Channa Nageswaran, Alan Day and Kim Hayward (TWI Ltd)
TWI Member Publication. September 2013
This paper presents the design and implementation of ultrasonic techniques for the monitoring of crack growth initiated in threads during a full scale resonance test on a subsea component. Ultrasonic techniques were required to detect the onset of cracking and subsequently monitor the propagation of the cracks until failure of the component by full through-wall breach. The techniques made use of a 256 element two-dimensional array probe in order to generate a sound field insensitive to the strong echoes generated by the geometry of the threads such that the initiation of the cracking could be detected early. The performance of the inspection techniques was evaluated by sectioning the failed component. The techniques continue to be refined in TWI through ongoing projects with a view to in-service monitoring of threaded subsea components aiming to capture the dynamic nature of fatigue failure.
Ultrasonic inspection techniques were designed and implemented to detect the onset of cracking, size the cracks and monitor their growth during a fatigue test on a large full scale component used in the oil and gas industry. The design of a phased array technique, its implementation, the results and the evaluation of the outcomes is presented. In particular, the use of a 5MHz 256 element two-dimensional array probe to implement electronic skewing was specified for improving the inspection capability for the required task. The phased array technique was complemented by a conventional single element probe pulse-echo technique to enable the complete inspection system to monitor the failure of the component when a crack, originating from the threading, propagates fully through-wall.
The component consisted of a threaded connection as part of the fluid transport system. The outside diameter of the component was approximately 20” (inch) and the component installed in the TWI fatigue testing rig was approximately 9m in length with a dry weight of approximately 18 tonnes, as shown in Figure 1. It was known that the primary mode of failure was by cracks which initiate at the threading and propagate through the pipe wall to failure. The component was fatigue tested in air at ambient temperature in rotating bending using the resonance technique. The ultrasonic inspections were done at the start of the test (the fingerprint scans) and at the end of fatigue test blocks which were composed of a number of cycles, ranging from 250,000 to 1,500,000 with nominal bending moment ranging from +/-300kNm to +/-700kNm.
Figure 1 - The test component installed in the fatigue testing rig in TWI
Two previous fatigue tests on two similar components had found that the ultrasonic inspection techniques used for monitoring the failure were deficient in through-wall depth sizing capabilities. The conventional techniques used earlier had been shown to detect electrical discharge machined (EDM) slots of through-wall depths 3, 5 and 10mm but had been found to be deficient in size discrimination. Hence a phased array solution was sought that would be able to size the cracks as they emerge from the threading and monitor their growth during the test through to failure.
A two-dimensional (2D) array probe was chosen to be able to electronically skew the sound beam to eliminate the strong corner echoes generated in the threading which would interfere with effectively interpreting the onset of cracking damage. The high number of elements (256) on the array allowed for very precise control of the sound field, which had to be projected over a long range due to the large size of the component.
The phased array technique was designed using the CIVA  and the UltraVision software  and executed through the DYNARAY system  containing 256 independent pulser-receiver channels. The inspection was automated by the MAGSCAN system . The conventional technique was implemented by the PS4 P-Scan system .
At the completion of the fatigue test the component was sectioned as part of the post-test failure analysis. The performance of the inspection techniques was evaluated and recommendations made towards further improvements for a future series of tests on similar components that are presently ongoing.
The component was a fluid transport system that contained a threaded connection. The primary aim of the project was evaluation of the design of this threaded connection in order to maximise the fatigue life of the component. Hence the two primary inspection goals were to:
- Identify the onset of damage on the corners of the threading,
- Measure the through-wall size of fatigue cracks as they grew
The component was fabricated in carbon steel which is isotropic to the propagation of the sound, ie the properties of the material remain identical regardless of the direction in which the sound propagates. The ferromagnetic nature of the steel allowed robotic scanning devices to be deployed for automated scanning where they remained attached to the component by magnetism, as described in Section 2. The component was large and held in a horizontal position (as shown in Figure 1); additionally, certain geometric features of the component design dictated the positioning of the probes to specific zones in order to achieve the inspection goals.
