Development of a Fatigue Sensor for Welded Steel Structures
Yanhui Zhang and Peter Tubby
Paper presented at International Conference Residual fatigue life time extension of in service structures, JIP 2006, Paris, France, 30 May - 1 June, 2006.
A recent collaborative project managed by TWI has developed a fatigue damage sensor suitable for welded joints in steel. Known as the CrackFirst TM sensor, it consists of a thin steel coupon with a pre-crack at its centre, attached to the structure adjacent to a critical weld detail. Cyclic loading applied to the structure induces a cyclic stress in the coupon, resulting in propagation of the pre-crack. This paper describes the sensor and its application, and presents results obtained from fatigue tests performed as part of its validation. The validation tests conducted under both constant and variable stress amplitude have demonstrated that the fatigue sensor system is suitable for welded joints in steel and is capable of providing advance warning of the rate at which the design life is being consumed. The sensor provides a reliable indication of design fatigue life consumed under both pulsating tension and alternating loading and it has a fatigue performance similar to Class F in the current fatigue design rules.
Although many types of fatigue damage monitoring systems have been developed and used in a variety of service conditions, the nature of the fatigue process in welded joints imposes special requirements and limitations on the fatigue sensors which can be applied successfully. In particular, the presence of fine crack-like flaws at the weld toe largely eliminates the crack initiation phase, which is a significant proportion of the fatigue life of unwelded components. Secondly, tensile residual stresses substantially increase the damaging effects of nominally compressive applied loading cycles.
2. Sensor design
The CrackFirst TM sensor, devised for welded steel structures and patented by Burdekin and Prescott in 1990, is based on fracture mechanics concepts. Figure 1 shows the final sensor developed. The sensor comprises a thin steelshim 51 by 20mm, with openings at each end, and a central slit, which acts as a starter notch for fatigue cracking. A series of coatings is applied to the shim to provide corrosion protection. The array of conducting tracks on either side of the central slit provides the means of detecting crack length in the shim: tracks severed by crack propagation from the slit result in loss of electrical continuity. The design of the track array, with twelve tracks on each side, allows the crack length on each side of the slit to be registered independently. Figure 1 is a fully assembled sensor, incorporating a steel reinforcing pad at each end of the sensor shim in the area where it is attached to the structure and connectors for the sensor electronics.
Fig. 1. CrackFirst TM sensor with connectors for electronics and reinforcing pads
When the sensor is attached to the target structure close to a critical joint, a fatigue pre-crack at the centre of the shim, introduced during manufacture, extends by fatigue crack growth under the action of cyclic stresses in the structure. The sensor design is such that the extent of crack growth in the shim is proportional to the Miner cumulative fatigue damage for a welded joint subjected to the same loading. By varying the dimensions of the shim, the fatigue sensitivity of the sensor can be closely matched to the family of fatigue S-N curves for different welded steel joints, for example those given in BS 7608.  Thus, a range of sensors could be produced to cover the known fatigue strengths for welded joints. In practice it is invariably the lower fatigue strength details, such as fillet welds, which govern fatigue performance of large plate structures. A single sensor matched to the lower joint classes, typically Classes E, F, F2 in the British Standard, will therefore meet the majority of requirements.
3. Sensor attachment
The sensor is attached to the structure by a combination of threaded studs and adhesive bonding. A capacitance discharge welding (CDW) gun is used to attach four threaded studs to the structure at the required locations. The sensor is then mounted on the studs and retained in place with nuts tightened to a specified torque. Locking nuts and thread locking adhesive are employed to prevent loosening in service. Adhesive bonding is achieved by injecting a two-part low temperature curing epoxy resin through the central opening in each reinforcing pad. Sufficient adhesive is introduced to fill the gap between the sensor and the substrate. A mask is used to prevent adhesive flowing out of the required bond areas, ensuring that the sensor is attached only at the ends.
Since welded joints in the as-welded condition generally contain high tensile residual stress at the point of fatigue crack development, both tensile and compressive applied load cycles will contribute to fatigue damage. In order to achieve the same response from the CrackFirst TM sensor, it is pre-stressed in tension during installation. Sensors should be installed as close as possible to the target joint so that they experience the same cyclic stress. The design of the pre-stressing jig is such that a sensor can be positioned within about 10mm of a weld toe. At this distance the sensor will experience the nominal stress in the member, as used in fatigue design assessment, and is not influenced by the local stress concentration due to the weld itself. Like a strain gauge, the sensor responds to the average stress over its active length - 50mm for the sensor described here. It is therefore better suited to fairly uniform stress fields, rather than steep stress gradients, and its size is such that it is not suitable for very small components. Sensor orientation also needs to be considered in the same way as that of a strain gauge, i.e. the longitudinal axis of the sensor is aligned with the direction of maximum principal stress.
In order to protect the sensor and the on-board electronics from mechanical damage and corrosion, a sealed enclosure is installed over the sensor installation, Figure 2.
Fig. 2. Typical sensor installation prior to sealing the protective enclosure (the sensor shim is beneath the sensing electronics)
4. Interrogation of the sensor and interpretation of sensor output
The on-board electronics unit checks the sensor status and records in memory the crack length on each side of the sensor. Sensor data can be downloaded to a PC via a communication port. Since the crack length changes relatively slowly the time interval between downloads can be large, of the order of months or years, depending on the frequency of the applied loading. A life of ten years is anticipated for the on-board replaceable battery supply.
The data indicating the crack length for each side of the sensor shim are used to estimate the expended fatigue life, expressed as a percentage of the design life for the appropriate joint class according to BS 7608.
