete at two ends of the rebar, and TOF changes can be monitored as the live loads, such as passing vehicle loads in a bridge, are applied. The TOF shift will be different for corroded and noncorroded rebars; from this shift, corrosion level in the rebars can be obtained.
Footnotes
Acknowledgements
Consultation with Dr Abhijit Mukherjee of Thapar University, India while designing the corrosion-inducing setup is gratefully acknowledged.
Funding
This research was partially supported from an NSF grant (OISE-0352680) of USA and the senior scientist award to the second author from the Alexander von Humboldt Foundation of Germany.