In the work presented here, the inspection was undertaken at the start of the fatigue test (fingerprint scans); then after a certain number of cycles the test was interrupted for inspection, and so on until there was a full through-wall crack. Hence the monitoring was done at discrete points during the cyclic loading of the component, which took approximately two months to failure.
Sending beams to the threading orthogonally, illustrated in Figure 2, leads to high amplitude echoes at corners marked 1 which can effectively mask the onset of cracking at the corners marked 2 which are most likely to crack.
Figure 2 - Sound beams from a probe being incident on corners of the threading marked 1 in an orthogonal plane which gives rise to complete reflection at the corners and consequently high amplitude echoes which mask much weaker echoes due to onset of cracking from corners marked 2
Hence by projecting the beam onto the threads in a skewed plane away from the perfectly orthogonal plane to the threading, the high amplitude echoes from corners marked 1 in Figure 2 were effectively eliminated so that the echoes from cracks emanating in corners marked 2 could be confidently and effectively detected at onset. One way to achieve this would be to design solid wedges for a conventional single element probe with a cut designed to refract the sound beams into the component along the necessary plane . Each wedge will be constrained to project the sound beam along a particular plane which means that several wedges would need to have been cut as part of the optimisation process for the inspection technique. Instead, a 256 element 2D phased array solution allowed the design and fabrication of a single optimised wedge to nominally project the sound in the orthogonal plane (as in Figure 2) as well as along a range of skewed planes by electronically changing the delay laws used to phase the sound field .
Figure 3 shows the design of the inspection technique using the CIVA simulation platform where the probe on its wedge (fabricated using Rexolite) is located on the outside surface of the component; note that only a sector of the component is shown. The design of the probe and wedge were optimised to provide the required sound field over a range of beam angles where the criteria was (1) field intensity at the location of the threads, (2) beam size at the threads to achieve imaging resolution and (3) uniformity over the full length of threading in the component so that only one circumferential scan was required on the outside surface of the component. TWI has confidence in the theoretical output of the CIVA platform as it has been validated for this task in TWI’s Core Research Programme [7, 8].
Figure 3 - Simulation of the inspection technique in CIVA in order to optimise both the 2D array probe and the Rexolite wedge
The first iteration in the probe design process assumed it to be a linear array probe with the lengthwise primary axis divided into elements with widths equal to width of the probe. The second iteration divided the first iteration optimised width of the probe further to create smaller but more numerous elements. The resultant 2D array then allowed control of the sound field in a full three-dimensional volume instead of just in the two-dimensional plane aligned to the primary axis of a linear probe. Note that the wedge optimised in the first iteration when the probe was one-dimensional, providing refraction along the primary axis plane only, was not modified to provide refraction along skewed planes, so that the skewing was achieved only through the use of electronic delay laws, with no physical aid to skewing the sound beams provided by the geometric cut of the wedge.
The total number of elements to make the 2D array was restricted to a maximum of 256 elements leading to an optimised matrix of 32 by 8 elements. The frequency was optimised to 5MHz to be able to propagate the required distance to the threads while providing sufficient field conditions to achieve the necessary sensitivity and resolution. Figure 4 illustrates the concept of an un-skewed beam in the plane aligned to the primary axis of the probe – which is identical to the orthogonal plane illustrated in Figure 2 - and the effect of skewing (or ‘steering’) the beam away from this plane.
Figure 4 - The views termed top and side when looking down onto the threads and sideways onto them, respectively, illustrate the path of a sound beam (as the red line) when the beam is confined to the orthogonal plane of Figure 2 (above) and when it
The scanning was done using an automated scanner which was able to remain attached to the component by magnetic wheels and move the probe and wedge around the full circumference of the component while maintaining the position of the probe with respect to the threads precisely. In addition the scanning was fully automated such that the operator could control and analyse the data at a distance from the component. Figure 5 shows the probe on its wedge and the scanner deployed on the component.
