This is a preliminary summary report describing fatigue testing conducted by the Transportation Technology Center, Inc. (TTCI) of the Association of American Railroads on cracked girder specimens from a span of the Mason Creek Bridge from Grande Cache, Alberta. Two spans of this bridge, which is owned by Canadian National Railways (CN) had recently been replaced because of fatigue crack problems in the welded assemblies. In-situ acoustic emission testing to ascertain the condition of fatigue cracks on this bridge had previously been conducted by DNL Infrastructure Technologies Inc. (DNL). The purpose of the fatigue tests at TTCI was to attempt to correlate field test results with laboratory results using acoustic emission testing and accelerated laboratory fatigue tests.
This report by DNL presents an abridged preliminary summary of results obtained during the TTCI tests, based on only cursory examination of the test data. A final report which includes detailed analysis of the TTCI test data as well as a comparative examination of field test data collected on the full spans at Grande Cache, will be provided at a later date.
The objective of the TTCI tests was to verify the correlation's among fatigue crack growth and the acoustic emission activity generated by a crack during controlled cycling on a fatigue-testing machine. The accelerated laboratory loading would attempt to simulate field loading by using 610 x 1220 mm (24 x 48 in) pre-cracked girder samples in four-point bending on a large MTS fatigue testing machine.
Design and fabrication of the test specimens and the test fixtures, calculation of a loading regime to simulate accelerated field service conditions and conducting the cyclic load tests were all performed by TTIC staff. Measurement and recording of the acoustic emissions generated by the crack as well as measurement and recording of the strain field in the crack vicinity were conducted by DNL. Acoustic emission (AE) testing was conducted from August 26 to 29, 1998. TTCI had previously cycled several test coupons to failure to determine suitable MTS loading conditions for the AE tests. As well they had conducted typical metallurgical tests to characterize the steel in the test specimens.
TTCI provided two optical telescopes that were focussed on existing fatigue cracks in the test specimen. These telescopes enabled tracking the progress and size of fatigue cracks which were within the field of view. Since the vertical stiffener blocked a portion of the test area, crack growth behind the stiffener could not be viewed. TTCI also mounted two strain gauge rosettes on the specimen to enable the measurement and calculation of principal strains in the vicinity of the fatigue cracks.
The following monitoring and data collection equipment was provided by DNL and installed on the test specimen to collect load, strain and acoustic emission data:
(The first system was provided by Mr. H.L. Dunegan of Dunegan Engineering Consultants Inc. (DECI), the inventor of the AE Smart 2000; the second system was provided by DNL.) These instruments were each set up with two transducers, a split frequency SE9125 transducer with integral 20 dB pre-amplifier and an SE-150-MI 105 kHz resonant trigger transducer. The two systems were identical except that the system provided by DECI contained an enhanced timing circuit.
The timing of the above systems was synchronized before testing to enable correlation among AE, strain and load data.
Before test initiation, TTCI had performed calculations and load tests on coupon specimens to determine appropriate settings for the MTS machine that would result in fatigue crack re-initiation in the test specimen. These tests determined that with the specimen pre-loaded to 18 kN (4 kip) and cycled at a frequency of 3 Hz between +18 and +980 kN (+4 and +220 kip), crack re-initiation would occur. Anticipated crack growth would be about 10 mm (0.4 in) in 8 hours.
Before AE testing, test specimens were visually inspected by TTCI, using both dye penetrant and magnetic particle techniques. Inspection results were documented to provide a description the initial state of each test specimen.
TTCI had designed the test fixture for four-point bending of the specimen. The lower flange of the specimen was supported on the MTS bedplate at two points, 533 mm (21 in) on each side of the specimen centerline. The upper web of the specimen was loaded through two points on the MTS head block, 178 mm (7 in) on each side of the specimen centerline. Since the MTS head block, which is powered by two parallel actuators, was not restrained from movement in the lateral or longitudinal specimen directions, four-point bending could be achieved only in ideal conditions. As is discussed later, the unrestrained actuator action resulted in a loading regime that could best be described as combined four-point bending with lateral web flexure. In all likelihood, this complex loading may more closely duplicate field loading on deep girders than does four-point bending. Figure 1 shows the test set up.
