Construction Weld Testing Procedures
Using Ultrasonic Phased Arrays

by Michael Moles* and Jinchi Zhang+

Figure 1-4
Figure 5-7
Figure 8-10

INTRODUCTION
Construction welds in pressure vessels and other components typically require testing to guarantee structural integrity. In the past, such welds were radiographed, though ultrasonic testing has become more prevalent in recent decades. These tests are performed to a code (discussed below). Perhaps more important to the practical engineer, all these tests have limitations, both on discontinuity detection and sizing.

In the last few years, a new technology has become available for testing welds - ultrasonic phased arrays. Phased arrays differ from conventional industrial ultrasonics in that beams can be focused, steered and scanned. While this permits new test techniques, it also means that codes originally developed for conventional ultrasonics (or even radiography) may be inappropriate for phased arrays.

---

As plate walls get
thicker, detection drops
rapidly with S-scans.

---

This paper uses computer modeling to analyze discontinuity detection using two common procedures: standard American Society of Mechanical Engineers (ASME) raster scans and sectorial scans (or S-scans), both easily performed by phased arrays. The emphasis is on midwall discontinuities, which are a known weakness of standard raster test techniques. A limited amount of experimental data is given to qualitatively support the general modeling conclusions and we present some recommendations.

 

Industrial Phased Arrays
Phased arrays use an array of elements, all individually wired, pulsed and time shifted. These elements are typically pulsed in groups of approximately 16 elements at a time for weld tests. With user friendly systems, a typical setup calculates the time delays from operator input or uses a predefined file calculated for the test angle, focal distance, scan pattern and so on (Figure 1). The time delay values are back calculated using time of flight from the focal spot and the scan assembled from individual focal laws. Time delay circuits must be accurate to around 2 ns to provide the required accuracy. Due to the limited market, complexity, software requirements and manufacturing problems, industrial uses have been limited until the last few years (Lafontaine and Cancre, 2000).

From a practical viewpoint, ultrasonic phased arrays are merely a method of generating and receiving ultrasound. Consequently, many of the details of ultrasonic testing remain unchanged; for example, if 7.5 MHz is the optimum test frequency with conventional ultrasonics, then phased arrays would typically use the same frequency, focal length and incident angle.

While it can be time consuming to prepare the first setup, the information is recorded in a file and only takes seconds to reload. Also, modifying a prepared setup is easy in comparison with physically adjusting conventional transducers. Using electronic pulsing and receiving provides significant opportunities for a variety of scan patterns.

 

Electronic Scans
Multiplexing along an array generates electronic scans (Figure 2). Typical arrays have up to 128 elements, pulsed in groups of 8 to 16. Electronic scanning permits rapid coverage with a tight focal spot. If the array is flat and linear, then the scan pattern is a simple B-scan. If the array is curved, then the scan pattern will be curved. Linear scans are straightforward to program. For example, a phased array can be readily programmed to test a weld using both 45 and 60 degree shear waves, which mimic conventional manual tests or automated raster scans.

 

Sectorial (Azimuthal) Scans
Sectorial scans use a fixed set of elements, but alter the time delays to sweep the beam through a series of angles (Figure 3). Again, this is a straightforward scan to program. Applications for sectorial scanning typically involve a stationary array, sweeping across a relatively inaccessible component like a turbine blade root (Ciorau et al., 2000) to map out the features and discontinuities. Depending primarily on the array frequency and element spacing, the sweep angles can vary from ±20 to ±80 degrees.

Sectorial scans are unique to phased arrays and can be used for weld tests. However, common sense (and computer modeling) indicates that the ultrasonic response will depend on angle of impact, location of the array and the thickness of the plate (Figure 4). S-scans will test a weld feature at a given angle for each array location, not at all angles. Thus, some discontinuities will be better positioned and oriented than others for detection; of course, this applies to all ultrasonic procedures, not just S-scans. However, the limitations of S-scans for construction welds have not been investigated yet.