Figure 5 - The MAGSCAN scanner holding the probe and wedge on the outside surface of the component. The threads were on the inside surface, beyond the flange to the left of the image
The technique was designed to generate a range of beam angles onto the threaded region of the component so that the resultant sector scan (S-scan) image is able to monitor the threaded area of interest. Figure 6 shows data using un-skewed beams onto the threads before the resonance test was started, where the signals from the corner 1 of the threading (see Figure 2) can be seen in the top and side views. The dominance of the signals from the threading in both views make it difficult to differentiate signals that may arise at thread corner 2, as illustrated in Figure 2. Note that the full circumferential distance around the component is represented in the top view, the scale along the bottom starting from the datum at 0 on the left to over 2000mm on the right.
When, however, the beams are skewed (in this case by +10º) away from the orthogonal plane, the strong echoes from the corner 1 of the threading are effectively eliminated, as shown in Figure 7. The echoes highlighted in boxed red in Figure 7 are from the corner 2 (see Figure 2) edges of threading which were cut through by two keyways present in the component to aid alignment. These geometric features, firstly confirm that cracking from corner 2 could be detected by the skewed beams and could also be used to establish required sensitivities to such cracking.
Figure 6 - Images of the top and side views of the threads using un-skewed beams in the orthogonal plane where the thread corner 1 (see Figure 2) echoes dominate the viewing space
Figure 7 - Images of the top and side views of the threads using beams skewed away from the orthogonal plane by 10º showing the effective elimination of echoes from corner 1 (see Figure 2) so that echoes from corner 2 could be more readily detected and interpreted
The sensitivity for the technique was also set using EDM notches of different depths into the component away from the threading in order to ensure that the tip of the cracks as they propagate towards the outside surface of the component could be detected. The sensitivity – or gain/amplification of signals – required was high in order to be able to detect the tip of a fatigue crack after initiation from corner 2. Figure 8 shows the data from an EDM notch 15mm in depth from the threads, highlighted in a dotted red box.
Figure 8 - Detection of the tip of an EDM notch highlighted in the dotted red box 15mm deep from the threads using un-skewed beams
Following a series of trials on full scale mock up specimens, the procedure was finalised to the use of skewed beams for the early detection of the cracking on the thread corners and then to use the un-skewed beams to monitor and measure the growth of the cracks away from the thread lines.
The results are presented in chronological sequence from the initiation of the fatigue test when the threads were free of damage, as presented in Figure 7, and onwards through the test with the onset of damage leading to complete wall breach.
The onset of damage was detected at approximately 35% of the full fatigue life of the component and the scan using the skewed beams at first detection is shown in Figure 9, adjacent to the alignment keyway. After further loading, the crack shown in Figure 9 was seen to increase in size with additional points of damage on the threading, as shown in Figure 10.
Figure 9 - Detection of the onset of cracking damage in the threads adjacent to a keyway
Figure 10 - Growth of crack shown in Figure 9 as measured in the S-scan as well as the presence of other cracks along the threading
The through-wall depths of the cracking could be confidently measured in the skewed beam scans using diffracted signals from the tips once their through-wall height exceeded approximately 6mm.
As the cyclic loading continued it was clear that the majority of the damage took place within three adjacent threads. As the cracks grew towards the outside surface of the component, the un-skewed beam scans were utilised to monitor the cracks. Figure 11 shows the direct but very weak diffracted echo detection of the tip of a fatigue crack using un-skewed beams late in the fatigue life of the component, close to final failure. Figure 12 shows the corner echo generated between the outside surface of the component and the fully through-wall crack which led to the final failure of the component.
Figure 11 - The weak tip diffraction echo from a fatigue crack using un-skewed beams; note also the existence of several regions of major cracking
Figure 12 - The echo from the crack (highlighted in dotted red box) which reached the outside surface of the component, leading to final failure
The performance of the phased array inspection system was evaluated both during and after the fatigue test. During the fatigue test, the phased array technique was run parallel to a previously trialled conventional ultrasonic inspection procedure based on the P-scan system. This allowed a level of cross-checking between the two systems. Figure 13 for example shows the detection of the onset of cracking damage using the conventional system, corresponding to the data presented in Figure 9 from the phased array system. Unlike the phased array system, the conventional ultrasonic system was unable to provide crack growth information and was relatively more difficult to discriminate actual damage in the early stages in comparison to the background noise – as between Figures 9 and 13.
