NDT.net, Germany, Joined Nov 1998, 617
Smallest detectable flaw size, especially for low frequency test (f= 0.5 MHz) and in general.
For an applied frequency of 0.5 MHz what will be the smallest flaw
that air coupled ultrasound could find?
Furthermore I want to enter into a general discussion about smallest detectable flaw sizes.
Is there a certain relation between flaw size and frequency?
- lambda/flaw dia ratio?
- Is this relation linear?
Preconditions: ideal reflection (shape and impedance).
|Bill Grandia |
Re: Smallest detectable flaw size, especially for low frequency test (f= 0.5 MHz) and in general. Yes, we estimate the minimal detectable flaw size to be in the order of the size of the wavelength.
The focal spot diameter in air for a 25mm diameter element with a focal distance of 35 mm is 1.2 mm (-3dB points) at 400 kHz, with a focal depth of 25 mm (-3dB). In through-transmission with the focal spots coinciding with each surface, the lateral resolution therefore is about 1.2 mm.
|Robert A. Day |
Milky Way Jewels, USA, Joined Nov 1998, 40
Re: Smallest detectable flaw size, especially for low frequency test (f= 0.5 MHz) and in general. Rolf -
I have to disagree with Bill Grandia's estimate of
minimum detectable flaw size because we rarely detect
flaws in air. Usually in the material under test. Using
the diffraction limit of the transducer [at best Lambda
the wavelength] is not a bad approximation to the
longest dimension of the reflector that may be detected
the other dimensions maybe substantially smaller.
An example would be detection of 0.003 inch notch in
steel tubing at 15 MHz. The wavelength in steel is 0.016
inch so we are detecting less than a fifth of a
wavelength. The notch is in this case 0.25 inch long so
for all practical purposes it is infinite. Since this
is an angle beam detection the diffraction limit tells
us little about the beam because of refraction caused
astigmatism but clearly we are in one dimension doing
much better than a wavelength.
We pay a heavy penalty however to detect at these levels
because we use a focused probe that has a beam width in
steel of about 0.03 inches. To achieve 20% beam overlap
we would need to scan at a pitch of 0.024 inch. The
speed of the scan would have to be really fast to get
reasonable throughput. In the example above the scanning
was done to get 3 hits on a 0.25 notch so the pitch was
about 0.08 inches. It was still a relatively slow test.
Actual flaw detection levels are very much a function of
the test parameters, frequency, focus, diameter, and
geometry. Many things can be done to improve detection.
In the steel tubing example we could go to aspherical
transducers that would achieve a better focus in the
steel and presumably detect smaller flaws. Actually in
this example signal-to-noise was still very good at
0.003 so improvement wasn't needed. Since flaw detection
concerns often happen when doing curved surfaces the use
of asphericals may be valuable.
In a low velocity material, like one we would do air
couple sound on, would have a velocity of 2.5
mm/microsec. So at 500 kH we would have a wavelength of
5 mm. We certainly would expect to be able to see
something this size in homogeneous material, like steel. The only way to significantly improve on this is to use near field imaging which only works if the material is very thin, not much thicker than the resolution we want to get. This isn't practical for most NDT situations but was demonstrated to a 200th of a wavelength by Eric Ashe many years ago. This principle is the basis of
Scanning Tunneling Microscope for which Bennig and Rohre
got the Nobel.
I haven't even touched on the probability of detection
issue which has to do with the random errors that creep
into any measurement process. Or more precisely to the
definition of what do you mean by detection. Assuming we
want some probability and confidence that a flaw is
detect, what numbers do we use and is it possible to
obtain that performance from a given ultrasonic test.
On possible helper in all this is computer modeling.
Most users do not currently have access to this type of
capability but I expect that will change soon. The Ames
Center for NDE has a program that allows approximate
modeling to be done on complex shapes and provides
reasonable prediction of time of arrival and amplitude
of flaw signals along with the other signals to be
expected. This is a great aid to analyzing some of the
problems associated with how small a flaw can I see?
You should visit their web sight at
http://www.cnde.iastate.edu/ to get more information.
Better modeling tools are in the works and will help
answer complex questions better than we can answer them
|W. Grandia |
Re: Smallest detectable flaw size, especially for low frequency test (f= 0.5 MHz) and in general. Dear Mr. Day,
I have to disagree with your reply . We certainly do not inspect flaws in air. We are using air as a couplant medium, and tests are usually performed in through-transmission mode.
We related Rolfs question to the lateral resolution - which should be in agreement with the Rayleigh criterium - and we were considering either rectangular or circular flaw shapes. Your example of an infinitely long notch deviates considerably from our considerations. Small lines, numerous tiny flaws are known to generate different effects, which are outside the scope of Rolfs question.
|Rolf D. |
NDT.net, Germany, Joined Nov 1998, 617
Re: Smallest detectable flaw size, especially for low frequency test (f= 0.5 MHz) and in general.
thanks for your answer with so much details.
With the attached picture I want to make
my questions more clear.
Rocky explained a very complicate subject,
I didn't wanted to release a so detailed discussion.
However If you like you could continue and try to keep
the general aspect on sound propagation for this workshop theme.
Further I could past this messages to a future workshop
on flaw detection, this could be a valuable start.
Besides the air coupling capability I wanted to get a more
detailed information about the smallest flaw size
for a simple shaped defect (FBH). Especially the area where the
curve starts to drop and further how the curve continue.
I can imagine that the curve drops within x dB per wavelength.
What will be x?
I remember from my time I did really ultrasonic (no virtual like now)
that we recognized conclusions in plastic pipes mostly in a range
of 1/2 to 1 wavelength, so this confirms what Bill stated.
Refer also to the attached article where transducer selection
for a given task is discussed.
Thanks again for your discussion.
|Bill Grandia |
Re: Smallest detectable flaw size, especially for low frequency test (f= 0.5 MHz) and in general. Dear Rolf,
Thanks for your clarifying remarks about the subject of flaw size
detection. For those interested to continue this route of discussion, I
have added my detailed response to Robert Day's remarks in the following
Thanks to Robert and Rolf for making such discussions possible.
Regards, Bill Grandia
REMARKS ON MINIMAL FLAW SIZE DETECTABILITY, ESPECIALLY FOR AIR-COUPLED ULTRASONICS
(In response to the discussion of Robert A. Day of September 11, 1996)
From all the reactions received, I am becoming aware that the principles of non-contact (Airscan) ultrasonics are not quite
understood by persons experienced in conventional fluid coupled ultrasonics.
In the reply from Robert, it states that at a sound velocity of 2.5 mm/msec at 500 kHz the wavelength is 5 mm (this would apply
for something like graphite epoxy material). This does NOT imply that the smallest detectable flaw size would be in the order of
Here is a misconception, which will be further explained in this writing.
Prior to the explanation, I would like to react on the quoted "analog sample" of a pulse-echo inspection at 15 MHz, to detect a
0.003" deep notch with a length of 1/4 inch.
This is an extreme example, which should not be used to compare with the method of a through-transmitted non-contact
ultrasonic inspection for the following reasons.
The detection of a small notch (0.003") requires a pulse-echo system, using either a spherical, cylindrical, or better yet, a
compound focused transducer. When such a transducer is used, let us say a wide-band 15 MHz, F#4 (which means that the
focal length is 4 x the element diameter, which is the 1 inch), the diameter of the focal waist area, based on a -3 dB drop, is 0.10
mm, while the depth of field becomes 11.5 mm. These dimensions are of course for water.
Of course, inside the steel, these dimensions change as a function of the difference of the L-wave sound velocity in water and
the propagation speed of shear waves in steel. Also the shape of the focal spot becomes somewhat distorted or flattened in
favor of the detectability for the shallow notch detection.
Now, to detect a 0.003" notch, we need a high frequency contents in the transmitting signal. Most commercially available
ultrasonic instruments feature a transmitting pulse rise-time of approximately 10 nanoseconds. The fastest pulse width
recovery is usually larger then 10 nanoseconds. So, we may reason that the fastest pulse width available is around 20
nanoseconds, representing a transmitter pulse frequency of 25 MHz. Of course, the water and steel are acting like a low-pass
filter and a considerable amount of the high frequency contents becomes attenuated.
Since we need to detect the 0.003" deep notch, the use of a narrow band 25 MHz transducer is less desirable. We may select a
wide-band 15 MHz center frequency transducer, having enough bandwidth to reach into the higher frequency regime. For this
reason, the bandpass of the receiving amplifier should be set for high pass filtering above lets say 10 MHz to take advantage
of the high frequency resolving power (Not all ultrasonic instruments have such a feature).
At the same time, low pass filtering under about 30 MHz is desirable to limit the overall bandpass as would be beneficial to
suppress the noise floor. Again, most instruments do not have this option.
From this discussion, it becomes apparent that the 0.003" notch becomes detectable, especially since the notch width is much
larger than focal waist width.
I would like to emphasize again that we are discussing a high frequency wide band condition.
For this reason, it is not quite fair to use the above discussed technique as a comparison to a narrow band 400 or 500 kHz
Airscan through-transmission method.
Despite the fact that Airscan uses a narrow band approach, the technique is very much capable of detecting small flaws ( in
the order of 1 mm).
The wave length in air at 400 kHz is 0,85 mm.
The focal waist diameter for a 1 inch (25 mm) diameter 400 kHz transducer, having a focal distance in air of 1.5 inch or 38 mm
equal to 1.2 mm with a depth of field equal to 25 mm in air.
According to Robert, the minimum detectable flaw size is about 5 mm because of his reasoning that the wavelength in the
material is about 5 mm, which is the misconception.
First of all, in our Workshop paper we have included an Airscan C-scan image (fig.7), where we show that we can definitely
detect a missing bond of a single wall in a honey-comb core structure. In this case we are detecting a flaw with a length
dimension of approximately 1 mm. Looking at the picture must make every body a believer. All we need now is to explain why
such a fine lateral resolution is possible.
For simplicity sake, lets discuss the focused non-contact inspection on a flat Graphite/Epoxy laminate.
The transmitting transducer is usually aligned perpendicularly to the surface of the part, such that the focal spot coincides
with the entrance interface "air to laminate". Opposed is a similar receiving transducer located at the sound-exit side.
After the transmitting sound beam (having a narrow focal width) hits the surface, there will be sound propagating in many
different wave modes in various directions within the test part. Wave modes are possible like longitudinal- , shear-, bulk- and
guided plate waves. Each mode will have different propagation speeds. However the longitudinal wave mode is the fastest
one. This means that at the exit side of the panel the L-wave component will be arriving first. In case of a simple laminate, this
is the mode we are interested in (this is not the case for honey-combs later explained)
The shortest travel occurring through a virtual cylindrical sound port, having a diameter of the focal spot waist width can be
detected and gated in its appropriate time domain. The amplitude of the L-wave component within the gate is monitored and
further processed via a data acquisition/ imaging routine.
Any flaw-like sound obstruction such as small voids, distributed porous contents, delaminations, resin starved conditions,
etc., can now be detected. For this reason the minimum detectable flaw size is in the order of the focal spot diameter (in air), or
even smaller if we take a sound drop of - 6 dB into account.
It was also suggested by Robert to check with the Ames Center of NDE in regards to the program that allows approximate
modeling to be done in complex shapes and providing a reasonable prediction of arrival and amplitude of flaw signals along
with the other signals to be expected. Unfortunately, Dr. Chimenty fromAmes has not contacted us for exchange of
information, and therefor my knowledge about their progress is only limited to a few published papers. However our group at
QMI has been working successfully with Airscan in excess of 10 years. During this period of time, we and a large group of
Aerospace customers have been using our technology with great success. Especially strong progress has been made since
the recent years after our instrument model "SONDA-007" was released.
Judging from honey-comb C-Scan (fig 7 of our workshop paper), the high lateral resolution capability can not be denied.
I also would like to present some inside into the evaluation technique for the inspection of organic honey comb materials
(such as solar panels) and even metallic honey comb structures.
When we align a set of transducers in a perpendicular manner with the honey comb sheet, we cannot expect that a
longitudinal wave travels through a thin core wall, having a wall- thickness much smaller then the wavelength. In this case, the
wave mode through the core wall is an Ao (antisymmetrical) wave, which has a slow phase velocity and therefore a longer
wave length. (pretty much like the ondulating motion in a rope when one end is fastened to the door knob, while the other end
is moved by hand with a brief up-and-down motion.) This wave type experiences a low signal attenuation. On the surface of
the part, the L -wave at the focal spot becomes refracted and a mode conversion occurs of which the Ao component is the one
that propagates towards the exit side. At this point the reverse mode conversion occurs. Using this technique, it is very
necessary to locate the monitoring gate such that the through transmitted at a location very closely to the core wall is
captured. We do not wish to get a through transmitted signal in the gate from a sound beam hittingthe center of the core
(drum /membrane effect), since as a consequence the sound will travel through all the adjacent core walls, hereby impairing the
detectability of a single core wall.
It is obvious that the received signal amplitudes are extremely small. To overcome this problem, we use a strong tone-burst
transmitting approach, while the receiver is preceded by a remote super-low-noise preamplifier. Good signal to noise ratios are
obtained in real time without the necessity of digital or real time averaging or without the use of pulse correlation or other
means of noise reduction.
It is also possible to inspect solid panels by using a single sided plate wave mode, hereby aligning the tranducers under a
critical angle to allow an Ao guided plate wave to travel through the material. The value of the transducer entrance angle is
related to the wall thickness and phase velocity as can be seen from a dispersion diagram. Or instead theangles can be simply
aligned to offer the maximum signal amplitude.
We have detected a delamination with a diameter of 5/16 inch in a PEEK material plate having the transducer focal spots at a
distance span of about 12 inches, hereby gating for the very first "signal-component-only" to arrive and captured within the
Hopefully, this writing will take away some of the confusions.
|Yosi Bar-Cohen |
R & D,
Jet Propulsion Lab (JPL), USA, Joined Nov 1998, 26
Re: Smallest detectable flaw size, especially for low frequency test (f= 0.5 MHz) and in general. Ultrasonic methods of detecting flaws are using the perturbation
effect of flaws on the propagating ultrasonic waves. The
delectability of the wave depends on the flaw characteristics,
i.e., size and the material content.
Material content - This characteristics is related to inclusion
flaws. A simplistic rule suggests that the larger the acoustic
impedance mismatch with the host material the greater the flaw
Flaw size - Flaws will scatter ultrasounds that are impinging on
them. A flaw that is larger than the wavelength will form a
backscattering directivety, i.e., reflection. Flaw size equal
or greater than the wavelength will cause a reflected amplitude
that is equal to the impinging wave amplitude. When the size of
the flaw is smaller than the wavelength, the scattered wave
amplitude drops exponentially as a function of the
flaw-size/wavelength ratio. As a general rule: the minimum
detectable flaw size is considered the value of half the
As an example, for a steel part that is tested at 0.5 MHz, the
velocity is about 6x10^6 mm/sec and the wavelength is
12 mm. Therefore, the smallest detectable flaw is 6 mm.
Overall this is only a guiding criteria. The actual detectable
flaw depends on numerous factors that include instrumentation
capability and inspector attentiveness & qualification. This
factors are used to determine the probability of detection (POD)
of the flaw.