Wei Liu et al. 1

Nondestructive Corrosion Monitoring of Prestressed HPC Bridge Beams Using

Time Domain Reflectometry

Wei Liu, Robert Hunsperger,

Dept. of Electrical & Computer Engineering, Univ. of Delaware, Newark, DE 19716

Michael Chajes, Degang Li,

Dept. of Civil & Environmental Engineering, Univ. of Delaware, Newark, DE 19716

Eric Kunz

VETEK Systems Corp. 6 Oak Road, Elkton, MD 21921

Abstract. A novel nondestructive evaluation technique for corrosion detection of embedded or encased steel

reinforcement or bridge cables using time domain reflectometry (TDR) is developed and demonstrated in

this paper. TDR has traditionally been used to detect faults in electrical transmission lines. It involves

sending an electrical pulse along the line and analyzing the echoes. By applying a sensor wire alongside of

steel reinforcement (such as a prestressing strand), a transmission line is created. Both analytical models

and small-scale laboratory tests have shown that TDR can be effectively utilized to detect, locate and

identify the extent of damage in steel reinforcement in this manner. After reviewing the TDR theory and

presenting laboratory results, the paper will discuss an actual field implementation of this technology. A

TDR system has been has been installed in Bridge 8F, a prestressed high-performance concrete (HPC)

adjacent box beam bridge recently constructed in Fredrica, Delaware. The primary purpose of the TDR

installation was to demonstrate that the system can be successfully applied to actual structures. After the

TDR instrumentation was installed, initial readings were taken to verify that the system had survived the

fabrication process. Differential comparisons of subsequently measured signals to the baseline readings can

reveal damage to the strands. Initial experimental results from this project are reported in this paper.

Wei Liu et al. 2

INTRODUCTION

In 1999, the Delaware Department of Transportation (DelDOT) received funds through the Federal

Highway Administration’s Innovative Bridge Research and Construction Program (IBRC) to design and

construct a high-performance concrete bridge in Fredrica, Delaware (later referred to as Bridge 8F). HPC is

concrete that is optimized for a specific application and often possesses qualities such as high strength, low

permeability, good workability, and excellent long-term durability. A two-span, prestressed concrete,

adjacent box beam bridge utilizing HPC in both the beams and deck was designed. The beams for the

bridge were fabricated by Concrete Building Systems (CBS) in Delmar, Delaware. Prior to casting, long-

term monitoring instrumentation was installed in several beams. Bridge 8F, which replaced a deteriorated

four-span structure, was completed in October 2000.

The performance of the HPC beams is being evaluated through long-term monitoring. The monitoring

began at the time of fabrication, and is continuing now that the bridge is in-service. One of the many

aspects of the long-term monitoring program is corrosion detection using time domain reflectometry

(TDR). In this case, if the HPC performs as expected, TDR measurements will confirm that no corrosion

damage is occurring.

ANALYTICAL MODEL AND LABORATORY RESULTS

Time Domain Reflectometry

Time domain reflectometry is a proven method of detection of physical problems in electrical and

optical transmission lines. A transmission line is a wave guiding system along which electromagnetic

waves can travel. It typically has at least two parallel conductors. Examples are telephone lines and

television cables. The key difference between transmission lines and conventional circuits is the size. A

transmission line is long compared to the signal wavelength. As a result, signals cannot travel

instantaneously from one end to the other, because there will be a propagation delay and phase change.

TDR is a well-established technique in the field of electrical engineering that has been used since

1940’s to detect faults in transmission lines. It involves sending an electrical pulse along the transmission

line and using an oscilloscope to observe the echoes returning back from the line. Any discontinuity will

Wei Liu et al. 3

cause a reflection. From the transit time, magnitude, and polarity of the reflection, it is possible to

determine the spatial location and nature of the discontinuity. (1)

A time domain reflectometer is usually configured as shown in Figure 1. The pulse generator generates

a fast rising step wave or pulse. This wave is launched into the transmission line. A high impedance

oscilloscope is connected to monitor the wave.

