The e-Journal of Nondestructive Testing - ISSN 1435-4934 - www.ndt.

net

TFMiTM: Using Intermodal Analysis to Improve TFM Imaging

More info about this article: http://www.ndt.net/?id=26036

Paul Holloway1 and Ed Ginzel2

1.
Holloway NDT & Engineering Inc, Georgetown, Ontario, Canada
e-mail: [email protected], website www.hollowayndt.com
2
Materials Research Institute, Waterloo, Ontario, Canada
e-mail: [email protected]

2021.05.03

Abstract
In recent years, phased array ultrasonic inspections have incorporated a variety of total focussing methods (TFM) to
process data collected from full matrix capture techniques. The results are images that can greatly assist in
characterization and sizing of indications. Codes have been written to standardise the techniques [1,2,3]. Because
these codes use a qualification process, they specify that the selection of propagation modes used in the TFM
construction are identified. Experience has demonstrated that not all combinations of paths are effective at detecting
indications in TFM. The user is left to determine which propagation mode produces the best result. In this paper we
illustrate how the process can be improved in both detection and reliability by simultaneously processing multiple
modes in combination.

Keywords: ultrasonic, phased array, FMC, TFM, total focussing method

1. Introduction

Total Focussing Method (TFM) is a term given to the mathematical processing of waveforms
collected by a phased array data acquisition process to provide an image of a volume under
inspection.

Acquisition techniques can be one of several options:

 Full Matrix Capture (FMC)

 Half Matrix Capture (HMC)
 Sparse Matrix Capture
 Plane Wave Imaging (PWI)
 Synthetic Aperture Focussing Technique (SAFT)

The most common approach is FMC, where each element of a phased array probe is used as an
independent transmitter and receiver. An illustration of the transmitting and receiving operations
using a small 4-element transducer is shown in Figure 1. The echo from each transmitted pulse is

Page 1 of 13
recorded as a separate A-scan for each receiving element. As a result, the simple 4-element FMC
dataset would result in 16 A-scans (blue arrows).

Figure 1: Simplified 4-element full matrix capture

For a transducer of n elements, it is possible to capture up to a maximum of n2 A-scans with a full

matrix capture approach. For a 32-element transducer, an FMC dataset would produce (32)2 or
1,024 A-scans, and a 64-element transducer would produce 4,096 A-scans. These figures represent
the number of A-scans per scan frame, or scan position. A small weld sample 300mm long
scanned in 1mm scan increments using 64-elements would require over 1.2 million A-scans. Raw
FMC datasets can become incredibly large and do not present useful images alone without further
processing. Capture techniques such as sparse matrix capture (SMC) and half matrix capture
(HMC) use fewer elements or store fewer A-scans which help in acquisition speed. However,
FMC datasets provide the most data and possibilities for imaging algorithms.

Having acquired the waveform data with one of the acquisition techniques, there are several ways
to process the data to obtain an image that can be useful. The TFM algorithm uses a process of
delay and summing of waveforms. Each received waveform can be calculated for its point of
origin in the region of interest (ROI), and signals from boundaries in the same location will add as
coherent sources. This results in larger amplitude summations where flaws or other discontinuities
are located. When plotted, signals are reconstructed as points of higher amplitude when there is
constructive interference from echoes received at each element. TFM can be applied as a
processing option regardless of the type of acquisition (FMC, HMC, PWI, etc.). TFM can be used
to synthetically generate a focus everywhere in the ROI by applying different virtual focal laws to
the collected A-Scans.

The same set of FMC data can be used multiple times using different combinations of
reconstruction parameters. Critical in the success of the TFM is the selection of the propagation

Page 2 of 13
mode that could occur for each pixel in the ROI. Flaws can be imaged from multiple directions
when boundary reflections are included in the path from transmitter to receiver. This applies to
both reflected and diffracted signals. Coherent signals in a pixel region arriving from multiple
modes and directions will improve results by increased reliability of detecting flaws and reduced
occurrence of artifacts [4,5].

