An acoustic emission approach to bridge management is presented through case studies. The study shows promise that with technological advances, acoustic emission sensors may be used as part of integral sensor systems to monitor the health of bridge structures. A prototype system is currently in use, but still requires advances in automated data assessment schemes. G.P. Singh Contributing Editor

 

Introduction
The deterioration of steel bridge members is caused by a combination of load and environmental factors. The unpredictable rate of deterioration makes it difficult for the bridge engineer to plan for repair or replacement. This difficulty is increased by the limitation of public funds available to state departments of transportation. The bridge engineer, therefore, must consider the safety of the public as well as the bridge’s role in the transportation system.

Most of highway bridge inspection is done using visual inspection. When a crack in a steel member is observed, one of the following actions may be taken: the bridge is closed until the member can be repaired or retrofitted; the weight capacity of the bridge is lowered and it remains in service; and/or the frequency of inspection is increased to monitor visually the further deterioration of the problem area, and the bridge remains in service.

In the latter two situations, a degree of uncertainty exists because the rate of deterioration is unpredictable, so a decision must be made regarding the increased frequency of inspection. Because of access limitations to various parts of a bridge structure, an inspection visit may be expensive even if only a single structural element is inspected. In addition, the need to inspect other bridges within the highway system might prevent more frequent inspection. Acoustic emission testing (AE) exploits the fact that most materials release energy in the form of stress waves when microstructural damage, e.g., crack growth, occurs. Research has been done to apply AE to steel bridge structures.

---

Research to date has provided a reasonable scientific base upon which to build an engineering application of AE as part of a bridge management program.

---

In what was perhaps the first application of acoustic emission for testing bridges, Pollock and Smith (1971) collected data during proof testing of a portable tank bridge for the British Ministry of Defense. They demonstrated that signals recorded in the field could be associated with test results on laboratory specimens.

The next year, scientists from the Argonne National Laboratory (1972) monitored emissions from a bridge on Interstate 80 in Illinois, followed by Hopwood (1973), who monitored emissions from eyebar members of a bridge. He found good transmission through eyebar members, although noise was a serious problem.

An extensive program funded by the Federal Highway Administration (FHWA) with Battelle Pacific Northwest in the late 1970s resulted in the development of a battery powered digital acoustic emission monitor (Hutton and Skorpik, 1975 and 1978). This device allowed data to be collected periodically and stored on erasable programmable read only memory chips for additional processing and evaluation. The study demonstrated the utility of frequency spectrum analysis and the potential for centralized signal processing. Again, in addition to emissions associated with bridge component damage, a large number of noise related signals were detected.

During the period from August 1980 to July 1982, the Kentucky Transportation Research Center used the digital acoustic emission monitor to periodically monitor a bridge on Interstate 471 and pointed out effects of traffic and rainfall as sources of emission noise (Miller, 1987).

In the early 1980s, the Dunegan Corporation, under contract from the West Virginia Department of Highways, examined the practical difficulties in long term acoustic emission monitoring of bridges (Hartman, 1983). The financial benefits of this kind of monitoring over the use of periodic ultrasonic, magnetic particle, or liquid penetrant inspection of known defects were discussed.

United Technologies Research Center, under contract from FHWA, performed laboratory and field tests to characterize acoustic emission signals from flaws and various noise related sources (Miller et al., 1983). They explored different approaches using both time and frequency domain representations of signals. Pattern recognition and source classification for filtering out noise and for discriminating between different damage related events, such as brittle fracture and fatigue, were demonstrated. To facilitate the study, a field worthy acoustic emission sensor capable of detecting a broad band of frequencies was developed during the course of the program.

Prine and Hopwood (1985) considered an acoustic emission weld monitoring system for both fabrication and in service evaluations of bridge components. They pointed out that signals from bridges depend on traffic volume and vehicle speed and weight, as well as on structural details and transducer characteristics.

In 1987, the University of Maryland monitored the Woodrow Wilson bridge on the border of Maryland and Virginia for the Maryland Department of Transportation. They found that the predominant peak frequency of noise emissions is distinctly lower than crack related emissions. Suitable software filters, designed to exclude signals whose time domain parameters do not fall within the range of parameters of crack related emissions, can eliminate most noise signals (Vannoy et al.).