Because of the reported potential problems with background hydraulic noise on large fatigue test machines, especially those which have not been modified with special rubber hydraulic hoses, DNL had requested that damping assemblies composed of alternating steel and aluminum plates be installed at each of the four specimen load points. TTCI had fabricated and installed damping assemblies that had the potential of reducing the MTS hydraulic noise transmitted to the specimen by up to 90%. Unfortunately, because of the design of the upper and lower specimen grips, the damping assemblies could be effective only in four-point bending. Grip contact with the specimen would occur with lateral web flexure resulting from unrestrained head block movement, essentially "shorting-out" the dampers. TTCI wedged teflon sheets between the grip arms and test specimen in an attempt to reduce noise transmission, however such approaches have previously been shown to be relatively ineffective.
Figure 1. Photograph of the Instrumented Test Specimen in the MTS Fatigue Machine.
TTCI adjusted, calibrated and operated the MTS fatigue machine throughout the test. DNL used T-connectors to sample values from the six strain gauges on the test specimen and from the two load cells on the MTS actuators.
The PAC AE system was set for linear source location with two transducers mounted just above the bottom flange angle of the specimen. The transducers were spaced at 600 mm (24 in). The speed of sound transmission in the specimens was experimentally determined to be about 3000 m/s (19.7x103 ft/s). The transducers were spaced so that the welded stiffener and the original fatigue crack were located at about 150 mm (6 in) from the nearest transducer. This configuration was used to reduce noise in the datasets. Pencil break tests were performed according to ASTM standards to complete system calibration. Signals from two strain channels were used as parametric inputs to the AE channels to enable correlating load with AE generation.
The two AE Smart systems were calibrated in a slightly different manner. The first calibration was performed to demonstrate that the AE system was correctly responding to In-Plane (IP) and Out-Of-Plane (OP) waves. Since fatigue crack growth activity is primarily carried by IP waves it is necessary to calibrate the AE Smart to clearly distinguish IP from OP wave sources. IP waves were created by pencil break tests on the edge of the test specimen, while OP waves were created by pencil breaks on the web surface near the transducers. A second calibration was then performed to force the system to respond only to AE events generated near the original fatigue crack. This was accomplished by using pencil breaks and observing the arrival times of AE events between the trigger transducer and the data transducer. Trigger transducers for both the AE Smart systems were mounted on the test specimen on the backside of the web, directly behind the original cracks in the stiffener weld. The data transducers were mounted on the web about 300 mm (12 in) from the original crack tip. Ratio and delta T filters were set in one of the AESMART 2000 systems such that only IP signals coming from the vicinity of the existing crack were recorded.
Testing began on specimen #1 at 09:15 on August 27. TTCI had decided to initiate the test by applying and maintaining an 800 kN (180 kip) load. This would enable characterizing the hydraulic noise from the MTS machine before start of load cycling. As a result of the unrestrained attachment of the head block, the load induced out of plane bending in the web and within 15s of load application the web buckled forward about 45o. The web deformation forced the original crack to close and caused the web stiffener to contact the lower flange of the specimen. The test was halted for obvious reasons to verify the instrumentation setup and to attempt to produce a solution to the less-than-ideal specimen setup. TTCI decided to continue testing this sample even though it appeared to be damaged beyond effective use. All transducers were removed and a cutting torch was used to remove the portion of the stiffener contacting the bottom flange to reduce mechanical noise. AE transducers were then remounted, the system re-calibrated and the loading re-initiated.
