MONITORING OF THE LEVEL OF RESIDUAL STRESS IN
SURFACE-TREATED SPECIMENS BY A NONCONTACTING
THERMOELECTRIC TECHNIQUE
B. Lakshminarayan, H. Carreon, and P. B. Nagy
Department of Aerospace Engineering and Engineering Mechanics
University of Cincinnati
Cincinnati, Ohio 45221-0070, USA
ABSTRACT. We have recently initiated the development of a noncontacting thermoelectric method
based on magnetic detection of local thermoelectric currents in the compressed near-surface layer of
surface-treated metals when a temperature gradient is established throughout the specimen. Beside the
primary residual stress effect, the thermoelectric method is also sensitive to the secondary "material"
effects of shot peening (local texture, increased dislocation density, hardening), but it is entirely
insensitive to its "geometrical" by-product, i.e., the rough surface topography. This method measures
only the weighted average of the near-surface residual stress, which is sufficient for quantitatively
evaluating the degree of thermally-induced stress release, but, in its present form, it is not suitable for
detailed mapping of the residual stress profile. Preliminary results are presented for shot-peened and
low-plasticity-burnished IN 100 nickel-base superalloy specimens to show that the technique is also
applicable to low-conductivity engine materials.
INTRODUCTION
Nondestructive evaluation (NDE) of the existing residual stress in the shallow
subsurface layer of surface-treated components could be very beneficial during
manufacturing to monitor and minimize process variations. Even more importantly, NDE is
absolutely necessary after extended service if residual stresses were to be taken credit for in
fatigue life predictions of critical components because of the very significant and highly
variable stress release that might occur at elevated operating temperatures. Currently, the
only reliable NDE method for residual stress assessment is based on X-ray diffraction
(XRD) measurement that is limited to an extremely thin («1 mil) surface layer, which is
approximately one order of magnitude less than the typical penetration depth of
compressive residual stresses produced by surface treatments. To obtain detailed depth
profiles of the residual stress distribution, successive layers are removed, usually through
etching or electropolishing, i.e., in a destructive manner. The removal of material also alters
the stress field, and thus requires theoretical corrections of the measured values.
Furthermore, since the method probes only the surface, the results can be easily skewed by
spurious effects in the extremely shallow top layer. In spite of the troublesome and
destructive sectioning required by the low penetration depth, XRD is probably the most
CP657, Review of Quantitative Nondestructive Evaluation Vol. 22, ed. by D. O. Thompson and D. E. Chimenti
© 2003 American Institute of Physics 0-7354-0117-9/03/$20.00
1523
accurate and reliable method for residual stress assessment in surface-treated metals. One of
the main reasons for this is that XRD methods are not significantly influenced by additional
variations in material properties such as hardness, plastic strain, or texture [1].
THERMOELECTRIC METHOD
It was recently demonstrated that self-referencing thermoelectric measurements can
be done in an entirely noncontacting way by using high-sensitivity magnetic detectors to
sense the weak thermoelectric currents around inclusions and other types of
inhomogeneities or material perturbations when the specimen to be tested is subjected to
directional heating or cooling [2-8]. External heating or cooling is applied to the specimen
to produce a substantial temperature gradient in the region to be tested. As a result, different
points of the boundary between the host material and the inclusion will be at different
temperatures, therefore also at different thermoelectric potentials. These potential
differences will produce opposite thermoelectric currents inside and outside the inclusion.
The thermoelectric currents form local loops that run in opposite directions on opposite
sides of the inclusion or imperfection relative to the prevailing heat flux. When the
specimen is scanned with a sensitive magnetometer, the magnetic field of these
thermoelectric currents can be detected even when the inclusion is buried below the surface
and the sensor is far away from the specimen.
It was also demonstrated that the thermoelectric method is entirely insensitive to
surface topography while it is uniquely sensitive to subtle variations in thermoelectric
properties, that are associated with the different material effects of shot peening [9-11].