In a study completed in 1991, the same group conducted extensive laboratory tests on full size A588 bridge beams (Vannoy and Azmi, 1991). Acoustic emission parameters of cracks versus noise on rolled, welded, and cover plated beams were characterized in both time and frequency domains. It was also determined that corrosion has no effect on the time domain parameters of emissions from cracks.

In a related study, Hariri (1990), also of the University of Maryland, sought to develop a database of signal characteristics from different bridge steels and various material and loading conditions, as well as from different part geometries and thicknesses for application on bridge structure AE. He showed that noise filters, dictated by the type of material, thickness, paint layer, and corrosion conditions of a monitored part, can be developed using ranges of acoustic emission parameters provided by such a database. Surface paint layer was found not to significantly attenuate acoustic emission signals.

A series of field tests done for the FHWA by the Physical Acoustics Corporation on several bridges and various bridge details emphasized the need for source location and guard sensors for filtering out irrelevant acoustic emissions events (Carlyle, 1993; Carlyle and Ely, 1992; Carlyle and Leaird, 1992). Acoustic emission was demonstrated for testing the effectiveness of retrofits as well as in finding new cracks.

In Canada, the Canadian National Railways sponsored acoustic emission monitoring on 36 railroad bridges over a period of three years (Gong et al., 1992). Using a known functional relationship between the emission count rate and the stress intensity factor range, they were able to classify cracks into five levels of severity. Spatial discrimination and filtering using parameter windows determined from laboratory tests on bridge steels were used to eliminate noise.

Prine (1993) further demonstrated the effectiveness of combining AE and strain gage monitoring with tests on three bridges in Wisconsin and California. In a departure from the usual crack characterization function of acoustic emission monitoring, a bascule bridge was tested to determine the cause of loud impact noises that accompanied the lifting and lowering of the bridge.

Overall, the research to date has provided a reasonable scientific base upon which to build an engineering application of AE as part of a bridge management program. In addition, continued advances in electronics, such as faster microprocessors, provide testing capabilities that were not possible even a few years ago. To a nondestructive evaluation method that relies heavily on instrumentation, these advances give extra encouragement that better results will be obtained through further studies.

Nearly all of the work to date has sought to use AE to detect the initiation of damage, locate it, and then monitor its increase in severity. The approach taken in this work limits the application to that of monitoring. From an engineering point of view this restriction is quite significant. The limit means that the size and complexity of the AE system required may be greatly reduced. Noise sources associated with the structure may be eliminated, since the location of the test source (the problem area) is known. Requirements of monitoring, to support decision making of the bridge engineer, make it possible to configure a system that provides constant surveillance and early warning of changes in the condition of a critical bridge component.

Several field tests on actual bridge structures are described to demonstrate the engineering approach outlined above.

 

Instrumentation
AE on the bridges was performed using an eight channel Spartan AT acoustic emissions data acquisition system manufactured by Physical Acoustics Corporation. The entire system consists of the unit and a 386 personal computer. All functions of the system are controlled by the software program SA-LOC running in a DOS environment.

Of the eight channels available, six were used for performing signal measurement. These contain the circuitry that output the time domain parameters of an acoustic emission event such as counts, rise time, energy, peak amplitude, root-mean square energy, and duration. The two remaining channels were used to digitize and store waveforms using the PAC TRA-212 transient recorder analyzer system. The data acquisition system also has up to four parametric inputs which were used to record load data during laboratory tests and strain gage output during the monitoring of the various bridge elements.

 

Sensors and Auxiliary Equipment
Two types of piezoelectric transducers, the R30I and the WD, were used. The R30I, a resonant transducer with a peak resonant frequency of approximately 350 kHz, has an integral preamplifier that provides a gain of 40 dB. The WD, a wideband transducer with a relatively flat frequency response between 100 kHz and 1 MHz, is a differential transducer that requires a separate external preamplifier. A preamplifier with a highpass frequency filter of 20 kHz was used with the wideband transducer.

Magnetic holddowns were used to attach the resonant sensors to the monitored parts, and strips of duct tape or a cyanoacrylate adhesive were used to mount the WD sensor. Except in cases where the adhesive was used, a thin layer of vacuum grease couplant was applied to the interface between the transducer and part surface to aid in the signal transmission.

The strain gage used was a general purpose gage designed for strain averaging measurement on large specimens. It has a matrix length of 62 mm (2.46 in.) and a width of 8 mm (0.32 in.). The gage was attached using a methyl-2 cyanoacrylate adhesive.