Bridge Cable Modeling

There are obvious similarities between transmission lines and steel strands used for prestressing

concrete and for cables in cable supported bridges. A steel strand is a good conductor embedded in a

dielectric (concrete). It can be modeled as an asymmetric, twin-conductor transmission line by adding a

sensor wire alongside of the cable. The cross sectional view of the strand/wire system is shown in Figure 2.

Physical defects of the strand, such as abrupt pitting corrosion, general surface corrosion, and voids in the

surrounding concrete, will change the electromagnetic properties of the line. These defects, which can be

modeled as different kinds of discontinuities, can be detected by TDR. (2)

For a thorough analysis of the wave propagation in this transmission line, one needs to solve Maxwell's

equations with boundary conditions imposed by the physical nature of the system under investigation. It is

also possible to represent a line by the distributed parameter equivalent circuit, as shown in Figure 3, and

discuss wave propagation in terms of voltage and current. The distributed parameter equivalent circuit

possesses a uniformly distributed series resistance R, series inductance L, shunt capacitance C, and shunt

conductance G. (R, L, C, and G are defined per unit length.) By studying this equivalent circuit, several

characteristics of the transmission line can be determined.

The characteristic impedance of the line, Z0, is given by

The propagation constant, γ, which defines the phase shift β and attenuation α per unit length, is.

CjGLjR

Zωω

++

=0

( )( )CjGLjRj ωωβαγ ++=+=

Wei Liu et al. 4

ω is the radian frequency, which is given by ω=2πf, where, f is frequency in Hz. The velocity at which

the voltage travels down the line can be defined in terms of β:

The distributed parameters of the transmission line can be calculated from the geometry and material

parameters of the cable. After making several necessary simplifications, the characteristic impedance is

given to a high degree of accuracy by the simplified expression (3)

The characteristic impedance of the line is a function of a, b, and d. Note that b is much smaller than a

and d (see Figure 2), and it remains the same value along the line. However, the radius of the steel cable, a,

may be changed if corrosion occurs. When b<<d,

This expression has a negative value. This means that the characteristic impedance will increase for a

small decrease of a. Since radius a always decreases at a corrosion site, corrosion will cause higher

characteristic impedance. This change of impedance can be detected by time domain reflectometry.

In order to utilize TDR to detect corrosion, the damage sites of a strand need to be modeled as electrical

discontinuities in a transmission line. Several physical defects are of great interest when considering the

durability of steel strands. Among them are abrupt pitting corrosion, general surface corrosion, and voids in

the grout.

General surface corrosion tends to reduce the radius of the strand on the order of a few percent over a

part of length of the line. It is modeled as a section of transmission line with different characteristic

impedance. The extent and length of the corrosion can be determined from the magnitude and duration of

the reflection, respectively. Abrupt pitting corrosion is a severe localized damage. Its length is small

compared to wavelength of the excitation signal. Therefore, it is modeled as an inductor in series with the

−−= −

abbadZ

2cosh

21 222

10 ε

µπ

22

220 1

21

adad

adadZ

−+−≈

εµ

π

βω=pv

Wei Liu et al. 5

line. A positive reflection from the site of pitting corrosion is expected. The location of the corrosion site is

obtained from the transit time. The reflection amplitude indicates the magnitude of the damage. Voids in

the concrete tend to reduce the dielectric constant and therefore increase the characteristic impedance. Also,

voids will change the velocity of propagation in the transmission line. Therefore, voids can be detected by

TDR, as well as corrosion.

Small-scale Laboratory Tests

Small-scale laboratory tests have been conducted to verify the effectiveness of TDR in locating and

characterizing simulated corrosion sites. Several 1-meter and 3-meter specimens, made from standard rebar

and seven-wire strand, with built-in defects were used to study the ability of TDR to detect damage sites.

TDR tests were performed on both grouted and ungrouted specimens. Preliminary testing indicated that

damage detection for the ungrouted specimens was very similar to that of grouted specimens since the

relative reflection magnitude did not depend on the dielectric constant of a uniformly filled medium.

