Академический Документы
Профессиональный Документы
Культура Документы
Abstract
The objective of this work was to analyze the residual stress state in spot welds made in an HY-100 steel disk by an autogenous
gas tungsten arc (GTA) welding process. An uncoupled thermal-mechanical finite element (FE) model was developed that took
into account the effects of liquid-to-solid and solid-state phase transformations. Effects of variations in mechanical properties due
to solid-state phase transformations on residual stresses in the weld were studied. Extensive experimental testing was carried out
to determine the mechanical properties of HY-100 steel. The residual stresses in the disk with the spot weld were measured by a
neutron diffraction (ND) technique. The FE results are in good agreement with the ND measurements. The results show that the
volumetric changes associated with the austenite to martensite phase transformation in HY-100 steel significantly affect residual
stresses in the weld fusion zone and the heat affected zone. © 1998 Elsevier Science S.A. All rights reserved.
Keywords: Finite element method; Gas tungsten arc; Residual stress state; Neutron diffraction
Table 1
Chemical composition of HY-100 material (wt.%)
C Ni Cr Mn Si Mo Cu P S Al
0.18 2.8 1.4 0.28 0.26 0.18 0.17 0.044 0.029 0.013
sional transformations. The diffusional transformation Thermal expansion of the material was determined in
kinetics are complicated because the austenite composi- the range between room temperature and the solidus
tion in the HAZ is not homogeneous. This is due to the temperature. A high speed quenching dilatometer was
rapid heating and cooling associated with the welding, used to carry out these measurements. The mean coeffi-
and insufficient time available for austenite homoge- cient of thermal expansion (CTE) was calculated from
nization especially at those locations where the on-heat- the dilatometer data using the solidus temperature as a
ing peak temperatures are just above the trans- reference (Fig. 2). The solidus and the liquidus temper-
formation temperature. It is known that inhomoge- atures for the HY-100 steel composition determined
neous austenite transforms more readily to diffusional using Thermocalc™ are 1741 and 1776 K, respectively.
products because of the ease of nucleation of the equi- Other properties, such as elastic modulus, Poisson’s
librium phases. The diffusional transformation kinetics ratio, thermal conductivity, density, and latent heat
also depend upon the austenite grain size, since austen- were found elsewhere [12,13].
ite grain boundaries are preferred nucleation sites for An autogenous GTA welding process was used to
the austeniteferrite transformation. The grain size make a spot weld at the center of a HY-100 steel disk
dependence becomes more complicated in the weld (19 mm in height and 75 mm in diameter) with a
HAZ because of the variation of grain size with posi-
welding current of 320 A and a voltage of 15 V, using
tion in the HAZ. Solidification may introduce composi-
4 mm diameter electrode. Welding arc time was 5 s with
tion fluctuations in the FZ with respect to
argon shielding, followed by a 5 s postweld purge with
substitutional elements. It is known that substitutional
argon gas. During welding, the disk was set on a table
elements play a major role in the kinetics of diffusional
and no additional restraints were applied. Fig. 3 shows
transformations [8].
The objective of this study was to analyze the resid- the welding set-up.
ual stress state in spot welds made in HY-100 steel by The disk was sectioned along the z axis through the
an autogenous gas tungsten arc (GTA) welding process. center of the weld. One half was used for the ND
A finite element (FE) model for residual stress analysis residual stress measurements and the other half was
in spot welds was developed. The analysis included used for microhardness measurements and metallogra-
both the mechanical property changes and the volumet- phy. Redistribution of residual stresses due to section-
ric changes due to the austenite martensite transfor- ing of the disk was not taken into account.
mation. Effects of variation in mechanical properties at
high temperatures and the variation due to transforma-
tion of austenite to martensite on residual stresses were
evaluated. The results were compared with experimen-
tal measurements of residual stresses [9 – 11] in the
HY-100 steel weldment using the neutron diffraction
(ND) technique.
2. Experimental
3. Fe modeling
thermal conductivity above the melting temperature g − 1 was used [13]. The tube-air convection coefficient
was assumed. The thermal effects due to solidification was calculated for the natural convection to the air at
of the weld pool were modeled by taking into account 293 K. Radiative heat transfer was assumed at the top
the latent heat of fusion. To account for heat losses, disk surface with an emissivity coefficient e=0.2 [16].
both the radiative and convective heat transfer at the Two computational steps were carried out to com-
weld surface were modeled. plete the thermal welding simulation. In the first step,
The overall heat flux was calculated as: the heating/melting of the material was modeled by
applying the heat flux. Cooling transients were calcu-
F =pEI (2)
lated in the second step, when the heat source was
where p represents the efficiency factor, which accounts removed and the disk was cooled to room temperature.
for radiative and other losses from the arc to the The heating time was 5 s, whereas the disk was cooled
ambient environment, E is voltage, and I electric cur- until the ambient temperature was reached.
rent. Heat flux distribution at the disk surface was
defined by the equation: 3.2. Mechanical analysis
− 3(r/r0)2
f(r) =f1e (3)
Rate-independent elastic-plastic constitutive equa-
where r is the radial coordinate with the origin at the tions were constructed from the uniaxial tensile test
spot weld center. The constant r0 represents the charac- results of the BM at various temperatures. A linear
teristic arc dimension in the r direction within which transition between the yield strength of the BM at 573
95% of the energy is transferred. The constant f1 can be K, the beginning of the martensite transformation, and
derived by integrating the function f(r) at the material the yield strength of martensite at room temperature, as
surface: calculated from the hardness data, was assumed. Both
yield and ultimate strengths were reduced to 5.0 MPa at
3
f1 =pEI (4) the melting temperature to simulate low strength at
yr 20
high temperatures. Elastic modulus and Poisson’s ratio
To account for heat transfer due to fluid flow in the were also given as functions of temperature. The elastic
weld pool, the thermal conductivity was assumed to modulus was reduced to a small value (5.0 GPa) at high
increase linearly between the solidus temperature and temperatures.
3000 K by a factor of three [4] (Z. Feng, Edison The same FE mesh as in the thermal analysis was
Welding Institute, 1997, personal communication). Fig. used here, except for the element type and different
8 shows the specific heat and thermal conductivity used boundary conditions. The analysis is based on the
in the analysis. temperature history calculated in the thermal analysis
The latent heat of fusion was specified to model the which represents the input information. The movement
heat released during solidification. The value of 247 J in the axial direction of the bottom disk surface was
B. Taljat et al. / Materials Science and Engineering A246 (1998) 45–54 49
restrained to appropriately model the actual welding the heating transient, and depending on the cooling
conditions. time, the decision was made whether the point under-
A subroutine to ABAQUS code was developed that went the martensitic transformation or not. For each
accounts for changes in mechanical properties and vol- point that underwent the transformation the material
umetric changes due to the phase transformation. The properties of the martensitic phase were applied.
program consists of two parts: The subroutine also provided the basis for calculat-
The thermo-physical and mechanical properties of ing the volume increase due to martensitic phase devel-
the BM were used for the entire model during heating. opment. At this stage, the volume change was
The temperature history of each integration point in the approximated by the introduction of a modified coeffi-
model was observed during cooling. Depending on the cient of thermal expansion. The program distinguished
peak temperature that a particular point reached during between the heating and cooling cycle of each point and
50 B. Taljat et al. / Materials Science and Engineering A246 (1998) 45–54
martensite formation during cooling, were carried out. that the effect of higher yield strength due to formation
Three cases were analyzed, as described previously, and of martensite does not affect the residual stresses in the
the comparisons are shown in Fig. 13. The results show weld. The assumption of linear decrease in yield
strength with temperature has a notable effect on the
residual stresses, which suggests that accurate tempera-
ture dependent mechanical properties are required.
One can conclude that in such analysis it is not as
important to account for changes in the stress–strain
data due to the austenite martensite phase transfor-
mation, as it is to account for the volumetric changes in
the material caused by the phase transformation, and to
provide accurate, temperature-dependent stress–strain
data for the austenitic and ferritic phases.
There are several possibilities for improving the
present model. The thermal analysis is based on the
conduction model and the assumption of a Gaussian
shape for the heat source. The accuracy of calculated
residual stresses depend on the calculated temperature
transients; therefore, future efforts will concentrate on
implementation of the arc-pool interaction effect (the
effects of arc-to-weld pool evolution and free weld
surface deformation) into the current thermal model. A
convective heat transfer model, which includes the fluid
flow effects in the FZ, is also being developed. These
changes would improve the calculated temperature
transients which have an effect on the calculated resid-
ual stresses.
The current model relies upon experimentally deter-
mined transformation kinetics to incorporate the trans-
formation plasticity effects. However, it is of
fundamental interest to develop modeling capabilities
that will predict the austenite transformation kinetics
under non-equilibrium welding conditions. As a first
step, it is necessary to model the HAZ grain structure
for a given weld thermal cycle. Previous studies [17]
using 2D Monte Carlo simulations captured the effect
of thermal pinning in the HAZ of a 1/2Cr-Mo-V steel.