Shielded RG50 coaxial cables 15 m (50 ft) in length connected the sensors (or the pre-amp of the wideband sensor) to the data acquisition unit. A portion of each cable close to the sensors was either looped around or taped to a secure part of the bridge to prevent the weight of the cable from pulling on the sensors and affecting the quality of the acoustic coupling between sensor and part surface.

A portable, gasoline fueled generator was used to power all instrumentation. Except for the need to periodically shut down the system for refueling, no problem was encountered with the power source.

 

Bridge Testing Setup Procedure
This section discusses procedures common to all bridge tests. It deals mainly with the steps taken to ensure that valid data are detected and stored during the tests.

With all sensors in place, the traditional pencil lead break test was performed for each sensor and source location sensor array. This test consists of breaking a 0.5 mm (0.02 in.) diameter pencil lead approximately 1.5 mm (0.06 in.) from its tip by pressing it against the surface of the piece. This generates an intense acoustic signal that the sensors detect as a strong burst. The purpose of this test is twofold. First, it ensures that the transducers are in good acoustic contact with the part being monitored. Generally, the lead breaks should register amplitudes of at least 80 dB for a reference voltage of 1 mV and a total system gain of 80 dB. Second, it checks the accuracy of the source location setup. This last purpose involves indirectly determining the actual value of the acoustic wavespeed for the object being monitored.

The software requires the user to enter a value of the acoustic wavespeed of the material being tested. In AE, this quantity could be anywhere between the velocity of longitudinal bulk waves and that of surface waves. The effectively measured wavespeed, however, may vary from test to test as influenced by the geometry and condition of the part being tested. An approximate value of this wavespeed can be determined using the differences in the time lead break signals arrive at two separate transducers.

Figure 1 — a) AE source location results for cracked pin showing no evident crack activity at suspected crack location, and b) source location results for new pin. (The numbers appearing in the squares represent the transducers.)

 

BRIDGE TEST
Comprehensive details of the AE results described briefly here may be found in a recent report (Clemeña et al., 1995).

 

Case 1 - New River Bridge
Acoustic emission monitoring was performed on the Route 460 westbound bridge over the New River in Glenlyn, Virginia, on September 24, 1993. This is a continuous span, multigirder bridge with sixteen pin and hanger and eight pin and hinge connections. Since 1990, suspected cracks in four pins have been monitored using ultrasonic inspection. It was decided to replace all the pins and hangers which were of A588 weathering steel with A276 stainless steel material. State Department of Transportation contractors were in the process of replacing the pins at the time the test was conducted.

Two pin and hanger connections on the western end of the westbound lanes were chosen for monitoring. One had a crack; the other was newly installed. The monitoring equipment was positioned under the bridge close to the pins which were about 4.5 m (15 ft) above the ground.

The pins are 300 mm long by 100 mm diameter (12 x 4 in.) at the widest section. A resonant sensor was attached to each end of both pins as close as possible to the axial center of the pin. Magnetic holddowns were used to attach the sensors to the cracked pin, and duct tape was used on the sensors mounted on the stainless steel pin. The data acquisition system was configured to perform linear source location, and an old pin that had been removed prior to the test was used to check the accuracy of the system settings. Since the pins being monitored were not exposed, the actual sensor setup was simulated on the old pin, and pencil lead break tests were performed along the length of this pin.

Live loading of the bridge was done exclusively by normal passing traffic, although the passing lane was closed due to repair work that was being done on the bridge. Data was collected for 1 hour and 52 minutes at the old pin, while the new pin was monitored for 18 minutes.

The newly installed stainless steel pin was monitored for a different purpose. It has been postulated that crack growth is by the pin seizing due to corrosion since this produces added torsion and bending loads not necessarily accounted for in the design of the pins. A freely rotating pin would surely cause rubbing on the mating pin and hanger surfaces that can be detected as rubbing noise by the acoustic emission sensors. AE can thus be used to determine qualitatively if a pin is seized or not.

The graph for the new pin (Figure 1) shows events that were detected over a period of 18 minutes. (Events from outside the region between the transducers appear to initiate at either one transducer or the other.) A total of 125 events occurred between the ends, compared to only 114 for the cracked pin which was monitored for nearly two hours. It is natural to expect that the old pin has more limited movement and this is shown by the results. However, the greater event rate for the new pin might also be attributed to mating surfaces that have yet to be smoothed by constant rubbing.