The following table is an overview of the primary propagation modes used with TFM for
inspection of carbon steel welds using a refracting wedge:

Table 1: TFM Propagation Modes for General Weld Inspection

Imaging Path Nomenclature Description

Two paths:
T-T (2T)
1. Transducer to discontinuity
2. Discontinuity back to transducer

Three paths:

1. Transducer to backwall
TTT (3T) 2. Backwall to discontinuity
3. Discontinuity back to transducer

(or reverse, i.e., to discontinuity first, then backwall)

Four paths:

1. Transducer to backwall
TT-TT (4T) 2. Backwall to discontinuity
3. Discontinuity to backwall
4. Backwall to transducer

Five paths:

1. Transducer to backwall
2. Backwall to discontinuity
TTTTT (5T)
3. Discontinuity to front wall
4. Front wall to backwall
5. Backwall to transducer

(or any other combination of 5 paths)

ISO [3] guidance states that in general, planar flaws are best detected when the imaging paths have
an incident angle and reflected angle on the flaw that is:

a) (about) perpendicular to the discontinuity orientation;

b) (about) symmetric to the normal direction of the discontinuity, or

c) according to Snell’s law if mode-conversion occurs at the discontinuity.

Page 3 of 13
The combination of paths and modes used to image a volume in an ROI becomes a challenge for
the operator qualifying a procedure using TFM. The selection of the right mode of propagation
often requires some foresight into the nature of the flaw under investigation (i.e. “we need to know
what we’re looking for before we start looking for it”). If the wrong mode is selected, the flaw
may be completely overlooked (porosity detected with PA sectorial and TFM TT mode, but missed
with TFM LTL mode, Figure 2 to Figure 4). Even if the recommended mode is selected (e.g., 3T
for a root connected crack), the image may fail to accurately represent the flaw if the reflecting
faces are tilted.

Figure 2: PAUT sectorial Figure 3: TFM TT mode Figure 4: TFM LTL-mode

In this paper we illustrate a new approach called TFMiTM that combines multiple, standard TFM
propagation mode images via multiplication of pixel values along with the introduction of a non-
linear amplitude factor. The multiplication approach not only greatly enhances signal-to-noise
ratio (SNR), but also improves reliability of detection and flaw characterization.

2. Basis of Intermodal Analysis

The basic TFMiTM image is generated by combining the images of a selected number of standard
TFM propagation modes by multiplication. Amplitude values at each pixel co-ordinate are a
product of the corresponding pixel values of each component mode. Coherent signals are
amplified, thus high amplitude responses in two or more modes are enhanced. Similarly,
incoherent signals are suppressed, which greatly reduces background noise. Any combination of
modes may be used in TFMiTM. This paper focuses on the results from the basic 2T, 3T, 4T and
5T combinations.

As an example, a TFMiTM image of a porosity cluster using a combination of the 2T, 3T, 4T and
5T modes would be generated as follows:

Page 4 of 13
 Figure 5: TFMiTM basic image, porosity sample (2T x 3T x 4T x 5T)

As an additional step, a non-linear amplification filter may be applied to reveal low amplitude
details while retaining the low noise floor. A comparison between a final TFMiTM image and an
image generated with simple intermodal addition is shown in Figure 6 and Figure 7. This
additional filtering process may not always be required.

Figure 6: Intermode addition Figure 7: TFMiTM image after application of a non-

linear amplification filter

3. Test Programme

To demonstrate the performance of TFMiTM on real flaws, a basic FMC was configured on several
flawed specimen samples fabricated by reputable manufacturers used in the NDT industry, plus
one additional specimen with an artificial 3-segment crack.

Page 5 of 13
All testing was performed with transducers mated to PAUT shear wave wedges to simulate field
conditions during weld inspection.

