Evaluating Phased Array and Time of Flight Diffraction

Evaluation of the Phased Array Transmit-Receive Longitudinal and Time of Flight Diffraction Techniques

Channa Nageswaran and Colin R Bird

Paper published in Insight, vol.50. no.12. Dec. 2008. pp.678 - 684.

Abstract

The use of two phased array techniques for the inspection of a representative dissimilar weld from the power industry was investigated. The performance of the phased array Transmit-Receive Longitudinal (TRL) and Time-of-Flight Diffraction (TOFD) techniques was compared to the best-practice baseline procedure using conventional ultrasonics, developed specifically for the nozzle component. Slots were implanted in several regions (parent, HAZ and austenitic weld body) under a layer of cladding. The performance criteria were the detection of the slots, through-wall sizing and signal-to-noise.

Both phased array techniques performed better than the conventional techniques with regard to sizing, satisfying the ASME Boiler and Pressure Vessel Code^[1] (2007) Section XI Appendix VIII. The Root Mean Square (RMS) error of the phased array TRL and TOFD techniques were 0.52 and 1.04mm, respectively. In addition the phased array techniques showed similar signal-to-noise performance to the conventional transducers which were matched to the curvature of the inspection surface. Based on these results the use of phased array techniques for the inspection of thick ferritic-austenitic dissimilar welded components is recommended, confident in their ability to outperform conventional techniques.

Keywords: phased array, dissimilar welds, TRL, TOFD, sizing and signal-to-noise.

1. Introduction

Dissimilar metal welds are an important class of joints used in several key industries. This includes joints between nuclear reactor pressure vessel nozzles and primary circuit pipes, and welds in corrosion resistant clad pipes used in the petrochemical industry. There are many engineering issues related to welds between ferritic and austenitic (eg nickel-iron-chromium Alloy 800 or 316 stainless) steel components; considerable research has been performed to ensure that the metallurgy is adequate for the required mechanical properties across the weld. The non-destructive testing (NDT) of such welds to assure weld quality/integrity at the point of fabrication and during service is complicated by their inherent inhomogeneity and anisotropy imparted by the large austenitic grains. Ultrasonic inspection is currently the key NDT technique in use for the inspection of thick section austenitic welds and this paper presents the results of an investigation into the detection and characterisation of slot-like defect using two phased array based techniques.

The present paper focuses on a representative dissimilar metal weld from a nuclear power plant. Nouailhas et al^[2] discussed the inspection problem faced in the power industry with regards to dissimilar metal welds (also termed bimetallic or transition welds in publications). The defects expected in service are classified into two major types: volumetric (inclusions, porosity) and planar (lack of fusion, disbonds).

In this paper, a phased array focused time-of-flight diffraction (TOFD) technique and a twin crystal phased array 2D matrix probe were used to determine whether it was possible to improve the signal-to-noise (S/N) and sizing performance. The performance of these advanced solutions was evaluated against the best practice conventional TRL ultrasonic probes specifically designed for the inspection of the component.

The component for this paper was provided by an industrial organisation from the power sector. The component has an OD of 391mm, as shown in Figure 1, representing a dissimilar joint between a clad ferritic component and a forged stainless steel component.

Two phased array based inspection techniques were studied in the present work: the twin crystal phased array transducer and a TOFD configuration using two linear phased array probes. The development and characterisation of the twin crystal phased array probe (referred to hereinafter as the TRLPA probe) is further discussed by Nageswaran and Bird in this issue of Insight, and shown in Figure 2.

Figure 3 illustrates the phased array TOFD (paTOFD) concept. In this method, the transmitting transducer focuses the sound field to a very small volume, within the test component, where the defect is expected to appear; the transducer in receive mode then focuses on any waves reflected from the same region within the material. The paTOFD configuration was implemented with two identical 32 element linear array probes of 2.25MHz frequency at a pitch of 1.4mm (manufactured by Vermon, France). The physical dimensions of each probe were 54mm in length, 42mm in width and 25mm in height.

As part of the technique development, the use of dynamic depth focusing (DDF) was investigated and employed in the paTOFD technique. DDF is a processing technique used on the reception of ultrasonic signals and is applied by constructing variable depth reception delay laws as a function of time; it has been shown to provide better beam lateral resolutions and to increase the S/N performance away from the focal point. ^[3]

2. Approach

The TRLPA and paTOFD techniques were optimised for inspection of the component using modelling. The component was then inspected using the two phased array techniques and the best practice conventional technique specifically developed for the inspection of the component.