No crack related signals were detected during the monitoring period.

 

Case 2 - Staunton River Bridge
The Route 29 bridge over the Staunton River in Altavista, Virginia, was monitored on July 19, 1994. Two sites, which were accessed using a Bridgemaster snooper truck, were chosen for monitoring. The first location was on a 9 mm (0.375 in.) thick cracked web of an inner girder that had been retrofitted with a 9 mm (0.375 in.) thick splice plate bolted to the web. The crack continued to grow despite the retrofit and had progressed past the splice plate and under the bolted angular connector to the diaphragm. Resonant sensors 1 and 2, spaced 155 mm (6.125 in.) apart, were set up for linear source location to detect activity at the lower exposed end of the crack. A third sensor was attached to the flange to function as a guard sensor against fretting noise coming from the flange bolts.

The second location was on the same girder where another crack was found on the web. Holes had been drilled at both ends of the crack to arrest further growth. Three R30I sensors were positioned for triangular planar source location to detect the presence of crack activity past the lower stop drill hole which is right above the flange weld.

The data acquisition unit and PC were set up on the bridge, leaving only the inner lane open to traffic. Loading was accomplished via normal bridge traffic. The first location was monitored for a total period of 1 hour and the second one for about 45 minutes.

No events coming from the crack tip were detected by the sensor array at location 1. For the triangular array at location 2, no events were detected at all.

Results show that the crack at location 1 was not active during the time the monitoring was performed. They also show that crack growth at location 2 had been successfully arrested by the drill hole. However, since monitoring time was limited, the results may not be fully representative of the cracks’ general behavior, especially since the test was not done during a rush hour period and the bridge was loaded on only one lane.

The main difficulty encountered in this test was the limited space available for sensor placement around the cracks. The relatively large size of the resonant sensors and the magnetic holddowns limited the options for positioning the sensors to less than ideal. This affected both source location accuracy and the ability to set up effective guard sensors. There is a need to adapt the instrumentation, particularly the sensors, to each individual application.

 

Case 3 — Moormans River Bridge
The Route 671 bridge over the Moormans River in Albemarle County was monitored on July 27, 1994. A load limit of three tons had been posted for the bridge due in part to a crack found on one of the 20 mm (0.75 in) square diagonal counters. This was the only defect that was monitored.

The part monitored had a simple geometry and the crack was well separated from possible noise sources which could propagate only from the ends of the one of the counters. Two resonant sensors were attached at either side of the crack at a distance of 150 mm (6 in.) from each other. These sensors were setup to do linear source location. A wideband sensor, to be used to record waveforms, was mounted close to the crack on the opposite side of the bar.

The AE equipment was set up away from the bridge. Being on a secondary road, the bridge was loaded intermittently. The crack was monitored for a total period of about 1 hour and 30 minutes during a steady rain.

Although signals were detected every time a vehicle passed over the bridge, only three events were recorded by the source location program for the entire period; none of them came from the location of the crack. The triggering threshold of the digital analyzer had been set at absolute minimum, yet no signals triggered it. No waveforms were thus recorded. The crack is benign and has become inactive. The unbroken appearance of rust covering the crack tends to support this conclusion.

 

Case 4 — I-81 South Exit Bridge over Route 29
Monitoring of the I-66 south exit bridge over Route 29 in Gainesville, Virginia, was performed on August 16,1994. A truck passing under the bridge had accidentally hit the lower flange of the northernmost girder, causing the web and a stiffener to deform and the welds to crack. The girder was repaired by replacing the stiffener, heat straightening the web, and rewelding the flange weld and damaged coverplate.

Repairs had just been completed when AE was conducted. Two resonant sensors, labeled 1 and 2, were attached to the coverplate for source location on the new weld as shown in Figure 2. Sensor 3 was installed on the flange above the first array. This sensor serves as a guard sensor to sensors 1 and 2 while also doing linear source location with sensor 4. These two sensors were set up to monitor the rewelded web to flange section. The remaining sensors, 5 and 6, were positioned as shown in Figure 3 to act as guard sensors against noise coming from the floor beams. The wideband sensor was mounted close to the coverplate weld to record waveforms.

 

Figure 2 — Sensor placement at coverplate showing sensors 1 and 2 positioned for linear source location to monitor new welds.