Because the researchers have access to the damage site when measuring specimens that are not embedded

in concrete, such specimens are more convenient to use to study the electromagnetic properties of the

simulated corrosion. As a result, bare specimens were mainly used in small-scale TDR tests.

The specimens were connected to the time domain reflectometer through standard 50Ω coaxial cables.

The far end of the specimen was connected to a terminating resistive load. A pulse was then sent down the

sample and the reflections shown on the oscilloscope. The terminating load was changed from an open to a

short to determine where the end of the sample was. The propagation velocity was then calculated.

Figure 4 shows the TDR reflection from a 3-meter steel rebar sample. This sample has 50% pitting

corrosion in the middle (1.55m from the front end). Pitting corrosion was simulated by locally grooving the

rebar specimens. The TDR measurement was made on a bare specimen (surrounded by air and not

embedded in concrete). The first step in the waveform corresponds to the generation of the step wave (point

A). The wave is launched into a coaxial cable, which is used to connect the sample to the measuring

system. The characteristic impedance of this coaxial cable is 50Ω . However, the sample has higher

impedance. As a result, there is a positive reflection at the beginning of the sample (point B). The wave

travels down the line at vp, the velocity of propagation. At every point that the excitation signal crosses, the

transmission line equations must be obeyed. However, there is a simulated corrosion site at point C. The

Wei Liu et al. 6

physical damage changes its electromagnetic properties. Therefore, the transmission line equations are not

satisfied and a reflection is generated at this point. The reflected wave is separated from the incident wave

in time. This time, T= TC - TB, is the transit time from point B to the mismatch and back again. At the end

of the sample, the wave goes up because the line is terminated by an open circuit (point D). The time

interval between points B and D is 23.0ns, which gives a propagation velocity of 2.61x108 m/s, i.e. about

87% of the speed of light. The location of the damage site is determined as 1.58m from point B since TC -

TB =12.1ns, which is the correct location.

TDR can not only locate the corrosion, but also reveal the severity of corrosion. Figure 5 shows TDR

returns from two seven-wire steel strand samples. The strands are 0.95m long and 1.27cm (1/2 inch) in

diameter. Corrosion was simulated by severing several wires of the strand specimen. The damage was

produced over a 7.5cm length, 44cm from the end of the sample. The TDR readings were taken on bare

strands. The first marker indicates the initial reflection from the front of sample, and the third marker

indicates the reflection from the end of sample 9.94ns later. The propagation velocity is 1.91x108 m/s. The

initial reflection is positive compared to the baseline before the peak, which shows that the characteristic

impedance of the sample is larger than 50Ω . The impedance is measured as 56Ω . It is close to 52Ω , which

is predicated by our model. Because the sample was terminated by a short circuit, the reflection from the

end of the sample is negative. The second marker indicates the reflection from the simulated corrosion site.

Note that an accurate location is identified. Experimental results indicate that the magnitude of the

reflection depends on the severity of the damage. The sample on the right has severe damage in which six

strands are severed, while the other sample has two severed strands.

The sensitivity and accuracy of TDR measurement depends on several parameters. A TDR parametric

study has been conducted. The following parameters have been identified through small-scale laboratory

tests:

• diameter of the sensor wire,

• distance between sensor wire and steel element,

• relative position of the sensor wire and damage site,

• system rise time of the measuring system, which describes how fast the signal is,

• water content of the surrounding concrete.

Wei Liu et al. 7

These factors need to be considered before TDR installation and measurement. (4)

Differential TDR Measurements

In field applications involving complex structures like an actual bridge, noise will be present in the

TDR measurements. One type of noise is random noise. Whether in the lab, or in the field, small amounts

of random noise will be present. However, it has been shown that one can average the results of several

measurements to effectively mask the random noise.

Another type of noise in the signal can be created by

• electric field disturbance caused by steel components near the cable being tested,

• variations in d, the distance between the steel cable and the sensing wire, since the characteristic

impedance depends on d.

