Actual Unloading Behavior and Its

Significance on Residual Stress

Hamid Jahed
Associate Professor
in Machined Autofrettaged Tubes
e-mail: hjahedmo@iust.ac.ir
A simple method for obtaining actual unloading behavior of high strength steels is pro-
Ghader Ghanbari posed. Based on this method loading-unloading stress-strain curves for NiCrMoV125
Graduate Student steel is obtained. It is shown that the actual unloading behavior tends to be nonlinear as
soon as unloading occurs. Using the variable material properties (VMP) method, residual
Mechanical Engineering Department, stresses induced by autofrettage process of a tube made of this type of steel are calculated.
Iran University of Science & Technology, Significance of employing actual unloading behavior is demonstrated by comparing the
Tehran, Iran 16844 results with ideal models such as isotropic hardening and bilinear models. There is at
least 30% over estimation of compressive residual stress at the bore if ideal models are
used. Also, the process of material removal is simulated by using commercial FEM soft-
ware to evaluate the recommended scenarios of removal on redistribution of residual
stresses at the bore. Resulting stresses after machining are also calculated by the VMP
method using actual behavior. It is shown that the final residual stress field differs sig-
nificantly if actual unloading behavior is used. 关DOI: 10.1115/1.1593070兴

Introduction
Autofrettage is used to introduce beneficial compressive hoop
stress at the bore of pressure vessels which enhances also the
fatigue life limits. Though the history of high pressure technology
is very old 共historical review and advancements is given by Ken-
dall 关1兴 and Perl and Arone 关2兴兲 and the theoretical understanding
of the pressurized thick-walled cylinders is well established, there
are interesting challenges still ahead.
There have been numerous papers, spanning almost two de-
cades, relating to the modeling of the unloading behavior of au-
tofrettage steels and the subject is still of interest.
Milligan et al. 关3兴 showed there exist a profound influence of
the Bauschinger 关4兴 effect, which serves to reduce the yield
strength in compression as a result of prior tensile plastic over-
load, in high strength steels. They also showed that the Bausch-
inger effect factor 共BEF兲 is a function of plastic strain. This de-
pendency was formulated later by Kendall 关5兴.
With the exception of few, none of the works regarding BEF
recorded the actual loading-unloading behavior. In many of these
investigations the uniaxial stress-strain curve was obtained to find
the values of the modulus of elasticity, Poisson’s ratio and yield
strength. However, a few measured and recorded the unloading
behavior. Stacey and Webster 关6兴 共Fig. 2兲 recorded experimentally
obtained stress-strain curve for AISI 4333 steel. It is seen from
these results that the unloading behavior is nonlinear from the
starting point and a hardening effect is experienced by the mate-
rial afterwards.
Chaaban 关7兴, in a rather complex manner, provided a method of
generating stress-strain curves as a function of plastic strain by
introducing fictitious thermal loads. This method, which was ba-
sically designed for use with a FEM commercial package, consid-
ers the dependency of BEF on plastic over strain. Chaaban et al.
关8兴 used bilinear model to calculate autofrettage residual stress.
Obviously, the method needs actual unloading curve to be
incorporated.
Chen 关9兴 proposed a bilinear approximation of the unloading
curve to account for both, the Bauschinger effect which reduces
the yielding in compression and also to account for the hardening
Contributed by the Pressure Vessels and Piping Division for publication in the
JOURNAL OF PRESSURE VESSEL TECHNOLOGY. Manuscript received by the PVP
Division March 13, 2003; revision received April 23, 2003. Associate Editor: M Perl.
effect after yielding. This method needs actual curve to come up
with slope of two lines of bilinear approximation.
Megahed and Abbas 关10兴 attempted to follow a full nonlinear
unloading profile which is a function of plastic strain, but the
analysis requires an empirical equation describing BEF depen-
dence on plastic strain. Their method is limited to materials whose
behavior is modeled by power law. The actual curve is needed to
come up with corresponding coefficient of power law.
In a similar problem, residual stress in cold worked fastener
holes, Possard et al. 关11兴 recorded the experimentally obtained
loading-unloading behavior of high strength Aluminum 2024
T351. This type of aluminum posses a similar behavior to high
strength steels. This work records unloading behavior at three dif-
ferent level of over strain. The change in unloading behavior due
to change in plastic strain is evident. Jahed and Dubey 关12,13兴
showed the significance of considering variable BEF in residual
stress calculation in autofrettage tubes. Parker and Underwood
关14兴 and Parker et al. 关15兴 reviewed and compared extensively
different unloading models.
The purpose of this paper is to introduce a simple method for
obtaining actual unloading behavior of autofrettage steels. A typi-
cal high strength steel is selected and corresponding first cycle
loading-unloading curve is obtained using a simple forward-
reverse torsion test.
Using the experimental results, different ideal models including
bilinear models are evaluated.
The variable material properties 共VMP兲 method of Jahed 关16兴
and Jahed and Dubey 关13兴 is employed to predict the residual
stress field due to autofrettage based on the actual unloading be-
havior. The significance of utilizing actual behavior as oppose to
ideal models is discussed. It is then suggested that, by using this
simple test, unloading material curve be measured prior to re-
sidual stress analysis using the VMP method.
Also, material removal influence on residual stress is studied.
The removal process is first simulated using ANSYS 关17兴 FEM
commercial software to verify recommended methods for analysis
of machined tubes. It is also shown that final residual stress after
material removal is depending on actual unloading behavior.
Test Procedure and Theory. A method of testing developed
by Baily 关18兴 is used to determine the first cycle material curve. In
this method a solid circular specimen is subjected to forward-
reverse torsion. The first cycle curve is obtained for points on the