 

Figure 3 — Drawing of repaired girder showing location of sensors 3 and 4, positioned for acoustic emission source location to monitor lower flange-to-web weld, and guard sensors 5 and 6.

 

The bridge riding surface was under repair at the time of the test. Only one lane was open to traffic which had to be slowed down as the bridge was crossed. Data was collected for a total of 1 hour and 25 minutes.

Although results from this particular monitoring test suggested no crack activity, this could not be a reliable gage of whether repair to the damaged girder was successful or not. The reason is that the bridge was not subjected to normal loading. It was most likely experiencing less dynamic loads during the test than in normal service.

 

Case 5 — Robinson River Bridge
The most extensive monitoring performed as part of this project was conducted on the Route 29 northbound bridge over the Robinson River in Madison County, Virginia. The bridge was tested on three separate occasions as work on the project and on an ensuing related study progressed. It was first tested on June 24, 1994, then on October 25, and again on December 8 and 9.

The bridge, built in 1934, has four steel girders extending over five spans with an overall length of 59 m (193 ft). The suspended span is supported by eight pin and hanger connectors on the north end and by corresponding pinned joints on the opposite ends. The 600 mm high by 165 mm wide (24 ´ 6.5 in.) hangers, with a 230 by 64 mm (9 ´ 2.5 in.) slot cut out of the middle, were fabricated from 15 mm (0.625 in.) thick steel plate.

During a regularly scheduled inspection in October 1992, cracks were found on two of the hangers. The east exterior hanger had one crack later measured at 0.9 mm (0.375 in.) long. Three similarly located cracks were found on the west interior hanger. Measurements done during the December AE showed the longest to be 36 mm (1.437 in.) long while the shorter upper crack was 6 mm (0.25 in). Two cracks are shown schematically in Figure 4. One of the upper cracks, which is not a through part crack, does not appear in the figure and is visible only at the surface of the internal slot. Bolted catch plates have been installed on both hangers to prevent collapse in the event of sudden failure of the hangers.

June 24 test. Both cracked hangers were monitored during the first AE. A resonant sensor was attached as closely as possible to each crack. Sensor 2 was positioned close to the crack on the east hanger while sensors 1 and 3 were mounted on the ends of the pin as shown in Figure 5. Sensors 4 and 5 monitored the two upper and the lower crack, respectively on the west hanger as shown in Figure 4. The wideband sensor was installed on the top pin of the west hanger.

A Bridgemaster snooper truck was used to access the hangers during setting up of the sensors. The truck was then taken off the bridge and both lanes cleared at the time of actual data collection. The bridge was loaded by normal noontime traffic and data were recorded for a total of 40 minutes.

In addition to signals recorded by the software program, waveforms were monitored and stored with a transient digitizer. One channel of the 2 channel system was assigned to the wideband sensor while the other channel was used for the resonant sensors.

Although source location was not intended during the actual test, analysis of the collected data made it apparent that in order to distinguish relevant signals from noise, spatial discrimination using source location was necessary. Due to the irregularity of the sensor placements, the length of the effective wave propagation paths between the sensors could not be ascertained. This information is necessary for source location calculations. Still, using the differences in the time of arrival of an emission at two or more sensors as well as the sequence of arrival at the sensors, it was possible to get the approximate location of the source of recorded events.

 

Figure 4 — Sensor placement at west inner hanger during the June 24 test. Location of cracks is also shown.

Figure 5 — East outer hanger showing dimensions and placement of sensors during the June 24 test.

A significant portion of all the detected events occurred close to sensor 4 at the upper crack, west hanger. Most signals were detected by only 1 or 2 sensors, but the high amplitude events regularly hit three sensors (4, 5, and 7). That the highest amplitude signals originated there was also evident from the recorder analyzer data. Waveforms were collected from the wideband sensor and sensors 1, 4, 5 and 6. The analyzer was set at the lowest threshold possible to maximize sensitivity. At this level, only the wideband sensor and resonant sensor 4 detected signals that were recorded by the analyzer. Apparently, acoustic emission activity during the period of monitoring was greatest near the upper crack, although it could not be ascertained then if the signals were coming from the crack itself or from the nearby pin. These results were the basis for the decision to concentrate the monitoring effort on the upper crack of the west hanger on the succeeding test.