While noise created due to these reasons can be relatively large, it is repeatable. Once a concrete girder is

instrumented, the location of the steel components causing noise, and the distance d between the steel

strand and sensing wire will remain constant. Therefore, the noise will be repeatable. Differential TDR

measurement can be used to effectively distinguish corrosion sites from repeatable noise. If several TDR

measurements are made for the same strand over a long time period, the later TDR results should be

identical to the former ones except for the corrosion sites. A differential comparison of stored signals with

newly measured ones can reveal corrosion that occurred between the two measurements. The differential

TDR method has been tested experimentally. Figure 6 shows TDR results obtained from a 1-meter seven-

wire strand bare sample. This sample has two severed strands over a 4.0cm length, 48cm from the front end

of the sample. From waveform 1, it is hard to tell whether or not the sample is damaged and where the

damage is. However, if this waveform is differentially compared with waveform 2, which is the TDR return

obtained from the same sample when it did not have any electrical discontinuities, the damage site can be

easily identified.

BRIDGE 8F EXPERIMENTAL PROGRAM

The effectiveness of the TDR corrosion detection method has been proven through laboratory tests. In

order to ready this technology for field implementation, full-scale experiments and field demonstrations are

Wei Liu et al. 8

necessary. Bridge 8F is the first field demonstration of the TDR corrosion monitoring technology. The

primary purpose of the demonstration was to show that the TDR system was robust enough to survive the

fabrication and erection process, and to show that usable signals could be achieved. While no corrosion is

anticipated in the HPC bridge in the near future, the repeatability of the signals over time will help to

validate the technology and confirm the performance of the HPC.

The complete experimental program involving Bridge 8F consists of two parts. First, a material testing

program has been initiated to determine the properties of the HPC used for the bridge beams. This program

involves a series of material tests conducted over a one-year period. Next, the performance of the HPC

beams is being evaluated through long-term monitoring. One of the many aspects of long-term monitoring

is corrosion detection. Corrosion monitoring with TDR began at the time of fabrication, and is continuing

now that the bridge is in service. Important aspects of the experimental program are provided in the

following sections.

History and Details of Bridge 8F

The existing bridge consists of four 8.67 m (28 ft. 5 in) spans supported by concrete encased I-beams

resting on solid concrete piers and abutments. The bridge was built in 1920 and rehabilitated in 1992. It

was classified as structurally deficient due to considerable spalling and salt infiltration of the piers,

abutments, and fascia beams. Soil conditions (clay layers beneath the marsh), have caused the structure and

approach roadway to settle over the years. These problems led DelDOT to pursue the replacement.

The HPC replacement bridge consists of two 19.0 m (62 ft. 4 in) spans consisting of adjacent

prestressed concrete box beams with a concrete deck to make the structure continuous for live load (see

Figure 7). HPC is being used throughout the bridge with low permeability limits to reduce the amount of

salt penetrating the concrete over the life of the structure and high strength to optimize structural members

and stretch span lengths.

Each of the 22 adjacent prestressed concrete box beams is 19.0 m (62 ft. 4 in) long and 0.686 m (27 in)

deep. The beams were fabricated by Concrete Building Systems (CBS) in Delmar, Delaware, and were all

produced in the same bed. The test program focuses on three beams, which are labeled B5, B7(4), and

B7(8) and can be seen in Figure 7. Each beam was prestressed using 12.7 mm (0.5 in) diameter, Grade 270

(1863 MPa), seven-wire low relaxation strands. Beams B5, B7(4), and B7(8) have 16, 18, and 18 straight

Wei Liu et al. 9

strands respectively. The strands were jacked to an initial tensile force of 138 kN (31 kips). After

approximately 18-hours after curing, the strands for each beam were released. Beam B5 was cast on August

30, 1999, while beams B7(4) and B7(8) were cast on October 4, 1999 and October 14, 1999, respectively.

Prior to casting, long-term monitoring instrumentation was installed in each beam.

TDR Instrumentation

A typical beam cross-section is shown in Figure 8. TDR monitoring wires have been installed alongside

a total of five strands in beams B5, B7 (4), and B7 (8). The wire is fully insulated, silver-coated copper

wire, which is commercially available. Both silver and copper are more corrosion resistive than steel.