The simulations are now being extended to two-phase
microstructures where the second phase is an insoluble
inclusion. Both 2 and 3D Monte Carlo simulations
have been carried out in the past to predict the influ-
ence of inclusion volume fraction, size, shape, etc. on
the pinned grain size (unpublished research). The re-
sults indicate that there is a significant difference in the
nature of grain boundary pinning by inclusions between
2 and 3D. Hence, in order to make a realistic prediction
of grain size in the HAZ of materials containing second
phase particles, one has to resort to a 3D simulation
than a 2D simulation.
5. Conclusions
An experimental program was carried out to deter- Navy No. N000014-92-F-0063 under US Department
mine material properties of HY-100 steel. These include of Energy contract DE-AC05-96OR22464 with Lock-
tensile testing, measurement of the CCT curve, metallo- heed Martin Energy Research Corporation.
graphic observations, and hardness measurements. The
hardness measurements show increase in hardness in
the HAZ and the FZ, which indicate the phase trans- References
formation from austenite to martensite. The hardness
values were used to determine the yield strength in the [1] N.J. Smith, J.T. McGrath, J.A. Gianetto, R.F. Orr, Weld. J.
HAZ and the FZ. The metallographic observations Res. Suppl., March 1989, 112.
[2] V.S.R. Murti, P.D. Srinivas, G.D.H. Banadeki, K.S. Raju, J.
show the martensitic structure in the FZ.
Mater. Proc. Tech. 37 (1993) 723.
The ND technique was used to measure residual [3] J.P. Balaguer, Z. Wang, E.F. Nippes, Weld. J. Res. Suppl., April
stresses in the spot weld. The FE results are generally in 1989, 121.
good agreement with the ND results. The results show [4] K.W. Mahin, W. Winters, T.M. Holden, R.R. Hosbons, S.R.
a considerable effect of phase transformation on resid- MacEwen, Weld. J. Res. Suppl., September 1991, 245.
ual stresses, which cannot be neglected in such analysis. [5] J.B. Roelens, 8th Int. Conf. on Pressure Vessel Technology,
The volume change during the austenite martensite Montreal, Canada, July 1996.
[6] A. Jovanovic, A.C. Lucia, Int. J. Press. Vessels Pip. 22 (1986)
phase transformation tends to relieve high tensile
111.
stresses, and this is the reason that the magnitude of [7] B. Taljat, T. Zacharia, X.-L. Wang, J.R. Keiser, Z. Feng, M.J.
radial and tangential stresses in the HAZ and FZ is Jirinec, Approximate methods in the design and analysis of
about one-half the stress magnitude in the BM adjacent pressure vessels and piping components, PVP ASME 347 (1997)
to the HAZ. The results also show that the effect of 83.
volume change on residual stresses is more significant [8] J. Dobbs, Masters Thesis, University of Alabama at Birming-
ham, Alabama, 1987.
than the effect of higher yield strength caused by the
[9] Z. Feng, Y.Y. Zhu, T. Zacharia, R.J. Fields, P.C. Brand, H.J.
phase transformation. The results also indicate the sig- Prask, J.M. Blackburn, Proc. of 4th Int. Conf. on Trends in
nificance of temperature dependent stress – strain data Welding Research, Gatlinburg, TN, June 1995.
for accurate prediction of residual stresses. [10] P.C. Brand, H.J. Prask, J. Blackburn, R.J. Fields, T.M. Proctor,
1994 MRS Fall Meeting in Boston, Proceedings.
[11] H.J. Prask, C.C. Choi, in: C. Ruud (Ed.), Practical Applications
of Residual Stress Technology, ASM International, Metals Park,
Acknowledgements
OH, 1991, 87.
[12] ASME Boiler and Pressure Vessel Code, Section II — Materials,
The authors would like to thank Dr C.R. Hubbard Part D — Properties, 1995 ed., The ASME, New York, 1995.
and Dr X.L. Wang for reviewing the paper. The re- [13] Touloukian, Y.S., Ho, C.Y., (Eds.), Properties of Selected Fer-
search was sponsored in part by an appointment to the rous Alloying Elements, vol. 3, McGraw Hill, New York, 1981.
Oak Ridge National Laboratory Postdoctoral Research [14] J.R. Cahoon, W.H. Broughton, A.R. Kutzak, Met. Tran., July
Associates Program administered jointly by the Oak 1971, 1979.
[15] ABAQUS Users’ Manual, Version 5.5. Hibbitt, Karlsson and
Ridge Institute for Science and Education and Oak Sorensen, 1996.
Ridge National Laboratory. The research was also [16] E.R.G. Eckert, Heat and Mass Transfer, McGraw Hill, New
sponsored by the US Navy, Office of Naval Research, York, 1959.
under interagency agreement DOE No. 1866-E126-A1, [17] B. Radhakrishnan, T. Zacharia, Metall. Trans. A 26 (1995) 2123.