Key observations after test re-initiation were as follows. With an initial cyclic load ranging from +18 to +133 kN (+4 to +30 kip) at a frequency of 1 to 3 Hz, the upper web of the specimen wobbled laterally in the test jig. The desired four-point bending could not occur because of no head block restraint. The AE systems were closely monitored for signs of noise since this was the only anticipated type of AE event at these low load levels. AE data were collected by all systems during loading but has not been analyzed. Visible inspection did not confirm any crack growth. Further cycling with a load ranging from +90 to + 267 kN (+20 to +60 kip) at 3 Hz indicated the presence of AE from crack activity but analysis is necessary to determine the source.
At 13:15 on August 27 it was concluded that this damaged specimen was unsuitable for test purposes and the testing of specimen #1 was stopped.
Specimen #2 had 25mm (1 in) existing cracks on each side of the welded stiffener at the boundary of the weld and the base metal.
Testing began at 16:40 after swapping specimens, re-mounting AE transducers and strain gauges and re-calibrating the AE systems. The applied load was cycled between +89 and +222 kN (+20 and +50 kip) at 1 Hz to initialize the test. Once again, the upper web of the specimen wobbled laterally and the loading regime could be defined as combined web-buckling with four-point-bending. This time, however, the web deflection was away from the surface of interest, tending to open the cracks and alignment was improved to reduce the relative magnitude of the lateral web deflection. At 18:15 the load was set to cycle between +89 and +356 kN (+20 to +80 kip) at 3 Hz. This began to provide a large dataset in the PAC AE system. Initial data analysis shows that the recorded AE events originated from the welded stiffener near the original fatigue cracks. Further analysis is needed to validate this AE in terms of exact location and statistical features.
The test was stopped at 19:00. Visual inspection did not reveal any additional growth in the original crack nor any additional cracks. Concerns were expressed that re-initiation of crack growth may not be possible with the existing test setup. One suggestion was to shim the stiffener to increase the stress in the vicinity of the original crack. TTCI decided to maintain the existing configuration and to increase the load in small, carefully monitored increments until crack growth was evident.
Testing of specimen #2 resumed on the morning of August 28, with continuous monitoring by all AE systems. Of the two AE Smart systems, only the Dunegan system was operating to full potential because of its enhanced filtering capability. The DNL system was set to filter based on IP vs. OP criteria but the triggering feature was inoperative due to a fabrication oversight.
At 11:00 the MTS machine was programmed to cycle from +89 to +534 kN (+20 to +120 kip) at 3 Hz. The original crack could be seen to open and close, causing the PAC AE system to record many signals from the crack vicinity. The AE Smart system did not respond to these signals so it was assumed that they were OP noise. Visual inspection confirmed that the original cracks had not grown and that there were no new visible cracks.
At 13:49, magnetic particle testing was performed to verify the size of the existing fatigue cracks. There were about 50,000 load cycles at this time and the cyclic loading was set to vary from +89 to +600 kN (+20 to +135 kip) at 3Hz. This inspection confirmed that although the original fatigue crack under observation was actively opening and closing, it had not grown and there were no other confirmed cracks in the vicinity.
At 15:05, the Dunegan AE Smart system began to collect AE events (Figure 2). These events occurred in clusters and continued with a combination of high and low counts. The MTS machine was allowed to continue cycling for a brief period between + 89 and +650 kN (+20 and +145 kip) at 3Hz. As the AE events began to occur less frequently with continued cycling, it was decided stop the MTS machine and inspect the specimen again. A magnetic particle inspection confirmed that the original cracks were not any larger but there were two new cracks in the web material. It appeared that there was a single point of origin slightly above the original crack that forked into two cracks. A preliminary study of the collected AE data indicates the AE was likely from a surface crack. Although microscopic analysis of the two cracks would have been required to confirm that they were surface cracks, they visually appeared as surface cracks. A detailed study of the AE data and strain data collected by all the systems is required to classify the AE and confirm its source.
It is important to note that even after discovery of the two new cracks, they were still impossible to see without the use of dye penetrant or magnetic particle techniques. The AE Smart however correctly detected and reported this fatigue crack initiation.
