Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Immersion Transmit-Receive Longitudinal phased array probe for stainless steel (December 2008)

Channa Nageswaran and Colin R Bird


Alison Whittle
Peak NDT Ltd
Unit 7, The Derwent Business Centre, Clarke Street, Derby, UK, DE1 2BU
01332 738 752

Paper published in Insight, vol.50. no.12. Dec. 2008. pp. 673 - 677.


The difficulties associated with the ultrasonic inspection of stainless steel welds are well known and researchers around the world are working on the problem of improving the inspection quality. There are many issues associated with inspection, including accessibility and critical defect sizes, but the most important issue regards the accurate positioning, sizing and characterisation of flaws. Recently, the use of phased array technology to address these issues have been implemented and studied by various institutions and organisations. An immersion coupled twin crystal 2D matrix phased array transducer was developed in order to improve the signal-to-noise performance and enhance the positioning and sizing accuracy. The Transmit-Receive Longitudinal (TRL) probe was optimised for inspection from near surface (sub-clad) to around 50mm in thickness.

Keywords: phased array, transmit-receive longitudinal, TRL, signal-to-noise.

1 Introduction

The austenitic weld material is composed of large columnar (dendritic) grains; the grains grow with the <100> crystallographic axis parallel to the maximum thermal gradient during solidification.[1,2] The large anisotropic grains result due to the absence of phase transition during cooling, leading to no grain refinement; the grains can be clearly observed in optical micrographs and measure several millimeters. This leads to grain boundaries having an appreciable effect on the propagation of sound, with the following implications for ultrasonic inspection:

  • Increased scattering from grain boundaries, leading to increased ultrasonic noise.
  • Reflection, refraction and mode conversions at grain boundaries.
  • Sound beam skewing due to the anisotropy and the inhomogeneity, leading to inaccurate sizing and positioning of defects;
  • Interference between reflections at multiple interfaces, leading to a phenomenon where nonexistent features are apparently observed (false calls).

The problems and limitations associated with the ultrasonic inspection of austenitic welds are well known throughout the power and petrochemical industries.[3] Past experimental studies have shown that the longitudinal wave is best suited to the inspection of austenitic welds.[4] To mitigate the effects of scattering, low frequency ultrasonic transducers are used such that the longitudinal wavelength is suitable for inspection of the coarse grained material; the conventional single crystal probes currently used for austenitic inspection rarely exceed 2MHz.[4] Focusing has been shown to be a route to improve the signal-to-noise performance of the inspection since increasing the intensity of the sound field at the target region improves the detectability of defects. [5,6 and 7]

A recent paper[8] which investigated the use of a large aperture twin crystal focused probe for the inspection of cast stainless steel welds concludes that the focusing method led to a signal-to-noise ratio improvement of 15dB at a focal depth range of 60-70mm. The transducer was 1MHz in frequency with a broadband signal which aids in the reduction of the scattering noise amplitude; the noise level is further reduced by separating the travel paths of the transmitted and received sound beams[9], which is the basic principle for using twin crystal probes. Disadvantages included internal echoes from within the wedge leading to a dead zone, reducing the ability to detect and resolve flaws near the surface and difficulties in implementing practical inspection using large probes due to geometrical constraints.

With the advent of instruments capable of addressing a large number of channels the development and application of phased array twin crystal probes has become possible. [10]

The French nuclear industry[11] has investigated the use of twin crystal phased array probes for the inspection of coarse grained components. These have shown an improvement of 10dB in the signal-to-noise when a 800kHz twin crystal phased array probe is compared to a 2MHz 64 element linear phased array probe. Investigations into optimising the frequency for austenitic weld inspection using twin crystal phased array probes was recently performed in the US power industry;[12] four probes ranging from 500kHz to 1MHz in frequency were used in the study and the researchers conclude that the 500kHz probe showed the best specular reflection performance, but it was not possible in all cases to detect the crack tip diffraction echoes.