Journal of Pressure Vessel Technology Copyright © 2003 by ASME AUGUST 2003, Vol. 125 Õ 321

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 09/16/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 2 Test specimen

The following constitutive shear stress-strain relation is as-

sumed:
␶⫽ f 共 ␥ 兲. (6)
Hence, Eq. 共5兲 will take the following form:

T⫽ 冕 0
␥a 2 ␲ f 共 ␥ 兲␥ 2d ␥
␪ l3
. (7)

Using the compatibility relation of ␥ a and ␪ l and differentiation

of equilibrium Eq. 共7兲 the following is obtained:
Fig. 1 Torsion of a circular section

␶ a⫽
2␲a3
1
冋 3T⫹ ␪ l
dT
d␪l
. 册 (8)

surface of the specimen. As oppose to tubular specimen where
plastic instability occurs very fast, here due to the elastic interior
points, large strains may be applied.
In this test following assumptions are made:
1兲 The material is homogeneous and isotropic.
2兲 Twist per unit length of the round specimen is constant.
3兲 Plane sections remain plane after deformation.
4兲 Initially straight radii remains straight after deformation.
Figure 1 shows deformation of a of circular section under tor-
sion.
Assuming above constraints the circumfrential displacement is
given by
In Eq. 共8兲, the total value of transmitted torque and the twist
angle per unit length ␪ l can be measured directly from the torsion
test. Also, the value of dT/d ␪ l can be calculated numerically or
graphically once the T⫺ ␪ l diagram is obtained from the test.
Moreover, the shear strain at the outer surface of the cylinder is
determined by the following equation:
␥ a ⫽a ␪ l . (9)
Hence, the shear stress-strain curve of the material is obtained.
The obtained curve 共data兲 may be used directly in the calcula-
tions of the residual stresses or be converted to normal stress-
strain curve with the use of effective values. For two common

u ␪ ⫽rz 冉 冊
d␪
dz
, (1)
yield criteria, i.e., von Mises and Tresca, equivalent values are as
follows:

where (d ␪ /dz) is the twist angle per unit length. Therefore with ␥
¯␧ M ⫽ , ¯␴ M ⫽ 冑3 ␶ , (10)
the use of second assumption the twist angle per unit length may 冑3
be replaced by ( ␪ /L), where L is the gage length of the specimen.
The only nonzero component of strain is ␥ z ␪ and hence there is 2
¯␧ T ⫽ ␥ , ¯␴ T ⫽2 ␶ , (11)
only one nonzero component of stress, ␶ z ␪ . The stress distribution 3
across section is given by
where ¯␧ M , ¯␴ M are the von Mises equivalent strain and stress and
Gr ␪ ¯␧ T , ¯␴ T are the Tresca equivalent strain and stress, respectively.
␶ z␪⫽ , (2)
L Test Equipments and Material. The tests are carried out in
where G is the shear modulus, r is the section radius. fatigue laboratory at IUST. Advanced torsion test equipment 共SM-
Torque 共T兲 transmitted by the section is 21兲 is used.