Instrumentation used was:

 Sonatest Veo+ 32:128 outfitted with development software

 Olympus 5L32-A31 transducer (5MHz, 32-elements, 0.6mm pitch)
o Full 32-elements used in FMC acquisition
o Mated to a 55-degree shear wave wedge
 Vermon-NDT A10L64-12 transducer (10MHz, 64-elements, 0.3mm pitch)
o Tests performed using 32 and 64-elements in FMC acquisition
o Mated to a 55-degree shear wave wedge

3.1. Artificial crack

To test the capabilities on crack-like reflectors, a 25mm thick steel block was manufactured with
three crack features varying in complexity from 1 to 3-segments (Figure 8). The vertical height of
each segment was 5mm, making the maximum height of the 3-segment crack 15mm.

Figure 8: Machined 3-segment crack feature

The crack cross-sections are shown in the section views. Each section was scanned from both
directions (probe on left/right) with PA sectorial, 2T through 5T propagation modes, and TFMiTM.
The probes were positioned at an offset distance typical for a double-V weld to represent inspection
conditions (Figure 9).

Page 6 of 13
 Figure 9: 3-segment machined crack probe positions

Neither the PA sectorial nor any of the regular TFM modes provide much information on the
complex shape of the feature. The TFMiTM group images the crack clearly from both skews with
extremely high signal-to-noise ratio in comparison with the individual modes (Figure 10, Figure
11).

Figure 10: Skew 90, Section A-A

Page 7 of 13
 Figure 11: Skew 270, Section A-A

The images from the smaller sections B-B and C-C are shown in Figure 12. Depth cursors are
provided in each image to validate sizing measurements. The full height in Section A-A (15mm,
above) and Section C-C (5mm, below) are sized accurately from both sides. Section B-B (10mm,
below) from the Skew 90 orientation was undersized by approximately 2-3mm, likely due to the
wide milling bit used to form the crack and the diminished tip diffraction echo.

Skew 90 Skew 270

Page 8 of 13
 Figure 12: 3-segment machined crack, TFMiTM images on Sections B-B and C-C

3.2. Real Weld Flaws

The following shows the comparison between TFMiTM and standard TFM modes on a variety of
realistic weld flaws. The projected views shown below the TFM frames are useful for displaying
the length (horizontal axis) and height (vertical axis) of reflectors.

#1 - Lack of Sidewall Fusion TFMiTM (2T-5T) TT-TT (4T)

#2 - Toe Crack TFMiTM (2T-5T) TT-TTT (5T)

Page 9 of 13
 #3 - Root Crack TFMiTM (2T-5T) TTT (3T)

#4 - Centreline Crack TFMiTM (2T-5T) TTT (3T)

#5 - Porosity TFMiTM (2T-5T) TT-TT (4T)

Page 10 of 13
The table below summarizes the height sizing performed with TFMiTM compared to other
techniques:

Table 2: Flaw Height Size Summary

PAUT (tip echo) TOFD TFM (TFMiTM)

#1 Lack of fusion 3.3 mm 4.2 mm 3.6 mm (TT-TT) 3.8 mm
#2 Toe crack 6.7 mm 6.5 mm 6.3 mm (5T) 6.0 mm
#3 Root crack 4.0 mm 5.1 mm 5.5 mm (3T) 4.4 mm
#4 Centreline crack 4.0 mm N/A 4.0 mm (3T) 4.0 mm
#5 Porosity 4.5 mm N/A 4.8 mm (4T) 4.9 mm

4. Discussion

As noted in the introduction, TFMiTM has the potential to improve reliability of detection, enhance
the detail of the imaged flaw for sizing and reduce the occurrence of unwanted image artifacts. In
some cases, standard TFM may provide no evidence of a flaw (Figure 13 vs. Figure 14 below).

Figure 13: Centreline crack, TT-TT mode Figure 14: Centreline crack, TFMiTM

Note that some combinations of TFM produce artifacts from mode conversion (false positives).
Intermode combination with TFMiTM can be used to suppress these artifacts.