These results indicate that the stress-dependence of the thermoelectric power in metals
produces sufficient contrast to detect and quantitatively characterize regions under
compressive residual stress based on their thermoelectric contrast with respect to the
surrounding intact material. In this paper we will demonstrate that the recently developed
experimental technique is sufficiently sensitive to weak material variations produced by
different surface treatments such as shot peening (SP) and low-plasticity burnishing (LPB)
and offers a feasible alternative to conventional ultrasonic and eddy current methods for
residual stress assessment.
THERMOELECTRIC RESIDUAL STRESS ASSESSMENT IN COPPER
Figure 1 shows a schematic diagram of the noncontacting thermoelectric method
used for the characterization of shot-peened copper specimens. The ends of the shot-peened
bars were simultaneously heated and cooled by running water and the temperature gradient
was kept at 2.3 °C/cm. Since the generated magnetic field is perpendicular to both the heat
flux in the specimen and the gradient of the material property, the magnetometer was
polarized in the tangential direction. The measurements were repeated after different
degrees of partial stress release to separate the individual effects of residual stress, cold
work induced texture and hardening, and surface topography on the measured
thermoelectric signature.
Figure 2 shows the peak-to-peak amplitudes of the magnetic signatures recorded on
nine C11000 copper specimens as functions of the shot peening intensity at different stages
of stress release. The partial stress relaxations at 235 °C and 275 °C resulted only in modest
30% and 50% drops in the magnetic signature, respectively, while the repeated annealing at
315 °C (second time) and 460 °C further reduced the amplitude. Finally, there was no
detectable magnetic signature left after full recrystallization at 600 °C. The measured
magnetic signature is essentially a linear function of the shot peening intensity and it
1524
fluxgate
gradiometer
shot-peened area
hot (cold)
water
cold (hot)
water
<==>
FIGURE 1. A schematic diagram of the noncontacting thermoelectric method as used for the characterization
of shot-peened copper specimens.
gradually decreases during relaxation to zero in fully recrystallized specimens. These trends
are very promising for the feasibility of nondestructive monitoring of thermal relaxation in
shot-peened copper specimens, but they do not provide unequivocal evidence whether the
magnetic signature is caused mainly by the presence of residual stress, the presence of cold
work, or a certain combination of both.
In order to better separate these two effects, we had a series of XRD measurements
conducted on the intact and partially relaxed specimens. Figure 3 shows examples of the
residual stress and cold work profiles recorded on specimens of two different Almen
intensity. These figures give a relatively detailed picture of the initial relaxation process,
though the accuracy of the very expensive and time consuming destructive XRD
measurement is obviously less than sufficient to precisely map these distributions and some
of the ruggedness of the profiles was certainly caused by experimental errors. The accuracy
of the residual stress measurement was estimated at approximately ±25 MPa while the
2nd relaxation at 3! 5
3rd
at 460 °
reerystaUiiitto at 600 °C
FIGURE 2. Average peak-to-peak amplitudes of the magnetic signatures recorded on C11000 copper
specimens as functions of the shot peening intensity at different stages of stress release.
1525
300
.§ 200* [
400
ji&
!2 1CIO
—J
-2CI0
0,2
0.4
Depth [mm]
02
0.4
Depth [mm]
0.6
1
0.6
14
8
CO
GO
o 100
INI
K
-200
0.2
0.4
0
0.4
0.6
Depth [mm]
Depth [mm]
— intact —<
OJ
235 °C
275 °C —O— 315 °C
FIGURE 3. Thermal relaxation of residual stress and cold work in Copper Cl 1000 for two different Almen
intensities (30 minutes in a vacuum furnace).
accuracy of the peak width measurement is approximately ±0.15°, which translates into
approximately ±10% plastic strain.
Figure 4 shows the thermal relaxation of the integrated residual stress, integrated
cold work, and peak-to-peak magnetic signature in Copper C11000 for different Almen
intensities. In order to avoid that the integrated profiles be dominated by the experimental
errors at large depths where the true values are very low, we assumed that the residual stress
and cold work are zero wherever their magnitude was less than their respective errors, i.e.,
25 MPa and 10%. A statistical comparison of these integrated residual stress and cold work
results to the magnetic signature revealed that in shot-peened C11000 copper, on the
average, -64% of the thermoelectric signal is due to residual stresses.
THERMOELECTRIC RESIDUAL STRESS ASSESSMENT IN IN100
The above experimental results clearly verify the feasibility of nondestructive
evaluation of thermal relaxation in shot-peened C11000 copper. Although the development
of a quantitative residual stress measurement method might require additional more
accurate and more detailed tests, the obvious next question to be addressed is whether the
thermoelectric method is applicable to other engineering materials of special interest to the
aerospace industry. In particular, it is very important to develop NDE techniques for highstrength, high-temperature engine materials such as nickel-base superalloys. We have
conducted additional thermoelectric measurements on two series of shot-peened and lowplasticity-burnished IN 100 specimens both before and after partial stress relaxation. Like
most engine materials, nickel-base superalloys exhibit much lower thermal and electrical
conductivity and more significant microstructural inhomogeneity and anisotropic texture
than single-phase pure copper. Therefore, we had to make certain modifications to the
experimental configuration previously shown in Figure 1. Figure 5 shows a photograph and
1526
03
OH
§
3
*S
&
2
6
70
60
50
40
30
201
10
0
2 6 10 14
10 14
Almen Intensity
mtact
Almen Intensity
235 C
• 275 *C
2
6
10 14
Almen Intensity
O 315°C
FIGURE 4. Thermal relaxation of the integrated residual stress, integrated cold work, and peak-to-peak
magnetic signature in Copper Cl 1000 for different Almen intensities.
a schematic diagram of the modified thermoelectric inspection system using noncontacting
forced air heating and cooling. In order to increase the temperature gradient in the region of
interest we used a so-called vortex tube to produce high- and low-temperature air streams
that are directed at the specimen by nozzles placed on the two sides of the fluxgate. The
vortex tube is balanced so that, at the air pressure used to drive it, the increase and decrease
of the outgoing air temperature on the hot and cold sides, respectively, are essentially the
same. This mode of forced-air heating and cooling assures that significant temperature
gradients arise only in the vicinity of the surface and the rest of the specimen remains at
room temperature. In this way, beside the obvious advantage of being noncontacting, this
type of thermal excitation is much less sensitive to bulk inhomogeneity and anisotropy in
the interior of the specimen, therefore it produces a lower background signal that could
interfere with the useful signal due to surface treatment.
The magnetic signatures produced by the forced-air system at ~30 °C/cm
temperature gradient are similar to but stronger than those obtained by the water-based
system on low-conductivity metals. Figure 6 shows examples of the magnetic signatures
recorded from six Almen 6 and six Almen 8 shot-peened IN 100 specimens before stress
relaxation. It should be mentioned that the relative scatter of the measurement is somewhat
increased by the thermal instability of the fluxgate due to the close vicinity of the heating
and cooling nozzles, therefore all signatures were measured three times and averaged to
obtain the peak-to-peak amplitudes reported later. Figure 7 shows the peak-to-peak
magnetic flux density as a function of the stress release time at 670 °C annealing
temperature for shot-peened (SP) IN 100 specimens. In spite of the considerable scatter in
the data, which is partially due to inherent variations from specimen to specimen and partly
due to the previously mentioned reduced reproducibility of the measurement, a clear
increase with peening intensity as well as a clear decay with increasing annealing time is
apparent in the magnetic flux density. Similar results were also obtained from lowplasticity-burnished (LPB) specimens of "low" and "high" grade (see Figure 8). Generally,
the magnetic signatures recorded on LPB specimens are about 50% higher than from the SP
specimens, which is probably caused by the deeper penetration depth of the residual stress
profile. It is also interesting that the relaxation is weaker and occurs slower in LPB
specimens than in SP specimens, which is probably caused by the lower level of cold work
1527
cooling
fluxgate
.
00
heating
surface
treatment
specimen
FIGURE 5. A photograph and a schematic diagram of the modified thermoelectric inspection system using
noncontacting forced air heating and cooling.
in the former case (unfortunately, X-ray diffraction data is not available at this point on the
residual stress and cold work profiles).