The test was continued at 16:00 but the MTS machine soon shut itself down automatically due to a malfunction in the setup program. After test re-initiation load cycling continued at previous levels until 17:10, at which time the test was discontinued due to noticeable slippage between the test specimen and the jig. The specimen was repositioned to enable test continuation in the morning.
Figure 2. Plot of Fatigue Crack Initiation and Growth as Detected in Specimen #2.
On the morning of August 29, DNL arrived in the test lab at the scheduled 08:00 start time. TTCI had however initiated cyclic loading at about 07:00 at the previous days levels of +89 to +650 kN (+20 to +145 kip) at 3 Hz. Since no data collection equipment was turned on, there are no AE data to analyze during this critical period. Since the primary goal of this test was to correlate crack growth with AE generation, the missing one-hour of data could conceivably have rendered the remaining test data inconclusive. The data from after this period have not been analyzed in detail but it is very likely that since two new cracks had initiated immediately before the 07:00 restart, the one-hour of unrecorded cyclic loading may have caused numerous other defect changes.
Examination at this time, of the load spectra applied by the MTS machine during start/stop phases revealed a potentially troubling aspect for data analysis. A load surge or spike occurred whenever the MTS machine was started and the initial load was applied. At small applied loads this was unimportant but at the higher loads this surge could be damaging. The restarting of the test at 09:00 illustrates this point. The load monitoring traces revealed a large spike and it was subsequently evident it had caused the original crack to grow significantly, by plastic deformation (tearing). Further analyses of load and AE data are needed to confirm if any AE was detected during this plastic deformation.
Shortly after test initiation more cracks appeared under the welded stiffener. With at least eight active crack tips in the vicinity of the AE transducers, it becomes nearly impossible to assign crack growth activity to any particular crack.
The final hours of the test were assigned to collecting data to distinguish machine noise and crack fretting signals from crack growth signals. It was decided to reduce the cyclic load to range from +89 to +178 kN (+20 to +40 kip) at 5 Hz and to collect sufficient data to characterize signals at this level. There should not be any AE from crack growth in such a data set. After collecting data for one hour, an additional visual inspection was performed to confirm that there was no additional crack growth.
The load was then increased substantially and data were collected to characterize global crack growth activity. Unfortunately a grip on the MTS machine broke and the test had to be discontinued at 11:38.
Fatigue testing of cracked girder specimens from a replaced span of the CN Mason Creek Bridge was conducted by TTCI from August 26 to 29, 1998. The 610 x 1220 mm (24 x 48 in) pre-cracked girder specimens were loaded in combined four-point bending with lateral web buckling, in an attempt to simulate field conditions. Throughout the cyclic loading, the strain field in the vicinity of the existing fatigue cracks was continuously monitored. As well, acoustic emission monitoring was conducted using three different types of AE systems. TTCI staff were assisted by staff from DNL and DECI, who provided the monitoring instrumentation and recorded the test data.
The CN Mason Creek Bridge, despite its low traffic volume and relatively young age, has continued to experience fatigue crack problems in a number of welded assemblies. CN has utilized the acoustic emission testing services of DNL at several times to assess the activity of fatigue cracks on this bridge and to provide additional background to assist in assessing the need for maintenance or repair. Acoustic emission testing is a passive, non-destructive testing technology that monitors the relative activity of fatigue cracks during normal cyclic loading, such as the passing of a train. Either when used separately or combined with other testing techniques, acoustic emission testing has the advantage of being able to detect whether fatigue cracks in fracture critical members are actively growing or have arrested. It also can measure the relative growth activity of existing cracks. When combined with other data, this can assist in establishing cost-effective maintenance, repair and replacement schedules. The objectives of the TTCI fatigue tests were twofold - to verify correlations among fatigue crack growth and the acoustic emission activity generated by a crack during controlled laboratory testing on large specimens and to attempt to correlate field test results on the same specimen with laboratory results.
Based on cursory analyses of the test data and an initial assessment of test results, the following preliminary conclusions may be made: