Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

14 Yan

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

The Open Access NDT Database

Xiang YAN 1, Jinchi ZHANG 2, Daniel KASS 1, Pierre LANGLOIS 2


1

Olympus NDT Inc, Waltham, MA, 02472. Phone:617-229-9314. Email:yanxiang2005@yahoo.com


Olympus NDT Inc, 505, boul. Du Parc-technologique, Quebec (Quebec) G1P 4S9, Canada. Email:
jinchi.zhang@olympusndt.com

Abstract
DGS curves for angle beam single element (UT) and phased array probes (PA) are calculated from semianalytical model using CIVA. The modelled DGS curves show improved accuracy in near field over the
universal DGS curves. This is validated by a set of measurements performed with angle beam probes.
Measurements were done with 2 and 4MHz integral wedge probes with Flat Bottom Hole reflectors
(FBH). The results show excellent agreement from half of the near field to about 5 times the near field
distances. The probes bandwidth is shown to influence the near field length and the attenuation
correction in the test sample. For a phased array probe having a broader bandwidth, the near field length
is slightly shorter than the mono crystal probe having a relatively narrower bandwidth. For the 4MHz
probes, the attenuation correction in the test sample is bandwidth dependent and is not a constant over the
sound path. A first order correction of the attenuation is sufficient for the measurement range up to 10
near field lengths.
Keywords: Single element, Phased array, universal DGS curve.

1. Introduction
The Distance Gain Size (DGS) technique is used in flaw detection and sizing by
registering flaw echo amplitude on DGS (AVG) diagrams, a series of curves
representing various disc size reflectors to sound path and gain. The first use of DGS
diagrams was reported by J.Krautkramer where he used FBH perpendicular to the beam
axis as the reference sizing echo [1]. The method was quickly used worldwide and the
diagrams are now supplied by the manufacturers to various types of probes [2]
including angle beam, straight beam and dual straight beam probes. The DGS curves
starts from the minimum detectable distance and ends at several near field lengths of the
probe. For mono crystal probes, the structure of the curve in the far field is solely
dependent on the FBH and probe size. The echo sensitivity from FBH in the far field
follows a simple relation [1],
AFBH

A0 ( b a )

(z N )

(1)

where a, b, z, and N are the probe diameter, FBH diameter, distance of FBH to the
probe surface, and the near field length of the probe, respectively.
Parameter A0 characterizes the probes sensitivity. For a particular probe type, the DGS
curves plot AFBH / A0 over z for various FBH size b. When the probes diameter a and
the near field length N are known, the DGS curves can be plotted simply by
renormalizing a set of universal DGS curves which plots AFBH / A0 over z / N for various
b/a. The universal DGS curves are based on the work of Krautkramer [1] and were
introduced in Germany around 1970s to be used for weld inspection. As its name

www.ndt.net/?id=9983

DGS Sizing Diagram with Single Element and Phased Array Angle
Beam Probe

implies, these curves are assumed to be representative for any probe type, regardless of
its size, shape, frequency or whether it is a normal or an angle beam probe. While this is
true in the far field, it is inaccurate within the near field of the probe. If the reflector is
inside near field zone of the probe, its response varies according to the amplitude
variation of the incident beam, which is influenced by the probes size, shape, frequency
and bandwidth. Especially for angle beam probes, the refraction of the sound beam in
the interface changes the effective probe size and the near field length which could leads
to uncertainties in the application of universal DGS curves. A recent investigation by
GE Sensing & Inspection Technologies in Huerth, Germany has shown that the near
field response from DGS diagram tends to underestimate the FBH amplitude compared
to measurement results with single element angle beam probes [3]. This is also proved
to be true in our model and measurement results shown later.
It is therefore difficult to characterize an applicable Universal near field curves. It
should be noted that in Krautkramers original work, the near field curves are obtained
based on a compromise derived from measurements and calculations for normal
incident and angle beam probes of various shape, size and frequencies. According to
European standard EN583, an empirical accepting sizing range is set to be greater than
0.7N with N been the near field length of the particular probe [4]. Although sizing is
problematic within near field, it is preferable to have more accurate and extended near
field DGS curves for many applications.
Basically there is no simple relation as Eq (1) within N of any type of probe. Kimura
et.al have shown analytical results to calculate the near field structure of the curves for
both normal and angle beam probes [5,6,7]. The result of normal probes has been
adopted in NIDS 2411-1980 (the standard of ultrasonic testing of steel forgings) [6].
Even though probes bandwidth and modal conversions are neglected, the results are
very complicated and show the explicit dependence on the probe type. While a closed
form expression for normal probes can be obtained, the results for angle beam probes
have to be calculated numerically [7]. When commercial modelling software is
available, calculations are made handy and can be easily adapted for different probe
types. For example, calculations of pulse echo responses from FBH for angle beam
probes are possible with CIVA, an ultrasound modelling software developed by French
Atomic Energy Commission. In CIVA, the probe size, shape, frequency and bandwidth
are input parameters. For an angle beam probe, the wedge geometry, velocity and
attenuation are wedge input parameters. The flawsFBH in particularcan be
positioned arbitrarily within the test sample. It is possible to model the pulse echo
response that includes modal conversions and sample attenuation. The receiving echo
can be displayed as an A-scan waveform. CIVA utilizes a semi-analytical method based
on Kirchhoff integral theory and the reciprocity principle. Semi-analytical approach
makes calculation much faster than Finite Element Model without loss of accuracy. A
two-way calculation from a FBH takes less than 1 second. In this paper, the echoes from
FBHs with 0.125 in. diameter are calculated with angle beam probes, and the results are
compared to the measurement data. The DGS curves generated from the model for FBH
are compared to those normalized from universal DGS curves. It is seen that in general,
the CIVA model results match the measurements within 1dB difference from 0.5N ~
5N.

With the capability of modelling phased array probes, DGS curves can also be
calculated for PA. The PA probes 4L16-DGS and 2L8-DGS (OlympusNDT Inc.) have
electronically steered refractive angles from 35 to 70 in steel. With a phase array
control instrument, such as OmniScan MX, S-scan and A-scan with specified angle can
be displayed on the same screen. As seen in Fig. 1, sizing can be performed on the Ascan while 2D visualization of the S-scan enhances the detection. DGS sizing with PA
reduces the need to use a number of UT probes and improves application flexibility.

Fig. 1 OmniScan MX display of DGS sizing with phased array probe 4L16-DGS. A-scan displays sizing
diagram with 60 refractive angle.

2. Model and measurement results


Three UT probes AM4R-8x9-45, -60, and -70 (OlympusNDT Inc.) and one PA probe
4L16-DGS are used for calculation and measurements at 4MHz. Three UT probes
AM2R-8x9-45, -60, and -70 (OlympusNDT Inc.) and one PA probe 2L8-DGS are used
for calculation and measurements at 2MHz. The control instruments are EPOCH XT
and OmniScan MX for UT and PA, respectively. Probes and instruments are calibrated
according to their standard procedures. The crystal dimensions for both UT and PA
probes are 8mm x 9mm. The bandwidths are 45% for UT and 87% for PA. The
refractive angles for UT are 45, 60, and 70, respectively for model -45, -60, and -70.
For comparison, 4L16-DGS and 2L8-DGS are steered at these refractive angles. The
sample material is 4340 steel block with several FBH inserts oriented at 45, 60. and
70. The FBH has a diameter of 0.125 in. and is oriented perpendicular to the incident
beam at various distances ranging from 0.5N~5N. The velocity of the steel is measured
to be 3.23mm/S during calibration. CIVA calculates the refractive shear wave
propagation in the steel that takes into account the modal conversions at wedge/steel
interfaces and at flaw boundaries. For every FBH position, the reflected echo is
displayed in the A-scan. The rectified max amplitude of the waveform is used as the
sizing amplitude.
2.1 UT model

The DGS curves (dashed lines) of the FBH calculated by CIVA for 2MHz and 4MHz
probes are shown in Fig. 2(a) and Fig. 2(b). These curves are compared to their
universal DGS curves shown as solid lines. In universal DGS curves the equivalent
diameter of the probe is 9.575mm. The near field lengths for 4MHz probes are 31mm,
28mm, and 25mm for model -45, -60, and -70, respectively. The corresponding near
field lengths for 2MHz probes are 17mm, 15mm, and 15mm. In the calculation, the
attenuation of the test sample is neglected. Clearly, when z>N, the two curves perfectly
match. When z<N, universal DGS curves are lower by as much as 5dB compared to the
model prediction. The closer the distance to the probe surfaces the bigger the amplitude
differences. The differences also increase as the refractive angle increases. The
universal DGS curves assume the probe has infinitely small bandwidth. In CIVA, the
bandwidth of the probe is intrinsically taken into account by using frequency dependent
coefficients. In addition, the near field structure in universal DGS curves does not
reflect particular probe type, its deviations from model outputs is expected.