The present paper describes the development of a unique immersion coupled TRL phased array probe for the inspection of an austenitic dissimilar metal weld between a clad ferritic Reactor Pressure Vessel (RPV) and its stainless steel primary circuit piping (the safe-end weld).

2 Approach

2.1 The problem

The technique was developed for an 85mm thick dissimilar weld configuration as illustrated in Figure 1 (known as the safe-end). The austenitic weld is a K-prep between an austenitic stainless steel nozzle and ferritic RPV section. Two test blocks, one with 4.5mm (test block A) and the other with 3mm (test block B) Side Drilled Holes (SDHs), were used to develop the technique. The SDHs were placed along the ferritic to weld interface (side 1) and the stainless to weld interface (side 2), as shown in Figure 1. Furthermore there was a nickel buttering layer between the ferritic to weld interface and an austenitic cladding layer 8-10mm thick.


Fig. 1. The safe-end weld showing the materials, K-prep 85mm thick weld, the SDHs along the buttering (side 1) and the SDHs along the stainless-weld fusion face.

Figure 2 shows the two inspection scan directions, defined as 'Ferritic Scan' (FS) and 'Austenitic Scan' (AS). The effect of the weld on ultrasonic propagation was investigated by scanning for SDHs 1-4 (in AS) and for SDHs 5-8 (in FS). The stainless steel parent was forged and hence its properties did not adversely affect inspection in direction AS; however, the effect of the cladding layer (due to anisotropy) is known to influence the propagation of the sound and lead to errors in the positioning and sizing of flaws.


Fig. 2. Illustration of the two inspection configurations termed 'Ferritic Scan', FS and 'Austenitic Scan', AS.

A compromise had to be reached between frequency and penetration: the higher the frequency the potentially better sizing ability but the poorer the signal to material noise ratio. For this reason a great deal of effort was invested in this stage of the project.

A section approximately 13mm in thickness was machined from a test block to study the microstructure; the dendritic grain structure of the weld is shown in Figure 3 and detailed micrographic analysis is shown in Figure 4. The stark change in weld microstructure across the dissimilar weld interface, as shown in Figure 4, has implications towards the propagation of ultrasound across the interface. Electron Back Scattered Diffraction (EBSD) was used to map the orientation within the grains, across different grains and to identify the degree of misorientation within grains in the weld. Figure 5(a) shows a scan of an area 2.8mm x 4.5mm on the weld using stainless steel 304 as the base for analysis of the crystallographic orientation on the surface. Note that the colour scheme does not give information as to the orientation in this representation.


Fig. 3. The etched 13mm thick weld specimen showing the dendritic grains in the weld region.


Fig. 4. The elongated grains within the weld (left) and the ferritic to weld interface (right)


Fig. 5. EBSD analysis of the austenitic weld specimen to determine the degree of anisotropy:

a) Identifiable grains within a 2.8mm by 4.5mm area (not to scale)
b) Orientation of crystals (not to scale)
c) Pole figure analysis; the pattern suggests a fibrous texture
d) Colour code to determine orientation of a point in (b); the dashed region shows the area in which most of the orientations lie

Figure 5(b) shows an inverse pole figure analysis to determine crystallographic orientation. It shows that what was identified in Figure 5(a) as a single grain (for example the large red region) actually contains regions of varying orientation. The pattern generated in the pole figure analysis, shown in Figure 5(c), seems to suggest that the texture is fibrous. In general, even the small region examined appears to contain multiple orientations as shown by the clusters in the pole figure analysis, confirming that the weld material is highly anisotropic.

The micrographic and EBSD analysis confirmed the grain sizes to be several millimeters, which supports past experience showing 2MHz to be a sensible highest inspection frequency. Trials using conventional probes and linear phased array probes were carried out to determine defect sensitivity at various frequencies and 2.5MHz was chosen for the TRL frequency, which pushed the upper frequency limit of current conventional probes used for austenitic weld inspection.