Therefore, the corrosion of steel strands will occur well before the corrosion of the monitoring wire.

Corrosion monitoring wires were wound loosely around the strands. Since the TDR signal wavelengths are

much shorter than the twisting length, this did not affect the transmission line geometry. The monitoring

wires were held in place with cable ties, as shown in Figure 9.

A section of coaxial cable was used for each strand to provide electrical access to the strand/wire

system. The connecting coaxial cable is about 8.5 m in length, which is longer than necessary and allows

for installation variances. To minimize noise, the coaxial cable to strand/wire connection was made inside

the concrete beam and was covered by insulating tape. A standard BNC connector was installed on the

other end of the coaxial cable. An HP54750A digitizing oscilloscope was used to take TDR measurements,

which is capable of generating a fast rising step wave with rise time less then 100 picoseconds.

The energy loss can be a major concern when TDR is applied to long samples. Fully insulated

monitoring wire is used to reduce energy loss. At high frequencies, even the standard coaxial cable used to

connect the measuring system and the steel cable will consume some energy. Therefore, the coaxial cable

should be as short as possible to reduce energy loss and improve the sensitivity of TDR measure ments.

Upon the completion of the bridge construction, the connecting coaxial cable was cut to the shortest length

allowable.

Experimental Results

Corrosion monitoring began at the time of fabrication. During the fabrication of each beam TDR

measurements were taken before and immediately after the concrete pour. Since then, TDR readings were

Wei Liu et al. 10

taken on each of the 5 strands on roughly a monthly basis. All TDR results were converted and stored

digitally for future comparisons.

Figure 10 shows the TDR return from stand #2 of beam B7(4) before the pour. The sample is

terminated with a short circuit (electrical contact between strand and sensor wire). A voltage drop marks

the end of the sample. The electrical length is 142.7 nanoseconds. The signal is noisy due to the presence of

other electrically connected conductors. Note that the time scale used in this figure is different from the one

in Figures 11 and 12.

Figure 11 shows the TDR return from the same sample immediately after the pour. Theoretical analysis

shows that both characteristic impedance and propagation velocity depends on the dielectric constant of

concrete. When the water content of concrete is high, energy loss is significant because of the conductance

between steel strand and sensor wire. As another consequence, the characteristic impedance is small. From

Figure 11, it is hard to tell where the sample ends.

As a side benefit, it is possible to measure the moisture content of the concrete by TDR measurements.

As expected, signals indicated that water content changed most significantly in the first day after concrete

was poured.

Figure 12(a) and 12(b) are TDR readings taken 9 days and 177 days after the pour, respectively. The

end of the sample can be easily identified indicating that energy losses for the embedded transmission line

are not a problem. The electrical length of the sample is measured to be 266 nanoseconds, which is

significantly longer then the one without concrete.

Figure 12(c) is the differential comparison of Figure 12(a) and 12(b). It shows the total changes

occurred in the 168-day time period. TDR returns from this strand are repeatable with minor changes likely

due to the small changes in the concrete’s water content. Furthermore, the beams have been moved around

in the fabricator's yard during the monitoring process. This change in the environment also can cause

minor changes in the waveform. The repeatability of the TDR returns demonstrates the effectiveness of

differential comparison.

As discussed in the previous section, electrical connections were remade upon beam installation. As

expected, there were changes on TDR reflections due to the change of environment and connecting coaxial

cables. Initial measurements were taken after the bridge construction was finished. These results are used as

Wei Liu et al. 11

new baselines. If the steel strand is corroded in the future, a differential comparison of newly measured

signals to the baseline should reveal the corrosion damage site.

Since Bridge 8F is an HPC bridge, corrosion is not likely to occur within the next 20 to 50 years.

However, the purpose of this Bridge 8F experimental program is to transit TDR technology from laboratory

to the field. Even with standard concrete, a new bridge is not likely to corrode within the first 10 years,

which is much longer than the lifetime of a typical research project. However, it has been demonstrated that

TDR systems can be implemented and made operational. Furthermore, the stability of the signal over a

long time period will provide useful validation regarding the integrity of the system, and of the HPC.