Fig. 2 Comparison of universal DGS curves (solid line) with CIVA model curves (dashed line). The
target is 0.125 in. diameter FBH. (a): AM2R8x9-45, -60, -70. (b): AM4R8x9-45, -60, -70.

In Fig. 3, CIVA recalculates the DGS curves including the sample attenuation
(0.014dB/mm at 2MHz and 0.05dB/mm at 4MHz). The attenuation coefficient is
obtained by best fit of calculated curves with backwall measurement. All flaw echoes
are normalized to the backwall echo which is immediately in front of the probe/steel
interfaces [1]. As seen in the figure, the model agrees with the measurements very well.
Except 2dB difference is found from measured amplitude to the model within near field
length of AM2R-8x9-45 and -60, the remaining measurements of the amplitude differ
from model within 1dB.

Fig. 3 Comparison of measured FBH responses with the CIVA model curves (a): AM2R8x9-45, -60, 70. (b): AM4R8x9-45, -60, -70.

2.2 PA model
The probes used are 4L16-DGS at 4MHz with 16 elements and 2L8-DGS at 2MHz with
8 elements. Their bandwidths are 87%. The delay laws are calculated to provide
refractive angles of 45, 60, and 70 in steel. The calculated curves and the universal
DGS curves for 0.125 in. FBH are shown in Fig. 4. The universal DGS curves has an
equivalent diameter a=9.575mm. The near filed lengths for the 4MHz probe are 28mm,
25mm, and 21mm for refractive angles of 45, 60, and 70, respectively. The
corresponding near field length for the 2MHz probe are 16mm, 14mm, and 13mm. The
near field lengths of PA probe are 3mm shorter than those of UT probes at 4MHz and
about 1~2mm shorter at 2MHz. This is due to the broader bandwidth of the PA probe. It
is seen from Fig. 4 that the universal DGS curves coincide with the model in the far
field. In the near field, universal DGS curves underestimate the FBH amplitudes by as
much as 6dB. In Fehler! Verweisquelle konnte nicht gefunden werden., CIVA
recalculates the FBH responses by taken into account the attenuation of the steel
(0.014dB/mm at 2MHz and 0.05dB/mmm at 4MHz). The results, again, agree with the
measurements very well. The differences in amplitudes over the whole measurement
range are about 1dB.

Fig. 4 Comparison of universal DGS curves (solid line) with CIVA model curves (dashed line). The flaw
is 0.125 in. diameter FBH. (a): 2L8-DGS, (b): 4L16-DGS.

Fig. 5 Comparison of measured FBH responses with the CIVA model curves for 45, 60 and 70
refractive angles with probes (a): 2L8-DGS and (b): 4L16-DGS.

From the FBH measurements, it is seen that for angle beam probes, CIVA can be used
to generate their individual DGS curves with improved near field accuracy. It also
shows that the sizing curves in the far field range (at least z>3N) are independent on
probe type. In this range, the DGS curves for any unknown diameter FBH r(k) can be
obtained from the curves of known FBHs with diameters r(m) and r(m+1) following the
relation:
DGSk = DGSm + ( DGSm +1 DGSm )

log r ( k ) log r ( m )

log r ( m + 1) log r ( m )

(2)

where r(k), r(m) and r(m+1) represent the diameters of the FBHs with the relation
r(m)<r(k)<r(m+1). DGSk, DGSm, and DGSm+1 represent the DGS curves for FBH with
size r(k), r(m), and r(m+1), respectively. For flaws with different geometry other than
FBH in the range z>3N, the DGS curves can be referred to Ermolovs equation [8]. In
the range z<3N, the DGS curves are completed from model calculations for any
particular case.

3. Attenuation effect
In sizing flaws, material attenuation is applied to the DGS curves. The attenuation is
assumed to be linear over the sound path. This only applies to the probes with infinitely
narrow bandwidth. The attenuation for high frequency wideband probes is expected to
decrease over distance due to the shift of its spectrum toward the low frequency range.
In CIVA, the reflections echoes from FBH are calculated with and without sample
attenuation. The differences in echo amplitudes are the attenuation corrections. The
steel attenuation is 0.014dB/mm for 2MHz and 0.05dB/mm for 4MHz, which follows a
frequency square law [9]. The attenuation correction for UT and PA at 2 and 4MHz are
shown in Fig. 6. At 2MHz, the attenuation corrections are in close agreement with the
steel attenuation. At 4MHz, the attenuation correction is smaller than steel attenuation
and is not linear over the sound path. The larger the probes bandwidth, the smaller the
attenuation at long sound path. A first order polynomial is found to better fit the data for
UT and PA.

PA ( dB/mm ) = 0.00006 z + 0.05

UT ( dB/mm ) = 0.00002 z + 0.05

(3)
(4)

where z represents the two-way sound path. The coefficients of z represent the
bandwidth effect on attenuation while the constants represent the steel attenuation.
When sizing range is within 3N, the correction factor from bandwidth is small and only
accounts about 1dB difference. For sizing at 6N, it accounts for more than 3dB for UT
and more than 5dB for PA, the bandwidth effect on attenuation can not be neglected.

Fig. 6 Attenuation correction over the sound path for sizing with UT (45% bandwidth) and with PA (87%
bandwidth). (a): 2MHz, (b): 4MHz.

4. Conslusion
We have demonstrated that the semi-analytical model is able to predict responses with
improved accuracy from disk type reflectors for both angle beam UT and PA probes.
This provides a theoretical basis for extending the sizing range of DGS curves closer to
the probe surface than what is now specified in EN583, whose sizing is based on
universal DGS curves. However, expanding the useful sizing range should be based on
extensive filed measurement on different type of reflectors. In this paper, we see that the
universal DGS curves could underestimate the FBH response in the near field. The
errors, however, may decrease as the frequency and refractive angle decrease. Probes of
4MHz with 60deg or 70deg angle give the biggest discrepancy.
The effects of bandwidth of the probe are also considered in the paper. First, the PA
probe which has 87% bandwidth has a shorter near field length than that of UT probes
which has 45% bandwidth. This effect is more pronounced at 4MHz than at 2MHz.
Model also shows that the bandwidth affects the attenuation correction. It is more
pronounced at higher frequencies and with a wider bandwidth probe. At 2MHz,
bandwidth of the probe can be neglected. At 4MHz, within 3N the bandwidth effect can
be neglected with only 1dB error introduced. As sizing range increases beyond 3N, the
bandwidth of the probe should be taken into account.
Acknowledegment
The authors would like to thank Ms Erica Schumacher for providing simulation
software through Magsoft Corporation ((www.magsoft-flux.com) and any valuable
technical supports.
References
1. J. Krautkramer., Determination of the size of defects by the ultrasonic impulse echo
method. British Journal of Applied Physics. 10; 1959: 240-245.
2. Berke, M., Kleinert, W.-D., Reinersmann, J.,Schlengerman, U., Volkmann, K.,
Krautkramer - Germany, Kritsky, M., Passi, A., G.Passi., Sonotron NDT Israel.,
New Way for Utilizing DGS Technology. 15th WCNDT, Rome, 2000.
3. http://www.ndt.net/article/ecndt2010/reports/1_03_64.pdf.

4. Distance Gain Sizing Technique, European Standard DIN EN583-2:2001 Part2:


Sensitivity and range setting.
5. K. Kimura., Calculation of AVG Diagram for Ultrasonic Normal Probe,. Trans.
National Research Inst for Metals, 22, 1980.
6. K. Kimura., S. Matsumoto., DGS diagrams for angle probes, 3rd European
Conference on NDT, Florence, Italy, 1984.
7. K. Kimura., S. Matsumoto., DGS diagram for Ultrasonic angle beams, JSNDI, Vol
48, No.2. 1999.
8. I.. Ermolov., The reflection of ultrasonic waves from targets of simple geometry,.
NDT 1972
9. B. A. Auld, Elastic properties in solids, Acoustic fields and waves in solids, 2nd
edition, Vol I , 1990.

You might also like