October 25 and December 8 and 9 tests. Only the cracked west inner hanger was monitored on the second test of the Robinson River bridge. Resonant sensors 3 and 4 were attached 175 mm (7 in.) apart on both sides of the upper crack (crack 1) as shown in Figure 6. The wideband sensor, which was to be used for recording waveforms, was mounted between the resonant sensors close to the crack using cyanoacrylate adhesive. The other three resonant sensors were used as guard sensors to detect noise coming from outside the crack zone. Guard sensor 1 was positioned to eliminate rubbing noise from the top pin while sensor 6 was used to eliminate noise from the lower pin and signals that may come from the lower crack (crack 2). Sensor 5 was mounted on the girder connector plate to filter out noise from the girder itself.

 

Figure 6 — Sensor placement at the west inner hanger used to monitor crack 1 during the October 25 test. Strain gage location is also shown.

 

Unlike the first test where AE parameters and emitted waveforms were recorded at separate times, the software was operated with the analyzer program running in the background so that both systems recorded the same events. To increase the sensitivity of the analyzer waveform recorder, an additional preamplifier was connected between the signal cable from the sensor and the analyzer input to further amplify signal levels. The pre-amps were set at 40 dB gain. One analyzer channel was used for the wideband sensor while sensor 3 was connected to the other analyzer channel. A digitization rate of 5 MHz was chosen, and the threshold levels were set so that the analyzer system only triggered when a vehicle passed over the bridge. Waveform size was set at 8K points which, at a pretrigger delay of 10 percent, allowed the recording of 1.47 m/s long segments. For the software, the threshold level had to be set at 30 dB to avoid low intensity background noise.

A further improvement to this particular monitoring situation over the previous tests was the use of a strain gage. This strain gage was attached to the left side of the link as shown in Figure 7. The conditioned strain gage output was connected to the data acquisition unit and recorded as parametric input 1. Thus, in addition to spatial discrimination, the setup provided for strain discrimination as an additional tool for use in distinguishing between fretting noise, crack face rubbing, and crack extension emissions.

 

Figure 7 — Location of sensors used to monitor crack 1 during the December 8 and 9 tests showing sensors 3 and 4, positioned for linear source location, and guard sensors 1, 2 and 6. Sensors 2 and 6 double as source locators for crack 2. Also shown is the wideband sensor close to crack 1 and the strain gage location.

 

Dirt and corrosion had accumulated and hardened between the hanger and the girder web extension right beneath the crack. Since these products can produce noise similar to crack signals, the space between the link and the girder was carefully cleaned. To further decrease the possibility of fretting noise coming from the vicinity of the crack, the area was sprayed with WD-40 lubricant.

As in the first test on this bridge, a Bridgemaster was used in setting up the sensors and strain gage. Both lanes were again cleared and the bridge was loaded by normal bridge traffic during data acquisition. Total monitoring time was 1 hour and 35 minutes.

The objective of the third test on the Robinson River Bridge was to gather more signal and waveform data from crack 1 and, in addition, to monitor the longer crack 2. Accordingly, the sensors were set up in two ways and monitoring time was split between the two setups. In the first test, which was done on the first day and part of the second, the sensor placement was similar to the October 25 test except that another resonant sensor, sensor 2, was attached close to the lower pin so that source location could be performed on crack 2. This sensor also doubled as a guard sensor for the source location sensor array at location 1. Sensor placement is shown in Figure 7. Sensor 5 had to be disabled since the maximum number of operating channels had been reached. Results of the October 25 test showed that guard sensor 5 was unnecessary.

 

Figure 8 — Location of sensors used to monitor crack 2 during the December 9 test showing sensors 2 and 6, positioned for linear source location, and guard sensors 1 and 3. Also shown is the wideband sensor close to crack 2 and the strain gage location.

 

Though unintentional, the distance between sensors 3 and 4 was increased to 185 mm (7.25 in.).

In the second sensor arrangement scheme, shown in Figure 8, the wideband sensor was moved close to crack 2 so that signal characteristics and waveforms from this crack could be recorded. Guard sensor 1 was attached to the end of the lower pin, while guard sensor 3 was used to eliminate noise coming from the upper pin. Sensors 2 and 6 remained in the same position for the purpose of linear source location on crack 2. The analyzer channels were connected to both the wideband sensor and to resonant sensor 6.