2.1 Technique optimisation through modelling

The CIVA^[4] modelling package was used to optimise the sound beam within the test component. Ideally, the microstructure of the weld material needs to be known so that the weld can be input into the model such that the distortive effects on the propagation of the ultrasound can be taken into consideration. However, since the test component could not be sectioned to evaluate the microstructural makeup of the weld, the simulations of the sound field and interaction with slot-like defects were performed on an isotropic material with velocities identical to the measured average velocities in the test component; as such the model did not incorporate the true anisotropic and inhomogeneous nature of the weld material or the cladding.

The aim of the study was to evaluate the performance of the two new techniques (TRLPA and paTOFD) in comparison with current procedures. Isotropic model outputs were used to optimise the parameters of the two techniques (wedge angle, coupling distances, probe separations etc) to ensure that the ultrasonic performance was sufficient to detect and size the implanted slots.

2.2 Calibration of the TRLPA and paTOFD techniques

The paTOFD technique was specifically targeted at detecting the diffracted signals, while the TRLPA technique was used to detect both the specular echoes and the diffraction signals. The delay law curves for both TRLPA and paTOFD techniques were generated for the MicroPulse 5PA instrument (manufactured by Peak NDT, UK) to execute the theoretically optimised inspection procedures. The probes used for the baseline scan were contact TRL probes specifically designed for the inspection of the component. The footprint curvature of the contact probes were matched to the outside surface of the component from which the inspection was performed.

Two calibration blocks were used: one whose measured velocities (longitudinal and transverse) were identical to that of the ferritic steel and one whose measured velocities were identical to the forged austenitic steel. The calibration was made on 3mm diameter side drilled holes.

2.3 Inspection

Spark-eroded slots with a semi-elliptical profile were introduced into the test component. Three sets of inspections were performed using:

1. Conventional TRL probes designed for the component (baseline).
2. The TRLPA probe in the same configuration used for the conventional probes.
3. The paTOFD configuration from the inside surface of the test component.

Figure 4 illustrates the cross section shape of the test component, weld geometry and position of slots. All slots are positioned on the outside surface, below the 6mm cladding which was applied after the slots were spark eroded. Three slot positions are defined: (I) embedded in the ferritic component, (II) embedded in the region of the Heat Affected Zone (HAZ) of the weld, and (III) embedded within the austenitic weld. The baseline and TRLPA inspections were performed from the outside surface of the component and the paTOFD inspection was performed from the inside surface. Note also the datum surface from which all probe positional measurements were made.

Figure 5 shows the dimensions of the test component cross section, component thickness (nominally 49mm) and the distances of the three slot positions from the datum. Table 1 shows the details of the slots within the test component; the tilt angle is defined in Figure 6. The edge just under the 6mm cladding is termed the 'root' and the edge within the substrate is termed the 'tip' (see Figures 6 and 7).

Four conventional TRL probes were used to generate the baseline inspection data:

1. 15° beam angle, 2.5MHz, 10mm diameter, semi-circular.
2. 30° beam angle, 2.5MHz, 10mm diameter, semi-circular.
3. 45° beam angle, 2MHz, 7.5 x 15mm, rectangular.
4. 70° beam angle, 2.5MHz, 8 x 15mm, rectangular.

All the probes generated longitudinal waves and inspected from the outside surface of the component. All the probes have dedicated single crystals in transmit and receive modes; twin crystal probes are recommended for the inspection of austenitic material as they improve the S/N performance by separating the travel paths of the beams in transmit and receive.^[5] All probes were contact probes (gel coupling), with shaped faces matching to the outside curvature of the component (OD 391mm). Furthermore, they align parallel to the component axis and are designed to inspect the flaws listed in Table 1.

The inspection was a blind trial, in that the operator was told the general region where the defects were present but was not given any information regarding their circumferential lengths or their through-wall sizes. The operator was required to size the defects' through-wall extent, the circumferential length, record the absolute gain setting when the target signal was at 80% FSH and collect the digital A-scan signals.

The performance criteria for comparison of the TRLPA and paTOFD techniques against the baseline conventional probes are:

1. Detection of the known slots (successful only when S/N>6dB).
2. Through-wall sizing.
3. S/N performance of the phased array techniques in comparison to established best practice probes for the test component.

The S/N is evaluated from the time-domain signals (ie A-scans) as the ratio between the maximum noise amplitude from the material ahead of the target and the target signal (when at 80% FSH).