冕
The material used in the tests is NiCrMoV125 steel with the
a
following chemical composition:
T⫽ ␶ z ␪ 2 ␲ r 2 dr, (3)
0 carbon 0.3–0.4, silicon 0.15–0.35, manganese 0.4 –0.7,
nickel 2.5–3.5, vanadium 0.08 –0.2, aluminum 0.05 and lower,
where a is the cylinder radius. Hence, elastic torsion formula is phosphorous 0.015, sulfur 0.015 and lower.
yielded, Test specimens are made from virgin material at different ori-
TL entations 共i.e., circumferential, radial and longitudinal兲. A typical
␪⫽ , ␶ z ␪ max⫽ Ta , (4) specimen drawing is shown in Fig. 2 共all dimensions in mm兲.
GJ J
Tests are conducted on 25 different specimens to reduce mea-
where J is the polar moment of inertia. surements error. Three different curves are fitted through regions
Substitution of ␥ ⫽r ␪ /L in Eq. 共3兲 gives the following: of elastic loading, plastic loading and unloading of the material.
Figure 3 shows the first cycle stress-strain curve at three different
T⫽ 冕0
␥a ␶ 2 ␲ ␥ 2d ␥
␪ l3
, (5)
plastic strain values 共1, 2, and 3 shown in Fig. 3兲.
Test results show a perfectly plastic behavior during loading
and a complete nonlinear behavior upon unloading. It is evident
where ␥ a is shear strain at the surface of the cylinder and ␪ l is that the unloading behavior cannot be modeled by any of the ideal
twist angle per unit length. models. Figure 4 compares the actual behavior and ideal models.

322 Õ Vol. 125, AUGUST 2003 Transactions of the ASME

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 09/16/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 3 Actual stress-strain behavior
Residual Stress: Actual Behavior Versus Ideal Models. The
results in previous section show that not only the BEF depends on
the amount of over strain but also material behaves in a nonlinear
fashion as soon as unloading starts. Hence, a nonlinear analysis
for the unloading is necessary. It is important to note that the ideal
models, like Chen’s 关9兴 bilinear model, need actual unloading
curve to come up with good approximation for linearization of
unloading behavior. If such curve is obtained it might be as well
used in the analysis directly. Since the VMP method is capable of
utilizing the actual behavior, it is believed to be the most accurate
method for residual stress prediction in autofrettage.
To compare the results of different models and to show the
significance of using actual unloading behavior, a cylinder of 146
mm ID and 305 mm OD of the same material used for experi-
ments in previous section is used for the analysis. The autofrettage
pressure is assumed to be 736 MPa.
Figure 5 shows the comparison of the results obtained from
ideal 共isotropic and kinematic hardening reveal the same results兲
models and that of actual behavior.
Here, unloading behavior at three different plastic strain 共i.e.,
0.005, 0.0152, and 0.0254兲 is available and for points in between
an interpolation is used. It is seen from Fig. 5 that there is a 100%
overestimation by the ideal models.
The closest model to actual behavior is that of the bilinear
model expressed as a function of plastic strain. The results of
using the bilinear model are compared with the actual behavior in
Fig. 6. While the results are improved by using bilinear model
Fig. 4 Actual behavior versus ideal model
Journal of Pressure Vessel Technology
Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 09/16/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Fig. 5 Residual stresses profile incorporating actual behavior
and ideal model
with appropriate slopes, still there is a 30% overestimation of
compressive hoop stress at the bore. Also, the bilinear model
shows a clear reversed yielding zone near the bore which is not
seen when actual unloading behavior is used.
Unlike the ideal and bilinear models, the actual unloading re-
sults do not show a sharp maximum residual tensile stress nor
does it suggest that it occurs at the elastic-plastic radius.
Material Removal Influence on Residual Stress. It is fre-
quently the case that material is removed from inside diameter
and/or outside diameter of a tube after autofrettage. This process
changes the residual stress field. To estimate the influence of ma-
terial removal on residual stress induced by autofrettage, different
methods have been proposed. Parker et al. 关15兴 have summarized
different scenarios.
To evaluate these common methods, the process is simulated
using ANSYS 关17兴 finite element commercial software. To do this,
a cylinder is loaded in three steps. First it is pressurized to autof-
rettage pressure and a nonlinear elasticplastic analysis is per-
formed. Then, the pressure is removed by performing another
nonlinear analysis. Here, a bilinear behavior is assumed for un-
loading analysis. In the third step, with the use of element birth
and death option of ANSYS 关17兴, removal is simulated by killing
corresponding elements. Here, different sequences may be studied
Fig. 6 Residual stresses profile incorporating actual behavior
and bilinear model
AUGUST 2003, Vol. 125 Õ 323
 Again, by comparing Figs. 7 and 8, the final stress field in the
machined tube is highly overestimated when ideal models are
used.

Conclusions
A simple test for obtaining actual loading-unloading material
behavior was discussed. It was shown that usual testing problems,
like plastic instability and buckling, will not happen in forward-
reverse torsion of a solid cylinder.
Material loading-unloading of DIN1.6959 steel was measured
and compared with the common ideal models. It was shown that
the material follow a nonlinear behavior as soon as it is unloaded.
The VMP method, which is capable of considering unloading
curve as it is, was implemented to calculate the autofrettage re-
sidual stress field. The calculated values was compared with esti-
mations of ideal and bilinear models. It was shown that the dis-
Fig. 7 Comparison of removal simulation using ANSYS and crepancy of the results at the bore is significant. It was then
VMP proposed that as long as the actual unloading behavior is avail-
able, the VMP method can be used for more accurate results.
Material removal process was simulated using FEM commer-
cial software. The recommended methods for consideration of ma-
subdividing the third step to two steps. That is, killing the inside terial removal were evaluated. Based on this evaluation, one of the
diameter elements first followed by killing the outside elements or consistent methods was used with the VMP method to calculate
vice versa. the final stress distribution upon material removal. It was shown
By using the VMP method and assuming same unloading again that actual unloading results of the final residual stress dif-
model as the one used in ANSYS, the following scenario matches fers significantly from the common models.
the simulation results.
Assume original cylinder of inner and outer radii as b and c,
respectively. Let the final dimension of machined tube be b ⬘ and Acknowledgments
c ⬘ . Also, let the radial residual stress be ␴ rR . The following four The authors would like to thank Hadid Industries for providing
steps are then performed. test samples. The help of Mr. Hosseinian and Mr. Ahmadi of Ha-
共1兲 Calculate the residual stresses of an autofrettaged cylinder did is acknowledged.
with dimensions b and c.
共2兲 Determine residual radial stress at b ⬘ and c ⬘ 共i.e., ␴ rR (b ⬘ )
and ␴ rR (c ⬘ ). Nomenclature
共3兲 Calculate the stress field in a cylinder with dimensions, b ⬘ ¯␧ M ⫽ von Mises equivalent strain
and c ⬘ , under internal pressure ␴ rR (b ⬘ ) and external pres- ¯␧ T ⫽ Tresca equivalent strain
sure ␴ rR (c ⬘ ). ¯␴ M ⫽ von Mises equivalent stress
共4兲 Superpose the results of step 3 on the results of step 1. ¯␴ T ⫽ Tresca equivalent stress
The results obtained from ANSYS is compared to the VMP ␴ rR (b ⬘ ) ⫽ Autofrettage radial residual stress at b ⬘
method using the above four steps 共Fig. 7兲. ␴ rR (c ⬘ ) ⫽ Autofrettage radial residual stress at c ⬘
The two results agree very well. Hence, the above four steps is ␶a ⫽ Shear stress at the surface of the solid cylinder
used with actual behavior. The same cylinder is considered and ␶ z␪ ⫽ Shear stress
the above four steps are performed using actual material loading- ␶ z ␪ max ⫽ Maximum shear stress
unloading behavior. The final residual stresses are shown in Fig. 8. ␥a ⫽ Shear strain of a point at the surface of the solid
cylinder
␥ z␪ ⫽ Shear strain
␪ /L, ␪ l ⫽ Angle of twist per unit length
␪ ⫽ Angle of twist
a ⫽ Solid cylinder outer radius
b ⫽ Inner radius of cylinder prior to machining
c ⫽ Outer radius of cylinder prior to machining
b⬘ ⫽ Inner radius of cylinder after machining
c⬘ ⫽ Outer radius of cylinder after machining
dT ⫽ Torque increment
dz ⫽ Longitudinal element
d␥ ⫽ Shear strain increment
d␪ ⫽ Angle of twist increment
d␪l ⫽ Angle of twist per unit length increment
d ␪ /dz ⫽ Angle of twist per unit length
f (␥) ⫽ Shear strain function
G ⫽ Shear modulus
J ⫽ Polar moment of inertia
L ⫽ Cylinder length
r ⫽ Cylinder radii
Fig. 8 Final residual stress field of a machined tube using ac- T ⫽ Torque
tual loading-unloading behavior u␪ ⫽ Circumferential displacement