Multiplication of modes with TFMiTM greatly increases the signal-to-noise ratio (SNR) of the
image. All images in this paper use the same colour palette with the same 0-100% range. The
white background in the TFMiTM images represents pixel values of less than 1% range. SNR
comparisons between standard TFM, intermode analysis by addition, and TFMiTM are shown in
Figure 15 to Figure 17. Note the SNR values shown are a ratio of the maximum pixel value in the
indication area compared to the maximum pixel value in an area away from the indication.

Page 11 of 13
 Figure 15: TT mode Figure 16: Intermode addition Figure 17: TFMiTM
(SNR: 24.6) (SNR: 6.8) (SNR: 86.9)

5. Conclusions

TFMiTM is a new approach to the basic TFM options. It provides improved detection and sizing,
improved geometric fidelity and flaw characterization, and reduces the chance of false positives
caused by artifact images. TFMiTM uses multiple propagation modes simultaneously to generate
an image, and removes the guesswork required when limited to single modes.

Currently, TFMiTM is in development by Holloway NDT & Engineering Inc. and Sonatest. All
images in this paper were generated using prototype software and/or post processed externally.

The use of intermodal analysis does not eliminate the need for examining mode responses
individually. It is still recommended that procedures include use of PA sectorial and standard TFM
propagation modes suited to characteristics of a known reflector. However, TFMiTM can reduce
the number of iterations that might be used to optimise detection and can greatly enhance
characterization and sizing of relevant indications.

5.1. Future Development

Possibilities for future development include:

 Pattern recognition between modes and a staging process for characterization to optimize
mode combinations, which may include self comparisons, inclusion of an addition frame,
and combining images from opposite skews
 3D weld flaw modeling based on frame stacking
 Additional study on interaction with propagation modes using mode conversions (TTL,
TLL, etc.)

6. Acknowledgements

We would like to express our thanks to Jonathan Lesage to detailed explanations of the various
processes associated with the FMC/TFM techniques. We would also like to thank Philippe Rioux
(Sonatest) and Steve Hudgins (Advanced Sound Solutions) for providing many of the samples
used to illustrate TFMiTM in this paper.

Page 12 of 13
References

1. ASME Boiler and Pressure Vessel Code, Section V, American Society of Mechanical
Engineers, Two Park Avenue • New York, NY • 10016 USA, 2019
2. ISO 23864, Non‐destructive testing of welds — Ultrasonic testing — Use of automated
total focusing technique (TFM) and related technologies. International Standards
Organisation, Switzerland, 2021
3. ISO 23865, Non-destructive testing — Ultrasonic testing — General use of full matrix
capture / total focusing technique (FMC/TFM) and related technologies. International
Standards Organisation, Switzerland, 2021
4. Busse, L J, Collins, H D, and Doctor, S R. Review and discussion of the development of
synthetic aperture focusing technique for ultrasonic testing (SAFT-UT). United States: N.
p., 1984. Web. doi:10.2172/6977775.
5. Sy, K., Bredif, P., Iakovleva, E., Roy, O., & Lesselier, D. (2018). Development of
methods for the analysis of multi-mode TFM images. Journal of Physics: Conference
Series, 1017(1). https://doi.org/10.1088/1742-6596/1017/1/012005
6. N. Portzgen, D. Gisolf, D. Verschuur, Wave equation-based imaging of mode converted
waves in ultrasonic NDI, with suppressed leakage from nonmode converted waves,
Ultrasonics, Ferroelectrics, and Frequency Control, IEEE Transactions (2008)
7. E. Iakovleva, S. Chatillon, P. Bredif, S. Mahaut, Multi-mode TFM imaging with artifacts
filtering using CIVA UT forwards models, AIP Conference Proceedings, Vol. 1581, 2014

Page 13 of 13