In the TRLPA technique, through-wall sizing was performed using longitudinal waves diffracted from the roots and tips of the slots; the through-wall size of the defect is defined as the vertical distance (ie normal to the outside surface) between the root and the tip.

For the paTOFD technique, edge diffraction of longitudinal waves was again the primary route to through-wall sizing. Figure 7 illustrates the technique configuration and the parameters (transmit/receive wedge angles, transmit/receive coupling distances, probe separation and position) which were optimised and set through modelling results. TOFD inspections using single crystal probes are effectively used in industry for through-wall sizing of crack-like defects. The transmitter probes generate highly divergent beams (small crystal size) to 'flood' the material with sound which then diffracts at defect edges; the diffracted signals are very weak (about 25dB lower in amplitude in comparison to a SDH at the same range) such that the received signals need to be highly amplified by low-noise electronics;^[5] however, the microstructure of austenitic material also generates echoes which are detected and amplified, leading to noise. Hence the present study aims to investigate whether the ability of arrays to focus and concentrate the sound energy will lead to sufficient S/N performance for the detection of the weak diffracted signals in austenitic steels, while eliminating spurious background noise.

Table 1 Slots, geometry and dimensions embedded within the test component; length is circumferential distance, size is through-wall and tilt is defined in Figure 6.

3. Results

3.1 Summary

Table 2 summarises the performance of the three inspections against the three criteria outlined in Section 2.3 - detection, sizing and S/N, which are fully outlined in Table 3. In general, note that the baseline inspections using four different beam angles (15, 30, 45 and 70°) did not lead to sizing the through-wall extent of any of the slots. However, the average S/N performance of the baseline inspection was, in general, better than both the TRLPA and paTOFD techniques.

Table 2 Summary of performance using the baseline technique and the two phased array techniques, along the three criteria described in section 2.3. Successful detection is only when S/N on sentencing is greater than 6dB and average S/N is based on all the slots that were detected in each category.

The data presented in Figures 9 to 14 show the data from the TRLPA and paTOFD techniques. The distance scales are in mm and the colour scale for signal amplitude is from white (low) through blue, green and red (high). Each slot from Table 1 is referred to in the text with its position assigned as, for example, slot III-5 implies slot 5 at position III and slot II-J implies slot J at position II.

Consider first slot III-5 which was detected by the baseline 15, 30 and 45° beams with the probe facing the ferritic steel (as defined in Figure 8) at an average S/N of 13.1dB, whereas the S/N of TRLPA and paTOFD techniques was measured to be 11.5 and 18.5dB, respectively. However, neither the TRLPA nor the paTOFD methods were able to size the through-wall extent of slots III-5 and III-6 as the root diffraction signal was not detected by either technique (Figures 9 and 10); Figure 14 shows an example of when the root signal is detected using the paTOFD technique (slot II-N) enabling through-wall sizing.

Figures 9, 11, 12 and 13 show data using the TRLPA technique and the images show the scan composed of beam angles from 20 to 80°; the data shown is the signal received at receive array. Figures 10 and 14 shows data using the paTOFD technique and the images show an unconventional display of TOFD data, which is essentially a pitch-catch arrangement. The scans of Figures 10 and 14 are composed of beam angles from 0 to 70° and the lateral wave synonymous with the TOFD inspections is visible at the high beam angles.

With the probe facing in the direction of the stainless steel (Ss), slot II-G was detected at all beam angles of the baseline scans at an average S/N of 20.5dB; the average circumferential length was evaluated to be 6.25mm, whereas the actual length was 5mm. The TRLPA technique also detected the slot with the length sized as 5mm at a S/N of 9.5dB. The paTOFD technique not only detected the slot but also sized both the length and through-wall extent as 5 and 2.4mm, respectively, at a S/N of 7.1dB; the actual through-wall size was stated to be 3mm, implying an absolute error of 0.6mm in the paTOFD sizing performance.

In the case of slot II-H, the baseline scans were only able to detect the slot in the (Ss) direction at an average S/N of 15.3dB and sized the length as an average 4.2mm (the actual length was 5mm). However, the signal from slot II-H was considered too weak for the TRLPA technique in the (Ss) direction. In contrast, the paTOFD technique was able to size the through-wall extent of the slot to 2.9mm at an S/N of 7dB, whereas the actual size was 2mm, implying an error of 0.9mm in the measurement. The wavelength of the sound in the material at a frequency of 2.25MHz is about 2.6mm; since a slot with a through-wall size of 2mm was able to generate root and tip diffraction signals, this would suggest that both the beam spot and temporal resolutions of the sound wave is sufficient for a minimum through-wall resolution of 2mm.

