Instrumentation for multi-channel vertical electrical impedance scanning of concrete bridge decks

Analog design

The analog subsystem is shown in figure 2. The voltage signal output of the DAC is a 3.3-V_peak-to-peak sinusoid at the desired measurement frequency, typically 100 Hz to 1 kHz. This sinusoid signal is precomputed on the microcontroller, and the voltage output is updated at 100 kHz using DMA. The stair-step output is smoothed by a second-order Sallen-Key reconstruction filter with a corner frequency of 3.1 kHz. The signal passes through a second-order Sallen-Key high-pass filter with a corner frequency of 19.4 Hz to remove the DC offset resulting from the DAC.

Zoom In Zoom Out Reset image size

After passing through the analog filters, the signal passes through a 51-kΩ current-sense resistor. A resistance of 51 kΩ was chosen because it balances the signal-to-noise ratio (SNR) for both the voltage and current signals over a large range of expected concrete impedance values, which are typically 1 kΩ to 1 MΩ. When the resistance is 1 kΩ, the $| {\boldsymbol{V}}\left(\omega \right)|$ signal is approximately 0.064 V_peak-to-peak, and $| {\boldsymbol{I}}\left(\omega \right)|$ is approximately 3.24 V_peak-to-peak. When the resistance is 1 MΩ, the $| {\boldsymbol{V}}\left(\omega \right)|$ signal is approximately 3.14 V_peak-to-peak, and $| {\boldsymbol{I}}\left(\omega \right)|$ is approximately 0.16 V_peak-to-peak. The current flow through the current-sense resistor is converted to a voltage signal using an instrumentation amplifier (INA121P, Texas Instruments) with a gain of 1. A 1.65-V DC bias is added to the instrumentation amplifier in order to center the signal between the 0-V and 3.3-V rails of the ADC.

After the signal passes through the 51-kΩ current-sense resistor, another instrumentation amplifier (INA121P, Texas Instruments) is used to measure the voltage difference between the impedance probe and the low-side port, again applying a 1.65-V offset to the output before sampling with a separate ADC. An additional amplifier in a voltage-follower configuration passes the signal from the probe to the guard ring.

Digital design

In the digital subsystem, the voltage and current signals are synchronously sampled at 100 kHz. Sampling is triggered by a programmable delay block (PDB), and values are stored in a double buffer using DMA. While one buffer is filled with samples, the other buffer is processed by the filtering scheme shown in figure 3. Each box represents a decimating low-pass filter, with the decimation factor labeled in the middle of the box. Information about the passband, stopband, and attenuation of each filter is also given in figure 3. The low-pass filters are necessary to reduce the effects of noise, aliasing, and other channel interference. The sine and cosine in the filtering schematic demodulate the signal at the center frequency [22]. Magnitude and phase are then estimated using in-phase and quadrature components of the signal that have been further processed.

Zoom In Zoom Out Reset image size

This overall signal-processing scheme is applied to both the voltage and current signals so that the complex impedance can be calculated using equation (2). An impedance measurement with an effective bandwidth of 40 Hz is recorded 98 times per second. This data collection rate is more than an order-of-magnitude improvement over that of previous VEI scanning devices, which employed a resistor-switching configuration and only generated measurements approximately four times per second [8].

The analog and digital subsystems resulted in the impedance analyzer shown in figure 4. The impedance analyzer consumes only 1.2 W of power (5 V at about 240 mA of current draw). The impedance analyzer weighs 72 g and is very small, having dimensions of 9 cm by 6 cm by 4 cm.