The strain gage was kept at the same location as on the October 25 test. The threshold setting on the software was decreased to 25 dB from the 30 dB level used in the previous test. It was apparent from the results of that test that some crack related signals had peak amplitudes of less than 30 dB. Analyzer threshold levels were likewise adjusted so that weak, long duration noise signals would not be detected. The digitization rate was kept at 5 MHz while waveform record lengths were decreased to 1.17 ms for the wideband sensor and 0.778 ms for the R30I sensor on the basis of results from the preceding test.

The hanger was monitored for a total of 44 minutes on the first day and 1 hour and 35 minutes on the second day using the first sensor arrangement. The bridge was open for full traffic. Sensors were then rearranged to the second setup scheme and data collected for 22 minutes. Monitoring times were not continuous and on occasion, data was collected only when large vehicles such as tractor-trailer trucks, passed over the bridge.

Spatial discrimination and waveform classification, in conjunction with load discrimination, were used to assess the extensive data collected during this testing. The details of this analysis are too extensive to include here. Overall it is clear that emissions from cracking were indeed detected, primarily during loading by large trucks. Analysis of this data suggests that it is possible to effectively monitor a problematic steel bridge member using a simple acoustic emission system operating with a reduced data analysis procedure.

 

Summary and Conclusions
A summary of the acoustic emission field tests is shown in Table 1, and the potential applications for monitoring are shown in Table 2.

In order to fully assess deterioration of steel bridge members by means of AE, extensive instrumentation is necessary. A large number of detection channels is required and monitoring and analysis must be carried out over long periods of time. Information obtained from the AE must be monitored. Since the nature of actual damage is not known initially, the placement of sensors would most likely not be ideal. The expense of this effort would not be offset by any savings, in that inspection requirements for bridges would not allow for substitution of testing for periodic visual inspection. Therefore such an approach would result in added costs.

The approach proposed here, however, supplements the present inspection procedure, provides increased assurance of safety, and possible early warning of changes occurring in critical members already identified visually as being problematic.

 

Table 1 Summary of acoustic emission bridge test
Bridge	Detail	Problem	Results
Route 460 New River, Glenlyn, VA	Pin & Hanger	Ultrasonic detected crack in pin	No crack activity
Route 29 Staunton River, Altavista, VA	Grider Web	a) Web crack: retrofitted with splice plate b) Web crack: arrested using stop drill holes	No crack activity
Route 671 Moorsmans River Albemarle Co., VA	Diagonal counter	Visible transverse crack	No crack activity
I-66 south exit over Route 29, Gainsville, VA	a) Coverplate weld b) Lower web-to-flange weld	New repair welds	No crack activity
Route 29 Robinson River	Pin and hanger	4 visible cracks in 2 hangers	Crack activity from 3 cracks

---

Table 2 Potential problem suitable for acoustic emission monitoring
Detail	Problem
Pin and hanger	Crack in pin Crack in hanger Corroded pin
Grider web	a) Web crack: retrofitted with splice plate b) Web crack: arrested using stop drill holes
Girder	New repair welds

 

Ideally, once a critical problem has been identified visually during a regular inspection of the bridge, a compact system could be installed on the bridge and used to monitor that specific problem area. Technologically, it is possible for such a system to be solar battery powered and to incorporate a cellular telephone for reporting to the bridge engineer’s office. This system would remain on the bridge until repair or replacement occurred. The system could then be used on another bridge, whereas the alternative approach outlined above would require a system for each bridge throughout its life.

Effective use of acoustic emission monitoring can reduce the frequency of extra bridge inspection, while actually increasing the level of safety, since the monitoring would be continuous, with reporting occurring possibly on a daily basis.

At present a prototype system, incorporating the capabilities described above is installed on a bridge, and is being used to monitor a critical member. Automatic data assessment schemes are being evaluated.

It is clear, based on the field experience of the authors, that even short term monitoring of actual structures will provide large amounts of data that often include few "significant" signals. This is due to the unpredictable nature of the deterioration load environment synergism. From an engineering perspective, methods must be developed to automatically assess this data in order for it to be used for practical applications. For example, the bridge engineer cannot afford to assess hundreds of hours of data to find critical signals.

One would expect, however, that with new bridge construction, recognizing the magnitude of the investment and advances in electronic instrumentation and sensors, that integral sensor systems would be installed to monitor the health of the bridge structure. Acoustic emission is likely to be one of the phenomena exploited. However, such health monitoring can only be beneficial if automated data analysis reduces the data to a critically important subset.

 

References
Argonne National Laboratory, "Formal Proposal for Acoustic Emission of Steel Highway Bridges," Submitted to the National Science Foundation, Apr. 1973.

Carlyle, J.H., "Acoustic Emission Monitoring of the I-10 Mississippi River Bridge," Phase Report No. R90-259, Physical Acoustics Corporation, Lawrenceville, NJ, 1993.

Carlyle, J.H., and J.D. Leaird, "Acoustic Emission Monitoring of the I-80 Bryte Bend Bridge," Phase Report No. R90-259, Physical Acoustics Corporation, Lawrenceville, NJ, 1992.

Carlyle, J.H., and T.M. Ely, "Acoustic Emission Monitoring of the I-95 Woodrow Wilson Bridge," Phase Report No. R90-259, Physical Acoustics Corporation, Lawrenceville, NJ, 1992.

Clemeña, G., M. Lozev, M. Sison, and J. C. Duke, Jr., "Acoustic Emission from Steel Bridge Members," Interim Report (FHWA/VTRC95-IR1), Virginia Transportation Research Council, May 1995.

Gong, Z., E.O. Nyborg, and G. Oommen, "Acoustic Emission Monitoring of Steel Railroad Bridges," Materials Evaluation, Vol. 50, No. 7, Jul. 1992, pp 883–887.

Hariri, R., "Acoustic Emission Investigation and Signal Discrimination in Steel Highway Bridge Applications," Ph.D. Dissertation, University of Maryland, College Park, MD, 1990.

Hartman, W.F., "Acoustic Emission Monitoring of Electroslag and Butt Welds on Dunbar Bridge, West Virginia," Report No. FHWA-WV-83-001, Dunegan Corporation, San Juan Capistrano, CA, 1983.

Hopwood, T., "Acoustic Emission, Fatigue, and Crack Propagation," Report No. 457, Kentucky Bureau of Highways, Oct. 1976.

Hutton, P.H., and J.R. Skorpik, "Acoustic Emission Method for Flaw Detection in Steel for Highway Bridges," Report No. FHWA-RD-78-97, Battelle Pacific Northwest, Richland, WA, 1975.

Hutton, P.H., and J.R. Skorpik, "Acoustic Emission Method for Flaw Detection in Steel for Highway Bridges," Report No. FHWA-RD-78-98, Battelle Pacific Northwest, Richland, WA, 1978.

Miller, R.K., ed., Nondestructive Testing Handbook, Vol. 5, 1987, p 330. ASNT, Columbus, OH.

Miller, R.K., H.I. Ringermacher, R.S. Williams, and P.E. Zwicke, "Characterization of Acoustic Emission Signals," Report No. R83-996043-2, United Technologies Research Center, East Hartford, CT, 1983.

Pollock, A.A., and B. Smith, "Stress-Wave Emission Monitoring of a Military Bridge," Nondestructive Testing, Vol. 30, No. 12, Dec. 1972, pp 348–353.

Prine, D.W., "Application of Acoustic Emission and Strain Gage Monitoring to Steel Highway Bridges," Proceedings of the ASNT 1994 Spring Conference, New Orleans, Louisiana, Mar. 1994, pp 90–92.

Prine, D.W., and T. Hopwood, "Improved Structural Monitoring with Acoustic Emission Pattern Recognition," Proceedings of the 14th Symposium on Nondestructive Evaluation, Southwest Research Institute, San Antonio, TX, 1985.

Vannoy, D.W., and M. Azmi, "Acoustic Emission Detection and Monitoring of Highway Bridge Components," Report No. FHWA-MD-89-10, University of Maryland, College Park, MD, 1991.

Vannoy, D.W., M. Azmi, and J. Liu, "Acoustic Emission Monitoring of the Woodrow Wilson Bridge," Report No. FHWA-MD-87-06, University of Maryland, College Park, MD, 1987.

* Engineering Science and Mechanics, Virginia Tech, Blacksburg, VA 24061-0219;

(540) 231-6063; fax (540) 231-9187

⁺ Virginia Transportation Research Council, Charlottesville, VA 22903

Copyright © 1996 by the American Society for Nondestructive Testing, Inc. All rights reserved.  

[ Materials Evaluation ]