Fig. 3. Fatigue test specimen with two CrackFirst TM sensors
5. Validation of sensor
A series of trials have been carried out to validate the sensor's performance. Figure 3 shows a welded test specimen undergoing fatigue testing with two CrackFirst TM sensors attached, one of which is connected to the sensor electronics. Specimens were manufactured from steel plate complying with BS 4360 Grade 50B. The fillet welds were made manually by shielded metal arc welding. Two CrackFirst TM sensors were attached to each specimen, adjacent to the joint and aligned parallel to the longitudinal axis of the sample. In most cases attachment was by CDW studs only, i.e. without the addition of adhesive. Tests were conducted under direct axial loading in air, using servo-hydraulic testing machines, and continued until complete separation of the specimen. The majority of tests were conducted under constant amplitude loading at stress ranges selected to achieve lives in the range 10 5 to 5 x 10 6 cycles. Values of the applied loading ratio, R, were 0.1 (pulsating tension loading) and -1.0 (alternating loading). A number of tests were also conducted with loading of variable amplitude.
An example is given in Figure 4 of the comparison between the fatigue samples and the CrackFirst TM sensors for some constant amplitude fatigue tests. All samples failed by fatigue cracking from the toe of the weld. The results plotted for the sensors show the endurances at which the last track on each side of the sensor was severed; results for the earlier tracks are omitted from this plot for clarity. Since two sensors were tested on each specimen, and each sensor gives two results (one for each crack tip) there are four sensor data points for each weld. The Class F mean and design curves from BS 7608 are shown for comparison.
Fig. 4. Fatigue test results obtained under constant amplitude loading
It will be seen that the results for the welds lie close to the Class F mean curve, i.e. the 50% failure probability curve, as expected for joints of this type. A small effect of stress ratio is evident, tests under alternating loading (R=-1.0) giving endurances similar to or slightly greater than those under fully tensile loading (R=0.1). In all cases the sensors failed conservatively before the welds. At R=0.1 failure of the last sensor track was close to the Class F design curve, which represents a nominal failure probability for the welds of 2.3%. Hence, as intended, the sensors gave a reliable and conservative warning of reaching the fatigue design life of the joint. An exception was the test at the highest stress range, 210MPa, where the results for the last sensor track were greater than expected. As shown in Figure 4, this test was repeated with sensors attached using the hybrid method, i.e. the combination of CDW studs and adhesive, giving an acceptable result. It was concluded that at this relatively high stress range, with a maximum tensile stress in the cycle of 233MPa, attachment by CDW studs alone was insufficient to guarantee adequate load transfer to the sensor, but this was corrected by adoption of the hybrid attachment method.
Figure 5 shows typical results for two sensors tested at a stress range of 90MPa, showing the full sensor output in terms of the number of severed tracks versus test cycles. The response is close to a linear characteristic and the prediction based on fracture mechanics. Typically the first track was severed at about 15% of the total life of the sensor and subsequent tracks failed regularly throughout the life.
Fig. 5. Typical results for two sensors under constant amplitude loading (stress range=90MPa, R=0.1) showing sequential failure of tracks
A number of validation tests were also conducted with loading of variable amplitude, using a loading history provided by one of the project partners (Caterpillar) obtained from strain measurements on the chassis of an articulated off-road truck. Stress ranges in the sequence were magnified to obtain failures within a maximum endurance of 30 million cycles. Test results for variable amplitude loading are plotted in Figure 6 in terms of the maximum stress range in the loading sequence versus the number of sequence repeats to failure. Results for the sensor are for failure of the last track. As observed under constant amplitude loading,all the sensors failed before the test specimens, with values in the range 0.44 to 0.84 for the ratio of the sensor endurance to that of the weld.
Fig. 6. Fatigue test results obtained under variable amplitude loading
Under the variable loading sequence adopted, the test welds gave lives significantly greater than predicted by Miner's rule, with values of the ratio experimental life/predicted life in the range 1.79 to 2.23. The sensors show a similar trend, but with values in the range 2.64 to 4.55. This is an important finding. It is well documented that under variable amplitude loading joints may survive longer than predicted by Miner's rule or may fail prematurely, depending on the type of loading sequence. Accurate prediction of such deviations from Miner's rule is not presently possible, even when the applied loading sequence is well characterised for example by strain measurements. On the basis of tests to date, it appears likely that the CrackFirst TM sensor will be a useful aid in predicting such behaviour. Clearly, investigation of other loading sequences is required in order to extend the validation, especially those for which Miner's rule is expected to give unsafe predictions due to crack acceleration.
In all tests special attention was paid to the performance of the attachment studs. No stud failures occurred during testing. On completion of each fatigue test the sensors were removed and the specimens were inspected carefully in the area of stud attachment by the magnetic particle method. No cracking was detected.
Field trials are also underway on an articulated truck and an excavator, with the assistance of Caterpillar who were a partner in the development of the sensor.
A fatigue sensor system for welded joints in steel, capable of providing advance warning of the rate at which the design life is being consumed, has been developed. Negotiations with potential licensees are in progress as the next step in making the system available commercially.
The CrackFirst TM system was developed through the collaboration of TWI Ltd., FMB, Micro Circuit Engineering Ltd., UMIST and Caterpillar Peterlee (a Division of Caterpillar (UK) Limited) in a project funded by the UK'sDepartment of Trade and Industry LINK Sensor and Sensor Systems for Industrial Applications Programme.
- BS 7608, 'Fatigue design and assessment of steel structures', British Standard Institute, London, 1993.