CONCLUSIONS
We presented experimental results that illustrate the potential for a new NDE
technique to detect plastic deformation and the presence of residual stresses that are very
difficult to characterize by other, more conventional NDE methods. It has been found that
the noncontacting thermoelectric method can be used to characterize the prevailing residual
stress in surface-treated specimens. This novel method is based on magnetic detection of
local thermoelectric currents in the compressed near-surface layer of surface-treated metals
when a temperature gradient is established throughout the specimen. Beside the primary
residual stress effect, the thermoelectric method is also sensitive to the secondary "material"
effects of surface treatment, but it is entirely insensitive to its "geometrical" by-product, i.e.,
the rough surface topography. Further research is needed to better separate residual stress
effects from secondary material effects, especially in the case of low-conductivity engine
materials like titanium alloys and nickel-base superalloys.
ACKNOWLEDGEMENTS
This work was supported by the Metals, Ceramics, and NDE Division of the Air
Force Research Laboratory under contract #01-S437-002-34-Cl. The authors would like to
express their gratitude to Mark Blodgett of AFRL for his valuable contributions to this
work.
1528
c
(5
5
p
o w
Y
0
0
m
0
0
w
0
0
m
0
0
w
0
m
0
w
O
'
w
w
P
V
I
T
=1.
m
m
Magnetic Flux Density [nT]
'd
0
c
2
E
d
B0v
KL
5'
0
BB
F
$i?
. - 4
0
0
m
w
0
0
0
0
c
w
w
P
Magnetic Flux Density [nT]
3
za
0
%
Q\
C
8
Magnetic Flux Density
[l nT/div]
Magnetic Flux Density
[I nT/div]
REFERENCES
1. Prevey, P. S., in IITTInternational (Gournay-Sur-Marne, France, 1990) pp. 81-93.
2. Hinken, J. H., and Tavrin, Y., "Thermoelectric Squid Method for the Detection of
Segregations", in Review of Progress in QNDE Vol. 19 (AIP, Melville, 2000) pp. 20852092.
3. Nagy, P. B., and Nayfeh, A. H., J. Appl. Phys. 87, 7481 (2000)
4. Carreon, H., Nagy, P. B., and Nayfeh, A. H., J. Appl. Phys. 88, 6495 (2000).
5. Nayfeh, A. H., Carreon, H., and Nagy, P. B., "On the Exploitation of Thermoelectric
Coupling for Characterization of Inclusions in Metals", in Review of Progress in QNDE
Vol. 20 (AIP, Melville, 2001) pp. 1686-1693.
6. Carreon, H., Nagy, P. B., and Nayfeh, A. H.,"Noncontacting Thermoelectric Detection
of Spherical Inclusions in Copper by Magnetic Sensing", in Review of Progress in
QNDE Vol. 20 (AIP, Melville, 2001) pp. 1694-1701.
7. Tavrin, Y., Krivoy, G. S., Hinken, J. H., and Kallmeyer, J. P., "A Thermoelectric
Method to Detect Weak Local Structural Chanages and Inclusions in Metal Parts", in
Review of Progress in QNDE Vol. 20 (AIP, Melville, 2001) pp. 1710-1716.
8. Maslov, K., and Kinra, V. K., Mat. Eval. 59, 1081 (2001).
9. Carreon, H., and Nagy, P. B., in NDE & Char. Eng. Mat. Reliability & Durability
Predict., AMD-Vol 240/NDE-Vol 18 (ASME, New York, 2000) pp. 31-39.
10. Carreon, H., Nagy, P. B., and Blodgett, M. P., "Thermoelectric Nondestructive
Evaluation of Residual Stress in Shot-Peened Metals", in Review of Progress in QNDE
Vol. 21 (AIP, Melville, 2002) pp. 1667-1674.
11. Carreon, H., Nagy, P. B., and Blodgett, M. P., Res. Nondestr. Eval. 14, 59 (2002).
1530