 

Combined Scans
Phased arrays permit the combining of electronic scanning, sectorial scanning and precision focusing to give a practical combination of displays. Optimum angles can be selected for welds and other components, while electronic scanning permits fast and functional tests. This introduces the concept of tailored tests to optimize detection, sizing and testing time. This approach is discussed in ASTM E-1961 for automated ultrasonic testing of girth welds in pipelines (American Society for Testing and Materials, 1998).

 

CODES
Test procedures are based on requirements stipulated by standards and codes and then generalized to address the day to day requirements for typical components or industries. Thus, the codes are the key issues, with procedures or techniques following. There are many different codes available for ultrasonic testing; mostly these codes are written for conventional manual tests, with some permitting automated procedures. Almost no codes mention phased arrays, though most do not specify the method of generating and receiving.

In North America in particular, the American Society of Mechanical Engineers code family is dominant (American Society of Mechanical Engineers, 2003). However, there are a number of other codes available for specific uses: API 1104 and API 620 (American Petroleum Institute, 1999; 2002), DNV OS F101 (Det Norske Veritas, 2000), AWS D1:1 (American Welding Society, 1999) and international codes such as EN 1714 (European Committee for Standardization, 1998). ASME Code Case 2235 (American Society of Mechanical Engineers, 2001) is unusual in that it specifies automated tests, computerized data collection, performance demonstrations and fitness for purpose. In effect, ASME 2235 aims at using modern developments in automated ultrasonics and fracture mechanics.

None of the codes specify the actual procedure to be used; ASME generally requires that two angles be used in a raster pattern for coverage and that calibration be performed on side drilled holes. However, calibrating on side drilled holes is not demanding for phased arrays, using either raster scans or S-scans, as side drilled holes are omnidirectional reflectors. More importantly, the side drilled holes may not be representative of the inservice discontinuities and hence can give misleading setups and procedures.

 

MODELING
Modeling was performed using commercial software. A simulated 30 degree half angle weld profile was used, with six 5% notches either on the weld centerline or fusion line, as shown in Figure 5. A 1 mm (0.04 in.) cap was modeled, but the root was effectively flat. Three wall thicknesses were used: 12.7 mm (0.5 in.); 25.4 mm (1 in.) and 50.8 mm (2 in.). Two basic procedures were used: ASME raster scans at 45 and 60 degrees and sectorial scans (S-scans) from 35 to 70 degrees. These are standard tests on a common weld design, though there are many more possible configurations.

The S-scans were modeled at two positions: near the weld and far away. Near to the weld was defined as 5 mm (0.2 in.) from the toe for all thicknesses, a typical location. Far from the weld was defined as 5 mm (0.2 in.) plus 10, 20 or 30 mm (0.4, 0.8 or 1.2 in.) depending on the thickness. For 12.7 mm (0.5 in.), this would mean a distance of 15 mm (0.6 in.) from the modeled weld toe; for 25.4 mm (1 in.), the far distance is 25 mm (0.98 in.) and for 50.8 mm (2 in.), the far distance is 35 mm (1.4 in.). These are admittedly fairly arbitrary distances, but not unrealistic.

 

MODELING RESULTS

12.7 mm (0.5 in.) Plate
The 45 degree tests detected all the root and cap notches as expected (Figure 6a).

The midwall notches were not so detectable. Figure 6b shows the centerline notch detected, but by an indirect reflection that depends on cap geometry. There is effectively no direct reflection from the notch. Note that the strong reflection occurs much later than predicted, which suggests why some operators call discontinuities far from the actual location.

The 60 degree tests were not so satisfactory; however, the ASME code requires detection on only one angle, so this was an acceptable result overall.

The S-scans showed more variable results. In the near weld position (5 mm [0.2 in.] from the weld toe), the root notches were detectable at high angles. At 50 degrees the signal was direct, but at 60 degrees the returned signal was a fortuitous bounce.