2.2 Optimisation of the TRL probe

As introduced, current best practice is to use large TRL probes for the inspection of austenitic welds. Furthermore, phased array probes generate diffraction grating lobes if the element size is greater than half-wavelength. To provide a large phased array probe, either it is necessary to use large elements or many elements. Hence to generate a probe with sufficient aperture to focus at the desired range without generating large diffraction grating lobes a probe with a minimum of 128 elements was required. Practical financial constraints must be taken into consideration as the cost of the array controller increases with number of available channels; the total number of elements for the probe was limited by the 128 channels available on the TWI MicroPulse 5PA array controller, manufactured by Peak NDT (UK).

For detection, a signal-to-noise performance greater than 6dB is recommended by TWI; furthermore, to maximise the sizing performance the beam widths at the inspection range needed to be minimised. The SimulUS [13] model was used to optimise and specify the array parameters of the TRL probe to achieve these objectives. The need to inspect for Stress Corrosion Cracking (SCC) under the cladding (10mm) and for flaws within the weld volume to a maximum depth of 85mm imparted conflicting limits on the array parameters. The primary requirements were to eliminate unfavourable grating lobes when steering from 40° to a maximum steering angle of 70° and suppress the strength of the transverse wave.

In general, the greater the number of elements available the easier it is to achieve good signal-to-noise because less compromises have to be made between element pitch and array size. The probe was designed with 64 elements for transmission and 64 elements for reception. Both transmit and receive arrays were operated symmetrically, that is the delay laws applied on reception of the sound mirrored the laws applied to generate the sound field. Figure 6 visualises the TRL probe, showing the characteristic parameters. The transmit and receive arrays are each 16x4 2D matrix.


Fig. 6. The design of the immersion coupled TRL phased array probe. The axes convention is based on weld testing where the probe is scanned along the axis parallel to the weld centreline (scan axis) and movement along the index axis translates the probe towards (or back from) the weld centreline

The phased array data representation used in this paper makes use of beams from 40° to 80° 'stacked' to create what is known as the sector or azimuthal scan.[14] All such sector scans shown in this paper are created in the index - depth plane.

In addition to the roof angle and the frequency (2.5MHz) the following parameters were determined through an iterative modelling programme:

  • number of elements in the index axis direction;
  • number of elements in the scan axis direction;
  • pitch in both index and scan axes directions;
  • wedge angle.

The use of 64 elements in the transmit array leads to better manipulation of the sound field through 'phasing'; similarly the use of 64 elements on reception allows the processing of received echoes with greater resolution. Hence the parameters above were optimised for use of all 128 channels available in the array controller.

Since the in-service inspection was to be carried out in immersion, the sound was coupled through water. As well as being more accomodating towards varying surface profiles, the lack of a separator (as would be required in a solid wedge) enables greater lateral steering (ie along the scan axis). Furthermore, the use of immersion coupling makes the design unique and provides insight into the performance of a previously unexplored type of twin crystal design.

A geometrical model in the scan-depth plane, Figure 7, was used to select maximum refraction angles along the scan axis and specify the value for g (half-width separation), which was set so that the transmitted and received beams did not overlap at the water to component interface, leading to cross talk. This required the iterative optimisation of the roof angle and the coupling depth h to allow the focusing of the sound beam over the required depth range in the component.


Fig. 7. Parameters of the geometric model to define the parameter g, which represents the physical separation of the transmit and receive arrays. The figure is a projection of the probe in Figure 6 along the scan-depth plane

As an example of the conflicting design criteria consider lateral steering versus maximum focal depth. For a given array configuration, smaller element size increases steering ability whereas larger element size enables a longer lateral focal distance. As introduced, a single probe was required to operate from just sub-surface to the greatest range possible. As an example of the modelling, Figure 8 shows the specified TRL probe steering the beam at 60° and focusing to a depth of 40mm along the index - depth plane. The figure shows the desired ultrasonic beam with no grating or side lobes.

Table 1 gives the probe specification and Figure 9 shows the manufactured TRL probe.


Fig. 8. The sound field generated by the transmit array of the designed probe, steering to 60° and focusing to a depth of 40mm (SimulUS).