Figure 13 - Detection of early damage adjacent to one of the keyways using the conventional ultrasonic system, corresponding to the data in Figure 9, highlighted in the dotted red box
Post-failure of the component a comprehensive failure investigation was carried out on the component to establish the mode of failure and verify findings of the monitoring inspections done during the fatigue test. The post-failure investigation included further inspections using magnetic particle inspection (MPI), eddy current testing (ECT), alternating current potential difference (ACPD) and visual analysis of sections through the threading.
The phased array ultrasonic technique was found to be effective in mapping the distribution of cracking, able to detect cracks of 3-4mm in extent. There were however errors in positioning cracks to the correct threads, with several distinctly different cracks being classed as one and cases where one real crack being reported as distinctly separate emanating from different threads, as discriminating the signals were exacerbated by the majority of damage being confined to three threads. An additional limitation was the masking of smaller cracks behind larger ones as the sound could not penetrate the larger one. These effects are illustrated in Figure 14 which shows two cracks from two separate threads.
Figure 14 - Two fatigue cracks from different threads
This performance in positioning and detecting masked cracks could be improved if there was complete freedom to position the probe on the outside surface of the component, at the cost of increased inspection times.
Where the weak tip diffraction signals from the cracks were detected (as shown in Figure 11) and where they were compared to sectioning data close to the corresponding circumferential section from datum, the through-wall sizing error between the phased array system compared to MPI/ACPD/sectioning ranged from 1 to 12mm. This error will depend on the local crack tip conditions with respect to the incident beam. As can be seen in Figure 14, the trajectory of the cracks change significantly as they propagate and the likelihood of their relative orientation to the incident beam being favourable for generation of sufficiently strong diffraction echoes is difficult to predict. If, as for the positioning of the cracks above, it was possible for significant movement of the probe on the outside surface of the component then the techniques can be designed to maximise this likelihood.
The following conclusions draw together the findings in this paper:
- Phased array ultrasonic techniques were successfully developed to monitor the failure of a large threaded component during a resonance fatigue test.
- The use of a high element count two-dimensional phased array probe allowed the use of skewed beams to become insensitive to the strong echoes from the threading so that the onset of damage could be detected early during the fatigue test.
- The technique was found to be effective in mapping the cracks in the threaded region but issues with positioning of cracks and discriminating between separate cracks close together remain.
- The through-wall sizing accuracy was found to be good with the maximum error found to be 12mm with cracks ranging in size from 20 to >50mm in through-wall size.
- It is believed that both the positioning and through-wall sizing performance can be improved if it was possible to move the scanning system over a larger surface area on the outside of the component. This would however incur greater inspection times.
- The techniques presented have been successfully applied and continue to be adapted for in-service monitoring inspection solutions for a range of threaded components critical to the operations of TWI Members.
References and footnotes
- CIVA (www.extende.com)
- Zetec Inc. ( www.zetec.com)
- Phoenix ISL ( www.phoenixisl.com)
- Force Technology (www.forcetechnology.com)
- Krautkrämer J and Krautkrämer H, ‘Ultrasonic testing of materials’, New York, Springer-Verlag, 1990
- R/D Tech, ‘Introduction to phased array ultrasonic technology applications’, Quebec, 2004
- Schneider C R A, Kleiner D and Williams S M, ‘Optimising focussed/phased array ultrasonic testing of clad pipe: Initial verification of CIVA 3D modelling software’, TWI Members Report 836, 2005
- Nageswaran C, Schneider C R A and Decourcelle N, ‘Validation of theoretical models for focused/phased array ultrasonic inspection’, TWI Members Report 896, 2008