The midwall notches were unpredictable. The optimum for the centerline notch was 50 degrees, based on a number of bounces (Figure 6c). As before, this type of reflection depends on actual array location, perhaps on cap geometry and on suitable bounces. Also, these multiple signals typically occur further down the time base and can lead to misinterpretation. The A-scan display shows the amplitude of the returned signal. Other angles showed lower amplitudes in this simulation.

The fusion line midwall discontinuity was only detected at 60 degrees, again through a fortuitous reflection off the plate bottom (Figure 6d).

For the far position, the centerline midwall discontinuity was detected at 55 degrees only, as shown in Figure 7a. Again, the signals bounced off the bottom of the plate before returning. Detection in this instance would depend on weld geometry and discontinuity characteristics.

Predictably, the far position S-scan detected the fusion line midwall notch clearly at a 60 degree incident angle (Figure 7b). This is a "one and a half skip" test and is a normal ASME procedure; it is also similar to the drawing in Figure 4.

In summary, S-scans detected all the notches in the welds for the 12.7 mm (0.5 in.) plate, at both the near and far positions. However, many of the signals came from fortuitous bounces and were neither direct nor predicted.

 

25.4 mm (1 in.) Plate
The results on the 25.4 mm (1 in.) plate were less encouraging. The root and cap notches were predictably detected, especially on the raster scans. However, the midwall discontinuities presented significant detection issues for both raster scans and S-scans. Figure 8a shows the near weld (5 mm [0.2 in.] from weld toe) centerline midwall notch S-scan at 45 degrees, which was the only S-scan to detect the notch at all. The signal is relatively weak and depends on multiple reflections and on the cap profile.

For the midwall fusion line notch in the near position, the results were much the same (Figure 8b). Only the 60 degree S-scan detected the notch, again by indirect reflections off the bottom of the plate.

In the far position (25 mm [0.98 in.] from the weld toe), only the 35 degree S-scan detected the notch, again by multiple reflections. The signal was strong, but far down the time base, which could cause significant positioning errors.

None of the S-scan angles in the far position detected the fusion line midwall notch, including the 60 degree "one and a half skip" test. Obviously, the predicted detection at 60 degrees depends on actual positioning of the array and on indirect reflections.

In summary, notch detection for 25.4 mm (1 in.) plate was worse than for 12.7 mm (0.5 in.) plate and unpredictable.

 

50.8 mm (2 in.) Plate
For the centerline midwall discontinuity using S-scans in the near position (5 mm [0.2 in.] from weld toe) and raster scans, there was no detection at any angle. In all cases, the beams were reflected away from the probe. Similar results occurred for the midwall fusion line notch; no detection was made at any angle.

In the far position (35 mm [1.4 in.] from weld toe), there was no detection of the midwall centerline notch at all. Likewise, there was no detection of the fusion line midwall notch at any angle. While signals were reflected around the plate as normal, in the thicker plates, these typically did not find their way back to the probe.

In summary, S-scans performed very poorly and unreliably on thicker plates.

 

EXPERIMENTAL RESULTS
Unfortunately, it is difficult and expensive to reproduce these modeled results experimentally, largely due to the problems and cost of making discontinuities. However, some preliminary experimental results using S-scans qualitatively support the modeling. Figure 9 shows a series of S-scans on a 25.4 mm (1 in.) plate looking at a fusion line artificial discontinuity. This discontinuity was measured at 19 mm (0.7 in.) below the surface (that is, two thirds of the way down to the root). The discontinuity is readily visible in the near S-scan positions, but gradually disappears in the far positions.

 

DISCUSSION
The modeling results here are discouraging, in the sense that phased array technology does not appear to be detecting discontinuities as well as expected. However, this is not necessarily the case; what is really happening is that a new procedure is being used that is neither tried nor tested in construction weld applications and the codes have not fully addressed this yet. One of the conclusions of the massive PISC II trials for the nuclear industry was that procedure was critical then (Bush, 1997) and is obviously just as critical now.

This is a procedural issue, not a technological issue; the test approach is the issue, not phased arrays. Phased arrays work well and offer many commercial advantages over conventional ultrasonics, including speed, flexibility and size. The problem is that S-scans (and raster scans) are not tailored to midwall discontinuities, or some corner discontinuities.