CONCLUSIONS

A novel nondestructive evaluation technique for detecting damage in embedded or encased steel

reinforcement or bridge cables using time domain reflectometry has been developed and demonstrated.

Asymmetric transmission line models apply to steel elements with sensor wires. The effectiveness of TDR

in locating and characterizing simulated corrosion sites has been demonstrated through small-scale

laboratory tests, and TDR instrumentation has successfully been installed in beams of an HPC bridge

currently under construction. Differential TDR measurements are being used to monitor the onset of

corrosion of steel prestressing strands in the HPC bridge. The nature and repeatability of initial

measurements have demonstrated that the method is viable for actual field use. Additional research is

underway to investigate the application of this method for detecting damage in steel reinforcement of

cables of existing structures.

ACKNOWLEDGMENTS

This work was supported in part by the National Science Foundation under grant CMS-9700164 and the

Delaware Transportation Institute under grant No.929.

Wei Liu et al. 12

REFERENCES

1. “Time Domain Reflectometry Theory" Hewlett-Packard Application Note 1304-2, Hewlett-Packard

Company, Palo Alto, Ca lif. 1998.

2. Bhatia, S. K., R. G. Hunsperger, and M. J. Chajes. Modeling Electromagnetic Properties of Bridge

Cables for Non-destructive Evaluation. Proc. of Int. Conference on Corrosion and Rehabilitation of

Reinforced Concrete Structures, Orlando, Florida, 1998

3. Liu, W. Nondestructive Evaluation of Bridge Cables Using Time Domain Reflectometry. Master’s

thesis, University of Delaware, 1998.

4. Liu, W, R. G. Hunsperger, K. Folliard, M. J. Chajes, J. Barot, D. Jhaveri, and E. Kunz. Detection and

Characterization of Corrosion of Bridge Cables by Time Domain Reflectometry. Proc. of SPIE Vol. 3587.

Newport Beach, California, 1999.

Wei Liu et al. 13

FIGURE 1 Functional block diagram for a typical time domain reflectometer

Pulse Generator

Oscilloscope

Transmission LineZl

Load

FIGURE 2 Twin-conductor transmission line geometry of a bridge cable with sensor wire, where a is the

radius of the steel strand, b is the radius of the sensor wire, and d is the center-to-center distance between the

strand and wire.

a

b d

Z Z+ ∆ Z

R ∆ Z L ∆ Z

G ∆ Z C ∆Z

Z+ 2∆ Z

R ∆ Z L ∆ Z

G ∆ Z C ∆Z

FIGURE 3 Distributed parameter equivalent circuit of a transmission line.

FIGURE 4 TDR return of a 3-meter rebar sample. The sample has 50% pitting corrosion in the middle.

AB C

D

Wei Liu et al. 14

FIGURE 6 TDR results obtained from a 95cm seven-wire strand sample before (waveform 2) and after (waveform 1) a

simulated damage is made to the sample. The differential comparison in the bottom curve reveals the damage site.

FIGURE 5 TDR returns from 95cm seven-wire strand cable samples.

Six Severed Strands Two Severed Strands

Wei Liu et al. 15

FIGURE 7 Plan and cross-section of HPC bridge 8F.

(a) Instrumentation layout of beam B7(4).

(b) Beam B7(4) in the yard of the concrete plant.

FIGURE 8 Beam B7(4).

Wei Liu et al. 16

(a) Monitoring wire. (b) Connection to coaxial cable

FIGURE 9 TDR instrumentation.

FIGURE 10 TDR return of strand #2 of beam B7(4) on October 4, 1999. (before pour)

FIGURE 11 TDR return of strand #2 of beam B7(4) on October 4, 1999. (immediately after pour)

Wei Liu et al. 17

(a) TDR return on October 13, 1999.

(b) TDR return on March 29, 2000.

(c) Differential comparison.

FIGURE 12 TDR returns of strand #2 of beam B7(4).