Zoom In Zoom Out Reset image size

A decade resistor (1434-G, General Radio) with a range from 0.1 Ω to 1 MΩ was used to evaluate the accuracy of the newly developed impedance analyzer. Figure 5 shows the measured voltage and current signals for resistances of 1 kΩ and 1 MΩ at a measurement frequency of approximately 190 Hz. Figure 5 also shows an estimate of the power spectral density of the signal using Welch's method to estimate the power spectrum. The recorded signals were within the desired peak-to-peak voltages for proper ADC measurements, the signal-to-noise ratio was at least 30 dB in the worst case, and the largest harmonic peak (at an integer multiple of 190 Hz) was more than 20 dB below the primary peak (at 190 Hz) in the worst case.

Zoom In Zoom Out Reset image size

An additional three sets of measurements were also obtained. First, resistor values between 0 and 10 kΩ were measured in step sizes of 100 Ω. Second, resistor values ranging from 0 to 100 kΩ were measured in step sizes of 1 kΩ. Third, resistor values ranging from 0 to 1 MΩ were recorded in step sizes of 10 kΩ. Each resistor value was measured for approximately 1 s by the impedance analyzer. The mean and standard deviation of the impedance magnitude for that second were recorded. The scanning frequency for this test was 568 Hz, which was selected because it was approximately in the middle of the scanning frequency range of interest.

The results of this experiment are shown in figure 6. Figure 6(a) shows the measured resistance compared to the actual resistance, with an ideal measurement being represented by the line of equality displayed in the plot area. Figure 6(b) shows the percent error of the measured resistance as compared to the actual resistance. The first seven points equally spaced between 100 Ω and 700 Ω have errors as high as 60% but are not shown since they are outside the designated measurement range. Figure 6(c) shows the estimated standard deviation of the measurement noise using approximately 100 samples collected for each data point. Figure 6(d) shows the standard deviation as a percentage of the actual resistance value. The horizontal axes of each subplot and the vertical axes of 6(a) and 6(b) are presented on a logarithmic scale where a value of 2 represents 100 ohms, a value 3 represents 1 kΩ, and so on.

Zoom In Zoom Out Reset image size

Figures 6(a) and (b) demonstrate that the impedance analyzer accurately estimated actual impedance in the range of 1 kΩ to 1 MΩ, which is the full range indicated in the desired specifications. The accuracy was consistently within 10 percentage points in the impedance range of interest and was usually accurate to within 5 percentage points. Below the low impedance measurement limit of 1 kΩ, the impedance measurements develop a distinct bias due to a non-zero voltage bias in the voltage signal assumed to be caused by noise and possible leakage currents. This bias is insignificant when the voltage signal is large but becomes significant at smaller resistances when the signal is small. Although outside the range of the decade resistor and above the high impedance measurement limit, a similar effect, not shown in this data, happens to the current measurement when impedance values larger than 5 MΩ are measured. Despite bias when measuring relatively low and high impedance values, the impedance analyzer has low variance. As shown in figure 6(d), the standard deviation of the error is often less than 1% of the actual value.

These evaluations demonstrate that the new impedance analyzer meets the desired specifications and is capable of measuring the impedance of a bridge deck to within 10% of the actual impedance within the designed measurement range and at high sample rates. The new design can also drive guard rings and be arranged in a multi-channel configuration. These features should allow the impedance analyzer to be successfully deployed in field studies to measure the protection provided to embedded rebar against chloride ingress through the surface of bridge decks.

A multi-channel VEI scanner was constructed using one impedance analyzer per channel. This scanner, shown in figure 7, consisted of a data collection unit, a multi-channel impedance analyzer, a water delivery system, a localization unit, a moving platform, LAEs, and probes.

Zoom In Zoom Out Reset image size

A pickup truck was used to tow the scanner, as shown in figure 7(a). The data collection unit, pictured in figure 8, was housed in the cab of the truck. The data collection unit was comprised of five time-synchronized, single-board computers that recorded multiple data types; the parallel configuration of the single-board computers allowed for recording of multiple data streams with large bandwidth.

Zoom In Zoom Out Reset image size

A waterproof box in the truck bed housed six impedance analyzers as shown in figure 9. Each impedance analyzer shared a common low-side port connection and was connected to a single USB hub. The hub was connected to the data collection unit via a 5-m USB 3.0 extension cable. Six 7.6-m BNC cables connected the center electrode and guard ring of the probes to the high-side and guard ring ports of the impedance analyzers. The low-side ports were connected with wires to the aluminum frame of the moving platform, which was electrically connected to the LAEs.

Zoom In Zoom Out Reset image size

A 1000-L water reservoir located in the truck bed was the source of water for the water delivery system. The water reservoir was connected to three pumps, each capable of supplying 20 l per minute, that were powered by a 12-V battery. The pumps supplied water to a sprinkler bar through three standard garden hoses, as depicted in figure 10. Multiple sprinklers were mounted to the bar to ensure sufficient application of water to the deck surface during scanning [10].

Zoom In Zoom Out Reset image size

A hitch extension with a localization unit was attached between the truck hitch and the moving platform [23]. The box contained two sideways-pointing light detection and ranging (LIDAR) units and a downward-facing camera, as shown in figure 11. A differential global positioning system (GPS) unit was also used for localization. The GPS receiver was located in the truck cab, and the GPS antenna was placed on top of the cab. The LIDAR units were used to determine the distance of the scanner from the parapet walls for estimating transverse position. The camera was used to identify the exact time the scanner crossed onto the bridge, and the GPS was used to estimate longitudinal position within 1 m [24].

Zoom In Zoom Out Reset image size

The hitch extension was connected to the aluminum frame of the moving platform using a solid steel hitch bar and a double-hinged bracket. A steel post extended 1.2 m upwards from the hitch bar, and an 1130-kg winch (Badland) was attached to the top of the post. The winch was used to lift the platform off the ground for easy relocation during scanning or to fully stow the platform for travel between bridges, as shown in figure 12.

Zoom In Zoom Out Reset image size

The LAEs and probes were attached to the moving platform as shown in figure 7. The front LAE was positioned in the open frame area in front of the line of probes, and the back LAE was positioned directly behind the line of probes. These LAEs created a continuous low-impedance connection to the rebar, even when the scanner was traversing joints between electrically discontinuous spans in a bridge deck [7, 11]. The six probes between the LAEs were attached using flexible connections to the aluminum frame. Two of the probes were attached to optional wings that extended the 2.4-m scanning width to 3.7 m. Each probe used a different measurement frequency to mitigate probe crosstalk. The frequencies used were 190, 272, 347, 431, 500, and 568 Hz. The LAEs, probes, and extension bars were removed when traveling between scanning sites as shown in figure 12(b).

The probes and LAEs were designed to be mechanically robust and also flexible to ensure consistent contact with the bridge deck during scanning. Guarded probes were made of flexible brushes. As shown in figure 13, the center electrode and guard ring of a probe consisted of numerous 26-cm strands of 0.8-mm-diameter 302/304 stainless steel cables folded in half and individually threaded through two separate holes in perforated metal strips. Four of these metal brush strips were arranged into a diamond shape to form the guard ring brush and attached to a plastic frame having dimensions of 0.6 m by 0.5 m by 0.02 m. The center electrode consisted of a separate metal brush strip attached to the center of the plastic frame. This flexible brush design allowed the probe to conform to the contours of the bridge deck surface and maintain a relatively consistent contact area while exhibiting excellent durability even on very abrasive deck surfaces. The diamond shape for the guard ring was preferred over a rectangular shape as it resulted in more uniform bending of the individual metal bristles.

Zoom In Zoom Out Reset image size

The LAEs were designed to slide easily over the bridge surface. The LAEs were sets of 7-mm-diameter galvanized steel cables connected in parallel to 2-m aluminum bars. The bars electrically connected the cables to the frame. The weight and flexibility of the cables allowed the LAEs to conform to uneven surfaces. These cables were oriented in the direction of travel to minimize snagging and collection of debris, which were both significant problems with previous designs. The cables did not exhibit noticeable wear after multiple passes across bridge decks.

For a field evaluation, the VEI scanner was used to scan a concrete bridge deck in northern Utah. This bridge deck, pictured in figure 14, was built in 1995. It has a two-span cast-in-place concrete deck with a bare concrete surface and is supported by continuous steel girders. The bridge deck is 78 m long and 13.6 m wide (as measured from the outsides of the parapet walls) and carries two lanes of traffic on an urban local road with an estimated average daily traffic (ADT) of 5000. The bridge deck is typically subjected to numerous deicing salt applications as part of winter maintenance each year.

Zoom In Zoom Out Reset image size

The field evaluation consisted of five bridge deck scans using the VEI scanner, where a scan included one pass in each lane of the bridge deck. Each pass scanned a width of 3.7 m and was performed at a speed of 5 to 7 km h⁻¹, which required approximately 150 s to complete. During each pass, a sufficient quantity of plain tap water was sprayed onto the deck surface to produce a thin film of water as shown in figure 14(b). The data were recorded using the data collection unit and were processed afterwards by reconstructing the scanning paths and associating impedance measurements with their respective locations.

This bridge deck was of particular interest for the field evaluation because delamination and cracking surveys had already been performed on this deck approximately 8 months prior to the VEI scanning. The high-resolution delamination map was generated through a detailed chain-drag survey that required about 60 man-hours, not including personnel who were responsible for traffic control. Crack mapping was completed more rapidly, with only those cracks that were clearly visible from a standing position, or about 1.5 m from the deck surface, being recorded. Crack mapping required about 2 man-hours to complete. Both the delamination map and crack map were digitized after the surveys were completed.

The maps produced in this study are shown in figure 15. Figures 15(a) to (e) are the VEI maps from the five bridge deck scans, while figure 15(f) shows the delaminations and cracks together on the same plot. VEI is displayed on a logarithmic scale, similar to the values shown in figure 6. In preparation of the maps, linear interpolation was used to compute values for areas between measurement points. (Linear interpolation was also used to replace invalid data collected by one probe that was determined after the scanning to have a faulty connection.).

Zoom In Zoom Out Reset image size

A visual comparison of the VEI maps with each other indicates a high level of consistency among the five passes. Pass 1 appears to have slightly higher VEI values than the subsequent passes, which was expected as the water applied to the deck was still soaking into the concrete; for this reason, pre-wetting the deck or applying multiple passes is recommended to obtain stable readings. Passes 2 to 5 have very similar results, with most discrepancies being attributable to minor spatial offsets and interpolation.

A visual comparison of the VEI maps with the delamination and crack maps, as illustrated in figure 16 for a portion of the deck, also indicates strong correlations. Specifically, almost all of the detected delaminations occur in areas with very low VEI values (<5 kΩ), and the very low VEI values beyond the edges of many delaminations suggest the locations of future delamination growth. Similarly, nearly every crack that was recorded in the cracking survey occurs in an area with low VEI values (5 kΩ to 50 kΩ). Additional linear features also characterized by low VEI values, but not corresponding to a recorded crack, may indicate cracks that formed between the cracking survey and the VEI scans, or they may indicate future crack locations. Because VEI values can be used to quantify the spatial extent and severity of compromised concrete, VEI maps may be more useful than delamination or cracking maps for delineating areas of a concrete bridge deck that warrant repair.

Zoom In Zoom Out Reset image size

Regarding the speed of data collection, the VEI scanning was completed at a rate comparable to that of visual inspection and more than 500 times faster than the chain-drag survey. Regarding data analysis, post-processing of the collected VEI data was automated and required less than an hour to complete. Given that the speed of VEI scanning is approximately the same as that of a street sweeper, nominally 5 to 10 km h⁻¹ [25], VEI scanning could likely be performed on non-highway bridges with minimal traffic control.