at a time, and the focusing is done in post-processing both in
transmissions and in receptions (two-way synthetic focusing)
(Flaherty et al. 1967). This means that (a) the SAF hardware
can be much simpler since only a few D/A output channels
are required (b) the SAF imaging speed can be increased
by limiting the output channels and (c) the SAF focusing is
achieved in both transmission and reflection, leading to better
resolution in a large inspection area.
It should be also mentioned that modern PA technology
is utilizing elements of SAF in the ability to focus in reception
through time backpropagation delay laws.
The objective of this paper is to present an experimental
prototype system for 3D imaging of internal rail flaws using
ultrasonic SAF techniques. An improved SAF beamforming
scheme is proposed based on sparse subarray firing to provide
high-contrast images in quasi real time (Huang and Lanza di
Scalea 2022). A sophisticated post-processing routine is devel-
oped to enable automatic rail flaw quantification without the
user’s judgement. The prototype’s hardware is packaged in a
battery-powered storage case for portability and ruggedness.
Validation tests were performed on a number of flawed rail
sections from the FRA rail defect library managed by MxV Rail
(formerly TTCI). The flaw images generated by the imaging
prototype showed a good match compared to the ground
truth established from rail break tests, especially in the case of
natural transverse-type defects.
The SAF Imaging Prototype
A portable imaging prototype was designed, assembled, and
tested to enable handheld ultrasound imaging of rail flaws
based on an enhanced SAF technique. As shown in Figure 3a,
the hardware components of the imaging prototype were
a multiplexer, a 12 V battery, a host computer, and a probe
comprised of a transducer array, a wedge, and an encoder
wheel. All the hardware components were screw fixed inside a
carry-on size storage case. The multiplexer (a high-speed data
acquisition system) that allowed multichannel data acquisi-
tion controlled the pulsed emission and reception to/from the
array. A 12 V battery was used to support the multiplexer for up
to 8 h of autonomous operation. The probe was composed of
a transducer array, a shear wedge, and an encoder, as shown
in Figure 3b. The transducer was a 64-element longitudinal (L)
wave linear array with a central frequency at 2.25 MHz. The
array was attached to a 55-degree wedge to generate direc-
tional shear (S) waves in the rail steel. The encoder recorded
ME
|
RAILROADS
Figure 1. Examples: (a) FRA
safety statistics data for all
track, roadbed, and structures
(2018–2022) (b) detail
fracture (DF) (c) transverse
fissure (TF) and (d) vertical
split head (VSH).
DF TF
VSH
All channels in transmission
with applied time delays
(physical focusing in transmission)
Complicated and expensive hardware
Slow imaging through focused scans
Selected channels in transmission
(synthetic focusing in both
transmission and reception)
Simpler hardware
Fast imaging possible through subarrays
PA SAF
Figure 2. Ultrasound imaging technology: (a) conventional phased arrays vs.
(b) synthetic aperture focusing.
52
M A T E R I A L S E V A L U A T I O N • J A N U A R Y 2 0 2 4
2401 ME January.indd 52 12/20/23 8:01 AM

the transverse position of the probe when scanned on the
rail surface, with a resolution of 16 counts/mm. The encoder
allowed the system to create 3D images from the individual
2D scans. The array was coupled to the wedge using conven-
tional ultrasonic gel couplant. The couplant was also applied
at the wedge/rail interface to compensate for the impedance
mismatch. A graphical user interface (GUI) platform was devel-
oped on a standard commercial laptop with a GPU available
for parallel computation. All steps of the signal processing algo-
rithms were programmed and automated in the GUI platform,
which enabled flexible configuration and result analysis for the
user’s convenience. As shown in Figure 3c, during testing the
user simply moves the probe on the surface of the rail section,
and 3D images of the scanned area are displayed in quasi real
time in the GUI.
Specific features of the image reconstruction algorithms
that were developed and implemented in the prototype are
discussed in detail in the following subsections.
Time Backpropagation Beamforming with a Transducer
Wedge
The time backpropagation algorithm (also known as
delay-and-sum or DAS algorithm) is widely used in SAF
imaging (Jensen et al. 2006). Dynamic focus is achieved both
in transmission and in reception by considering the ray path
connecting the transmitting transducer element, the focus
point, and the receiving transducer element. An image is
built by summing the backpropagated signals through all
transmitter-receiver pairs of the transducer array. Considering
transmitters i =1, 2,…, M and receivers j =1, 2,…, N, the DAS
beamformed SAF image is constructed as: ​ ​
(1)​ I(y, ​ z)​ =​ ∑
i=1​
M ​ ​ ∑
j=1​
N ​ ​ Ai​j​​​(​τ​ij,yz​​)​​​​
where
the time of flight (TOF) ​​ ​ ij,yz​​​ is the propagation time of the
ray path from the transmitter Ti(yi, zi) to the focus pixel
P(y, z) and back to the receiver Rj(yj, zj).
Ultrasonic array
(2.25 MHz, 64 elements)
Wedge
(55° shear wave)
Encoder
(16 counts/mm)
Case
Case
Battery
Multiplexer
Probe holder
(array+encoder)
Laptop computer
(MATLAB GUI)
Data
Data
Power
Laptop GUI
Handheld probe
Scan
direction
Figure 3. Portable imaging
prototype: (a) main
components (b) array-
wedge probe (c) prototype
during scanning of a rail
section in the laboratory.
J A N U A R Y 2 0 2 4 • M A T E R I A L S E V A L U A T I O N 53
2401 ME January.indd 53 12/20/23 8:01 AM