Table 1 The TRL probe specification

Frequency, 2.5MHz Total number of elements, 128
Wedge angle, 13° Number of elements along index axis, 16
Roof angle, 3.1° Index axis pitch, 2mm
Parameter g, 1.86mm Number of elements along scan axis, 4
58mm (L) x 40mm (W) x 55mm (H) Scan axis pitch, 3mm

Fig. 9. The TRL probe specified by the present study, manufactured by Vermon (France). Note the removable separator between the transmitting and receiving arrays, which was used to confirm that the probe performance was better without the separator

3 Performance trials

Two standard 2MHz longitudinal wave single element austenitic probes were selected for the performance comparison: 45° (WSY 45-2) and 60° (WSY 60-2) beam probes. These probes represent those commonly used in the inspection of nuclear components. Sector scan data on test block B from the TRL probe is used to illustrate the inspection performance in general and S/N performance of the TRL probe at corresponding beam angles to the single element probes is shown in Table 2, collected on test block A. Note that in all cases the target holes were set to 80% full screen height (FSH) and the S/N is defined as the ratio between the maximum noise amplitude from the material ahead of the target and the target signal (when at 80% FSH). The S/N is measured from the A-scans along the beam angles for both the single element probes and the TRL probe.

Table 2 S/N performance of the TRL probe on test block A generating beams at 45 and 60° in comparison to single element probes (refer to Figure 1).

SideScanHoleAverage S/N performance (dB)
TRL beam angle (°)Single element probe
4560WSY 45-2WSY 60-2
1 AS 4 13   14  
1 AS 3 10   7.7  
1 FS 4 >20   18  
1 FS 3 19.4 15.3 7.5 9
1 FS 2 7.9 6 7.1 7
1 AS 2   13   4.6
1 FS 1   <3   6.2
2 AS 8 11.6 17 15 11.3
2 AS 5 2.5   6.3  
2 AS 6   3.3   6.5

The cases presented in Table 2 have been chosen to illustrate the relative S/N performance of the TRL probe in comparison to dedicated single element probes generating beams at 45 and 60°. For the detection of hole 4 through the weld (AS) the S/N of the TRL probe was 13dB and the single element probe was 14dB, illustrating comparable performances when illuminating targets near surface. However when looking at the deeper hole 3 the single element S/N is 2.3dB weaker than the TRL probe. As would be expected, their performance when targeting hole 4 through the ferritic parent (and cladding) is greater than 18dB.

Note that the S/N performance of the TRL when detecting hole 2 (51mm deep) through the ferritic parent at a beam angle of 60° is 6dB (as opposed to 7dB with the single element probe) which is on the limit of the minimum S/N performance recommended by TWI; when detecting the deepest hole 1 at 68mm the S/N performance of the TRL is unacceptably below 3dB. However, when detecting hole 2 through the weld the TRL S/N is 13dB, while that of the 60° single element probe is 4.6dB, illustrating the advantages of focusing the sound field onto the target.

3.1 Inspection of SDHs along the buttering layer (side 1) in AS

Inspection of the SDHs 1-4 in AS requires the sound to travel through the austenitic weld leading to distortion of the beam. Figure 10 shows the detection of hole 4 in test block B. The near surface performance of the TRL probe is good, as evident from the lack of noise before the target. By separating transmit and receive paths through the medium the noise induced by the coarse grained microstructure is reduced; focusing also increases the signal-to-noise performance of the detection. The detection of hole 4 by the 45° beam single crystal probe led to an average signal-to-noise measured at 14dB, while the TRL probe performed at 13dB (see Table 1).

The data presented in Figures 10 to 15 show the sector scan from the TRL probe. The weld overlay and target positions are dimensionally accurate and the scales are in mm. The colour scale for signal amplitude is from white (low) through blue, green and red (high).


Fig. 10. Detection of hole 4 (side 1) along the 45° beam by the TRL probe in AS (test block B)

Figure 11 shows the sector scan of SDHs 2-4 with the beam focused at hole 3 in test block B in AS. The path of the sound beam is through the weld material where beam forming also takes place. Note the positioning error of hole 4, which is about 4 and 3mm along the depth and index axes, respectively. Figure 12 shows the increase in beam width at larger depths (hole 2 in test block B); the beam widths generated by the TRL probe at large depths are consistent with predictions by SimulUS. It can also be seen that the signal-to-noise is still acceptable (greater than 14dB).