Sectorial scans have many useful applications: stress corrosion crack detection, testing for creep damage, hydrogen induced crack tests, shaft crack tests, small diameter tube tests, weld tests, nozzle tests, testing composites, rapid manual tests with good imaging and special applications (Dubé, 2004). The modeling shows that thin plates (less than 25 mm [0.98 in.] as a guideline) are generally tested using S-scans. However, thick plate tests are both unreliable (don't find the discontinuities) and unpredictable (results depend on probe position, cap geometry, discontinuity location and character). This is particularly true for midwall discontinuities; however, it should be pointed out that ASME rasters have significant limitations for midwall effects as well.

Despite the major benefits of computer modeling, it has significant limitations. All beams are calculated as rays with software, while actual ultrasonics is more complex. Discontinuities are simulated as flat reflectors, while real discontinuities are typically more omnidirectional (this will influence detection). S-scans are performed from specific locations, whereas in reality probes may scan from anywhere. The software also does not include diffraction. However, diffraction signals are typically 20 to 40 dB below pulse/echo signals and would be ignored in normal pulse/echo tests. One obvious step would be to compare raster scans and S-scans in the next plate test trials.

Despite these limitations, the implication of this study is clear: S-scans have severe limitations for thick construction welds. In particular, S-scans are both unreliable and unpredictable for critical midwall discontinuities. In contrast, ASME raster scans are unreliable but predictable for midwall discontinuity detection; since raster scans cover the whole area, if the discontinuity is well oriented, the operator can expect to detect it using automated scanning with raster scans. In practice, S-scans are probably more reliable with manual than automated scans, since the operator can scan around the weld as normal and quickly look for discontinuities.

The ASME Boiler and Pressure Vessel Code (2003) is not clear on the applicability of S-scan tests using a single pass. The code simply states that the search unit and beam angle selected shall be appropriate for the configuration being tested. This paper would argue that the angles are inappropriate since discontinuities are clearly going to be missed.

So, what can we do to improve these tests?

• First, use time of flight diffraction as well as ASME raster scans or S-scans. Time of flight diffraction is very good at detecting midwall discontinuities and is permitted under ASME Code Case 2235 (2001). Figure 10 shows a scan using both pulse/echo and time of flight diffraction; the midwall discontinuity at location 106.3 is not detected on any of the pulse/echo channels, but is clearly detected by time of flight diffraction.
• Second, it may be desirable to use multiple S-scan passes to improve discontinuity detection; however, this will slow down tests significantly and require the use of a second mechanical axis.
• Third, one excellent solution is to use tailored scans, along the lines of the ASTM E-1961 zone discrimination technique (ASTM, 1998). Also, it is easy to use a tandem probe arrangement with phased arrays, which is a good technique for midwall discontinuity detection.
• Fourth, it may be necessary to refine the codes to ensure that S-scans have higher discontinuity detection probability. Specifically, ASME Code Case 2235, which uses side drilled holes for setup and also performance demonstrations, may need to specify alternate reflectors and scans for S-scan tests.

 

CONCLUSION

• Computer ray tracing shows that both ASME and S-scans have detection limitations, particularly with midwall discontinuities.
• ASME rastering is more consistent, but misses the same discontinuities each time.
• S-scan detection depends on gate, geometry, skip pattern, wall thickness and location of the probe and is less consistent than raster scanning.
• As plate walls get thicker, detection drops rapidly with S-scans. Plates above approximately 25 mm (0.98 in.) show low detection rates.
• Limited experiments qualitatively support this modeling.

 

RECOMMENDATION

• Perform trials comparing ASME raster scans with S-scans to verify these results.
• To improve probability of detection, use tailored tests, if possible, and always use time of flight diffraction as well as raster or S-scans.
• Discourage S-scan tests for thick walled construction welds.
• Modify codes to ensure greater reliability from S-scans.

 

ACKNOWLEDGES
Chris Magruder, of R/D Tech, performed the experimental scans. Ed Ginzel, of the Materials Research Institute, Waterloo, Ontario, critiqued the paper. The software used was supplied by UTEX Scientific Instruments, Inc.

 

REFERENCES
American Petroleum Institute, Standard 1104, Welding of Pipelines and Related Facilities, 19th edition, Washington, DC, American Petroleum Institute, 1999.

American Petroleum Institute, API 620, Design and Construction of Large, Welded, Low-pressure Storage Tanks, 10th edition, Washington, DC, American Petroleum Institute, 2002.

American Society for Testing and Materials, ASTM E-1961-98, Standard Practice for Mechanized Ultrasonic Examination of Girth Welds Using Zonal Discrimination with Focused Search Units, West Conshohocken, Pennsylvania, ASTM, 1998.

American Society of Mechanical Engineers, ASME Code Case 2235-4, Use of Ultrasonic Examination in Lieu of Radiography: Section I and Section VIII, Divisions 1 and 2, New York, American Society of Mechanical Engineers, 2001.

American Society of Mechanical Engineers, "Nondestructive Examination," Boiler and Pressure Vessel Code, Section V, Article 4, New York, ASME, 2003.

American Welding Society, AWS D 1.1:2000, Structural Welding Code - Steel, 17th edition, Miami, Florida, American Welding Society, 1999.

Bush, S.H., "Ultrasonic Examination of Heavy-section Steel Components PISC-II and PISC-III Action 2 as They Apply to Nonnuclear Thick-walled Pressure Vessels," Welding Research Council Bulletin 420, New York, Welding Research Council, 1997.

Ciorau P., D. MacGillivray, T. Hazelton, L. Gilham, D. Craig and J. Poguet, "In-situ Examination of ABB l-0 Blade Roots and Rotor Steeple of Low-pressure Steam Turbine, Using Phased Array Technology," 15th World Conference on NDT, Rome, Italy, 11-15 October 2000.

Det Norske Veritas, DNV OS-F101, Submarine Pipeline Systems, Appendix D, Det Norske Veritas, 2000.

Dubé, N., Introduction to Phased Array Ultrasonic Applications - R/D Tech Guideline, M. Moles, ed., Mississauga, Canada, R/D Tech, 2004.

European Committee for Standardization, EN 1714: Non Destructive Examination of Welded Joints - Ultrasonic Examination of Welded Joints, Brussels, European Committee for Standardization, 1998.

Lafontaine, G. and F. Cancre, "Potential of Ultrasonic Phased Arrays for Faster, Better and Cheaper Inspections," NDT.net, Vol. 5, No. 10, October 2000, <www.ndt.net/article/v05n10/lafont2/lafont2.htm>.

 

* R/D Tech, 73 Superior Avenue, Toronto, Ontario M8V 2M7, Canada; (416) 831-4428; fax (416) 255-5882; e-mail <michael.moles@rd-tech.com>.

 ⁺ R/D Tech, 505 boul. du Parc Technologique, Québec PQ G1P 4S9, Canada; (418) 872-1155; fax (418) 877-0141; e-mail <jinchi.zhang@rd-tech.com>.

Copyright © 2004 by the American Society for Nondestructive Testing, Inc. All rights reserved.

 

---

 
Copyright © 2010 by the American Society for Nondestructive Testing, Inc. ASNT is not responsible for the authenticity or accuracy of information herein. Published opinions and statements do not necessarily reflect the opinion of ASNT. Products or services that are advertised or mentioned do not carry the endorsement or recommendation of ASNT.

IRRSP, NDT Handbook, The NDT Technician and www.asnt.org are trademarks of the American Society for Nondestructive Testing, Inc. ACCP, ASNT, Level III Study Guide, Materials Evaluation, Nondestructive Testing Handbook, Research in Nondestructive Evaluation and RNDE are registered trademarks of the American Society for Nondestructive Testing, Inc. ASNT exists to create a safer world by promoting the profession and technologies of nondestructive testing.