On the other hand, the paTOFD configuration was able to detect only very weak signals from the roots of slots II-J and II-K, which were identical in length to slot II-H, but were 3 and 4mm in through-wall size, respectively. The TRLPA technique was able to detect both slots II-J and II-K (at an average S/N of 8.1dB) and was able to correctly through-wall size slot II-K as 4mm (see Figure 11). In addition, note that slots II-G and II-J are nominally identical (see Table 1) but only II-G was detected and sized by the paTOFD technique. This discrepancy in performance is reflected throughout the data sets of all three inspection techniques and illustrates the inconsistency associated with the inspection of dissimilar metal welds.

Of the remainder of the slots at position II, the paTOFD technique was able to through-wall size slots II-L, II-N, II-O and II-P with errors of 0.6, 1.7, 1.3 and 0.6mm, respectively, and the TRLPA technique was able to size slots II-M, II-N and II-O with errors of 0.2, 0.1 and 0.4mm, respectively.

In the case of slots at position I (I-Y, I-Z and I-a), the paTOFD technique could not be configured for inspection due to space constraints but both the baseline probes and the TRLPA probe were able to detect all three slots. The baseline 15° probe had an average S/N of 15.4dB, while the 30, 45 and 70° probes performed with average S/Ns of 13.5, 17.4 and 17.4dB, respectively. The overall average S/N performance of the TRLPA technique was 14.9dB. Hence the TRLPA probe's S/N performance was similar to that of the specific conventional probes designed for the inspection of the component. However, unlike the specific probes, the TRLPA technique was also able to perform through-wall sizing: the sizes of slots I-Y, I-Z and I-a were measured to be 3.9, 5.9 and 7.7mm, respectively - with associated absolute errors of 1.1, 0.1 and 0.7mm, respectively. As an example, Figures 12 and 13 show the detection of the root and tip of slot I-Z embedded within the ferritic substrate, using two different focal laws optimised for the corresponding depths.

Furthermore, the average lengths sized by the baseline probes for slots I-Y, I-Z and I-a were 5.2, 7.6 and 9.6mm, respectively, whereas the actual stated lengths were 8, 10 and 12mm. Hence, the average error in length estimation was 2.5mm for the baseline probes, while the average error in the length estimation by the TRLPA technique was 2mm. Therefore, the TRLPA technique's length sizing performance was also comparable to that of the baseline inspection procedures.

Considering all the results from all the slots at position II, the average S/N performance of the baseline, TRLPA and paTOFD inspections were 18.4, 10.3 and 7dB, respectively. A minimum S/N performance of 6dB is recommended by TWI and based on this criterion the paTOFD technique could be rejected as a viable option for primary detection but since it was able to size the through-wall extent of six slots (II-G, II-H, II-L, II-N, II-O and II-P), while the TRLPA technique sized four slots (II-K, II-M, II-N and II-O) and the baseline technique was able to size none, paTOFD could be used as a secondary route for sizing.

The average S/N performance of the baseline probes in detecting the two slots (III-5 and III-6) at position III was 14.7dB, whereas the average S/N performance of the TRLPA and paTOFD techniques were 9.7 and 20dB, respectively. Based on the minimum 10dB S/N criterion the TRLPA technique would be considered marginal; however, it would be possible to further optimise the TRLPA technique (in terms of probe frequency, pitch, ultrasonic characteristics etc) to improve the S/N performance.

Based on the slot set available for the investigation, the RMS error in the through-wall sizing of the TRLPA and paTOFD techniques are 0.52 and 1.04mm, respectively. The RMS error of 1.04 is similar to the value of 1.041mm quoted by Buttram^[6] for a similar manually applied phased array TOFD configuration. Hence, both techniques investigated in the present paper satisfy the requirements of the most stringent supplement of the ASME code^[1] (Section XI Appendix VIII), which requires the RMS error to be less than 3mm (Supplement 2).

3.2 Limitations

The S/N performance of the paTOFD technique in detecting the slots at position II was below 10dB. Reasons for this include the use of an optimised setup which required a larger water path (22mm) as opposed to a water path of 10mm in the configuration used for slots at position III. The subsequent front surface reflection signals led to an increase in the measured noise at the lower beam angles (below 35°) as shown in Figure 14 for the detection of slot II-N. The average gain setting for inspection of slots at position II was 52.4dB as opposed to an average gain setting of 44.9dB for the detection of slots at position III. The requirement to transmit through the weld body to approach the slots at the HAZ, compounded by a strongly scattering dissimilar metal interface is likely to have led to the weak signals.

The S/N performance of the TRLPA technique was about 6dB weaker than that of the baseline inspection. This is partly due to the TRLPA probe being designed for inspections that required defects deeper within the component thickness to be detected. Also, the probes of the baseline inspection were specifically designed for the inspection of the component with their contact surface curved to match the outside (inspection) surface of the component; this would have the effect of maximising focusing effectiveness and minimising losses across the interface. A TRLPA probe specifically optimised and designed for the inspection of the sub-clad defects present in the component should show better S/N performance.

It was not possible to evaluate the through-wall sizes of the two slots (III-5 and III-6) at position III using any of the three techniques. Unlike in the ferritic part where there was a clad-ferritic substrate interface, there was no property mismatch at the root of the slots embedded within the weld. Hence, a strong signal was not generated at the root for the paTOFD technique; similarly, no root signal was generated when inspecting from the outside surface using the TRLPA technique. It is possible that the distortive effects of the austenitic cladding prevented the generation of a strong root signal and introduced large errors. For example, the tip-diffracted signal in the TRLPA case for slot III-5 (see Figure 9) appeared at a depth of 13.3mm from the surface; assuming that the cladding layer was 6mm thick, it implies that the through-wall size was 7.3mm which is an error of 4.3mm compared with the actual size (3mm).

3.3 Implications of using phased array techniques

The TRLPA and paTOFD techniques were able to identify root and tip diffraction signals to enable through-wall sizing. This was largely made possible by the ability to focus the sound and display the data in a sector range such that signals could be indentified clearly. In addition, the TRLPA probe was able to replace several fixed-angle conventional probes in one unit, through the ability of the array to steer the beam and by focusing the sound pressure to specific depths, the sensitivity to defects as small as 2mm was increased.

The extension of phased array technology to TOFD applications was shown to be feasible by the present paper; the standard phased array instrument used in the project (the MicroPulse 5PA) was shown to be effective, in terms of signal amplification capacity, to detect diffracted echoes up to a depth of 43mm in an austenitic weld.

Furthermore, by replacing the standard TOFD single crystal probes (which generate highly divergent beams) with arrays, the full versatility of phased array technology is brought to bear on the inspection problem.

In general, the cost of choosing a phased array solution to industrial problems is relatively high in comparison to traditional approaches, due to the increased computing requirements, cost of specialised probes, advanced instrumentation, and need for highly trained operators. However, when compared to the cost implications of reduced inspection times, the phased array option rapidly becomes a cost-effective choice.

4. Conclusions

1. The application of two novel phased array techniques for the inspection of a 49mm thick dissimilar metal welded component with a 6mm clad layer were evaluated successfully.
2. Both techniques are ready to be adapted for application to industrial components.
3. Both phased array techniques are able to detect diffracted signals such that through-wall sizing was possible, compared to conventional TRL probes where sizing was not achieved.
4. The near surface performance of the TRLPA technique was excellent, enabling a 4mm sub-clad slot to be detected and sized.
5. The phased array probe is able to replace several fixed-angle probes.
6. The RMS error in through-wall sizing (as specified in the ASME Boiler and Pressure Vessel Code (1) (2007) Section XI Appendix VIII) of the TRLPA and paTOFD techniques was 0.52 and 1.04mm, respectively, making both techniques acceptable for industrial application.
7. The length sizing capability of both phased array techniques was comparable to the baseline inspection.

Acknowledgements

Industrial Members of TWI funded the work as part of the Core Research Programme.

References and footnote

1. ASME Boiler and Pressure Vessel Code, American Society of Mechanical Engineers, New York, 2007.
2. B Nouailhas, G V C Nguyen, F Pons and S Vermersch, 'Ultrasonic modeling and experiments: An industrial case: Bimetallic weld in nuclear power plant', Journal of Non-destructive Evaluation, Vol 9, No 2/3, pp145-153, 1990.
3. R/D Tech, 'Introduction to phased array ultrasonic technology applications', R/D Tech Inc., Canada, ISBN 0-9735933-0-X, 2004.
4. CIVA website, www-civa.cea.fr
5. J Krautkrämer and H Krautkrämer, 'Ultrasonic testing of materials', Springer-Verlag, New York, ISBN 0-387-51231-4, 1990.
6. J D Buttram, 'Manual ultrasonic phased array technique for accurate through-wall sizing of planar discontinuities in dissimilar metal welds', Materials Evaluation, pp 62-66, January, 2007.