Fig. 11. Detection of hole 3 (side 1) along the 45° beam by the TRL probe in AS (test block B)


Fig. 12. Detection of hole 2 (side 1) along the 45° beam by the TRL probe in AS (test block B)


3.2 Inspection of SDHs along the weld to parent stainless interface (side 2) in AS

Figure 13 shows the sector scan with the beam focused at hole 8 in test block B. The focused beam is severely distorted and/or split due to the presence of the fusion interface. However, Figure 14 show that as long as the beam does not traverse the austenitic weld, the other SDHs along the fusion face (holes 6 and 5 in test block B) can be detected with good signal to noise. This is reflected in the corresponding tests using the single crystals WSY 45-2.


Fig. 13. Detection of hole 8 (side 2) along the 45° beam by the TRL probe in AS (test block B)


Fig. 14. Detection of hole 6 (side 2) along the 45° beam by the TRL probe in AS (test block B)

3.3 General discussion

Consider the data shown in Section 3.1 which show errors in the horizontal positioning of the vertical SDHs on side 1 with respect to each other - this is most pronounced in Figure 11. This is likely due to the distortive effects of propagating through the anisotropic weld, as inspection through non-weld material correctly positions the target. Errors in excess of 5mm have been noted in different probe positions, with the velocity of the material changing depending on the angle of sound beam propagation. In contrast consider data in Figure 15 which shows the detection of the SDHs on side 1 of test block A in FS where the sound does not propagate through the austenitic weld, only the clad layer. The SDHs are vertically aligned with little error in horizontal positioning. This is a graphic illustration of the distortive effects of the austenitic weld leading to errors in positioning and sizing of discontinuities. Note however that the measured errors in the relative vertical positioning of the SDHs when the sound propagates through the weld material were measured to be in the region of 2mm, which could reflect upon the process of weld deposition (downhand Manual Metal Arc).


Fig. 15. Detection of hole 3 (side 1) along the 45° beam by the TRL probe in FS (test block A)

One of the motivations in the design of the present TRL probe was the improvement of the signal-to-noise performance in austenitic welds. Conventional twin crystal probes have been shown to improve signal-to-noise performance for inspection of austenitic welds. These probes perform very well but are limited to specific beam angles and operating ranges. Further they are rarely constructed with focused crystals or lenses due to cost implications; furthermore, if they are focused the inspection range becomes very limited. Phased array focusing and steering capabilities enable a single probe to provide the functionality of several probes. There are a number of different styles of phased array TRL probes,[10] the early designs were two linear phased array probes with a wedge separating the transmitting and receiving sides. These were limited to focusing in a single plane only (along the index axis) which limited the advantages of focusing by making the phasing capability along the scan axis redundant. To take full advantage of phased array capabilities, the present design of a 2D array allows the potential to focus along both the index and scan axes. The range of effective operation within the material is limited by the roof angle and the crossing point of the beams along the scan axis. As described earlier, a decision was made to limit the separator so that it did not reach the front of the probe and to use water for coupling, which proved to be a beneficial design and concept change. There was concern that the transmitting and receiving beams would cross talk and spoil any advantages gained by focusing. This concern proved to be false with the probe performing better than expectations. The focusing was shown to be effective from a depth of 10mm (sub-clad) to about 50mm within the component. A further advantage due to the use of immersion coupling was the better accommodation of surface profile variations which are inherent to large austenitic welds in industry.