324 Õ Vol. 125, AUGUST 2003 Transactions of the ASME

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 09/16/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 References
关1兴 Kendall, D. P., 2002, ‘‘A Short History of High Pressure Technology From
Bridgman to Division 3,’’ ASME J. Pressure Vessel Technol., 122, pp. 229–
233.
关2兴 Perl, M., and Arone, R., 1994, ‘‘An Axisymmetric Stress Release Method for
Measuring the Autofrettage Level in Thick-Walled Cylinders—Part 1: Basic
Concept and Numerical Simulation,’’ ASME J. Pressure Vessel Technol., 116,
pp. 384 –388.
关3兴 Milligan, R. V., Koo, W. H., and Davidson, T. E., 1966, ‘‘The Bauschinger
Effect in a High-Strength Steel,’’ J. Basic Eng., pp. 480– 488.
关4兴 Bauschinger, J., 1881, ‘‘Ueber die Veranderung der Elasticitatagrenze und dea
Elasticitatamoduls Verschiadener Metalle,’’ Zivilingenieur, 27, pp. 289–348.
关5兴 Kendall, D. P., 1998, Discussion of paper ‘‘The Bauschinger Effect in Autof-
rettaged Tubes—A Comparison of Models Including the ASME Code,’’ by A.
P. Parker and J. H. Underwood, Proceedings, ASME Pressure Vessels and
Piping Conference, San Diego, CA.
关6兴 Stacey, A., and Webster, G. A., 1988, ‘‘Determination of Residual Stress Dis-
tribution in Autofrettaged Tubing,’’ Int. J. Pressure Vessels Piping, 31, pp.
205–220.
关7兴 Chaaban, A., 1985, ‘‘Static and Fatigue Design of High Pressure Vessels With
Blind-End and Cross-Bores,’’ Ph.D. thesis, University of Waterloo, Waterloo,
Ontario, Canada.
关8兴 Chaaban, A., Leung, K., and Burns, D. J., 1986, ‘‘Residual Stress in Autofret-
taged Thick-Walled High Pressure Vessels,’’ ASME PVP, 110, pp. 55– 60.
关9兴 Chen, P. C. T., 1986, ‘‘The Bauschinger and Hardening Effect on Residual
Stresses in an Autofrettaged Thick-Walled Cylinder,’’ ASME J. Pressure Vessel
Technol., 108, pp. 108 –112.
Journal of Pressure Vessel Technology
Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 09/16/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use
关10兴 Megahed, M. M., and Abbas, A. T., 1991, ‘‘Influence of Reverse Yielding on
Residual Stresses Induced by Autofrettage,’’ Int. J. Mech. Sci., 33, pp. 139–
150.
关11兴 Poussard, C., Pavier, M., and Smith, D. J., 1995, ‘‘Analytical and Finite Ele-
ment Predictions of Residual Stresses in Cold Worked Fastener Holes,’’ J.
Strain Anal. Eng. Des., 30, pp. 291–304.
关12兴 Jahed, H., and Dubey, R. N., 1996, ‘‘Residual Stresses Calculation in Autof-
rettage Using Variable Material Properties Method,’’ ASME PVP, Residual
Stress in Design, Fabrication, Assessment and Repair, 327, pp. 181–188.
关13兴 Jahed, H., and Dubey, R. N., 1997, ‘‘An Axisymmetric Method of Elastic-
Plastic Analysis Capable of Predicting Residual Stress Field,’’ ASME J. Pres-
sure Vessel Technol., 119, pp. 264 –273.
关14兴 Parker, A. P., and Underwood, J. H., 1998, ‘‘Influence of the Bauschinger
Effect on Residual Stress and Fatigue Lifetimes in Autofrettaged Thick-Walled
Cylinders,’’ Fatigue and Fracture Mechanics: 29th Volume, ASTM STP 1321,
T. L. Panontin and S. D. Sheppard, eds.
关15兴 Parker, A. P., Underwood, J. H., and Kendall, D. P., 1999, ‘‘Bauschinger Effect
Design Procedure for Autofrettaged Tubes Including Material Removal and
Sachs’ Method,’’ ASME J. Pressure Vessel Technol., 121, pp. 430– 437.
关16兴 Jahed, H., 1997, Variable Material Property Approach for Elastic-Plastic
Analysis of Proportional and Nonproportional Loading, Ph.D. thesis, Univer-
sity of Waterloo, Waterloo, Ontario, Canada.
关17兴 ANSYS, 2000, Element Manual, Ver. 5.7.
关18兴 Bailey, J. A. ‘‘Fundamental Aspects of Torsional Loading,’’ Metal Handbook,
8, ASM International, Materials Park, OH, pp. 139–144.
AUGUST 2003, Vol. 125 Õ 325