The present design of the phased array TRL probe was made possible due to the availability of a 128 parallel channel instrument - the MicroPulse 5PA. The instrumentation and the TRL probe itself are a relatively expensive solution in comparison to conventional methods but the result is a replacement of at least six conventional twin crystal probes leading to advantages in the nuclear environment such as the reduction in potential material contamination of the reactor pressure vessel. Since the TRL probe can be used to generate sector scans, several flaw sizing advantages can be achieved:

  1. No tilt error due to the position of the probe with respect to surface variations.
  2. No coupling errors.
  3. The ultrasonic beam travels through a more constant grain structure. If a single crystal probe of one beam angle is moved in conventional scanning the beam is travelling through different material for the top and bottom of a flaw. With sector scanning there is material change but it is greatly reduced. Differing material attenuates and skews the beam to different degrees. The project results show that the through-wall errors between adjacent holes are small which should lead to small flaw sizing errors.
  4. Sizing errors can be reduced if the exact material structure is known in the component but this is rarely the case. But by minimising the ultrasound path variations and by focusing, both the signal-to-noise performance and the through-wall sizing performance should be improved.


  1. In conclusion, the design effort was successful in providing a phased array TRL probe with a large focused inspection range and good signal-to-noise performance in a typical austenitic weld.
  2. The use of an immersion twin array probe proved to be successful, providing a solution to reducing coupling errors.
  3. The probe achieved very good performance for detection of sub-clad targets.
  4. Phased array TRL probes are very complex. It is concluded that modelling is both essential and cost-effective to achieve desired performance.

5 Acknowledgements

This paper results from work carried out on the CRAFT Framework VI project RIMINI (COOP-CT-2004-512984), where the aim was to develop NDT methods to decrease in-service inspection times on critical components in nuclear power plants.


  1. J A Ogilvy, 'The influence of austenitic weld geometry and manufacture on ultrasonic inspection of welded joints', British Journal of NDT, pp 147-156, 1987.
  2. A Juva and J Lenkkeri, 'The effect of anisotropy on the propagation of ultrasonic waves in austenitic stainless steel', Proc of Reliability of the Ultrasonic Inspection of Austenitic Materials, Belgium, 1980.
  3. G Maes, P Hansoul and P Dombret, 'The PISC parametric study on the effect of cast austenitic steel macrostructure on the capability of ultrasonic examination', Proc. of the 12th Int. Conf. on NDE in the nuclearand pressure vessel industries, p 197, 1994.
  4. R Schmid, 'Ultrasonic testing of austenitic and dissimilar metal welds',, Vol 2, No 12, 1997.
  5. J A Ogilvy, 'Ultrasonic beam profiles and beam propagation in an austenitic weld using a theoretical ray tracing model', Ultrasonics, Vol 24, No 11, pp 337-347, 1986.
  6. J A Ogilvy, 'The influence of austenitic weld geometry and manufacture on ultrasonic inspection of welded joints', British J. of NDT, pp 147-156, May 1987.
  7. J A Ogilvy, 'On the use of focused beams in austenitic welds', British J. of NDT, pp 238-246, July 1987.
  8. Y Kurozumi, 'Development of an ultrasonic inspection technique for cast stainless steel', Insight, Vol 44, No 7, pp 437-442, 2002.
  9. J Krautkrämer and H Krautkrämer, 'Ultrasonic testing of materials', Springer-Verlag, New York, ISBN 0-387-51231-4, 1990.
  10. M Delaide, G Maes and D Verspeelt, 'Design and application of low-frequency twin side-by-side phased array transducers for improved UT capability on cast stainless steel components',, Vol 5, No 10,2000.
  11. S Mahaut, J-L Godefroit, O Roy and G Cattiaux, 'Application of phased array techniques to coarse grain components inspection', Ultrasonics Vol 42, pp 791-796, 2004.
  12. M T Anderson, S E Cumblidge and S R Doctor, 'Low frequency phased array techniques for crack detection in cast austenitic piping welds: A feasibility study', Materials Evaluation, pp 55-61, January 2007.
  13. A Whittle, 'Preliminary steps to validate a beam model for ultrasonic phased arrays', Insight, Vol 48, No 4, 2006.
  14. R/D Tech, 'Introduction to phased array ultrasonic technology applications', R/D Tech Inc., Canada, ISBN 0-9735933-0-X, 2004.

For more information please email: