Sie sind auf Seite 1von 7

Dave Supplement for 10/01 9/27/01 9:33 AM Page 246

Heat Generation in the Inertia Welding


of Dissimilar Tubes
A reduced thermal model is developed that accurately captures joint temperatures
and provides guidance in weld parameter development

BY V. R. DAV, M. J. COLA, AND G. N. A. HUSSEN

ABSTRACT. The transient thermal re- and the temperature profile in the region model is, however, sufficiently accurate
sponse in inertia welding is difficult to near the weld can have a significant im- to be used as a reduced-order thermal
capture analytically. Heat is generally pact on flash formation, heat-affected model for process optimization and pa-
dissipated over time scales of less than zones, and joint strength. Cooling rates rameter development in the inertia weld-
one second, an order of magnitude faster are closely related to joint temperature ing of tubes. A more accurate truth
than direct-drive friction welding. The profiles and they directly influence the model, such as a finite element model,
present work critically examines the na- residual stress state developed in the can then be used to further examine a
ture of the heat generation term through joint. The issue of residual stress becomes more limited set of interesting parameters
an analysis of experimental data. The more significant in dissimilar material and quantify flash evolution and residual
method presented here determines the joints. It is therefore desirable to have a stress formation.
heat generation term for the inertia weld- means of rapidly and accurately estimat- The first published analytical solution
ing of dissimilar tubes (tube thickness ing peak joint temperatures and cooling to the transient thermal history during fric-
small in comparison to radius so that ra- rates based on input parameters and rou- tion welding is generally attributed to
dial effects are neglected) solely based on tinely gathered inertia welding machine Rykalin, et al. (Ref. 1), although several of
m a ch i n e - g e n e rated data, namely the performance data. the same era researchers from the former
curve of angular speed vs. time and the The present model offers such esti- Soviet Union also discussed heat input
magnitude of material burnoff. A simple mates based on an analytical solution during friction welding (Refs. 24). The
approach to determining the heat alloca- and two proposed parametric represen- mathematics upon which this analytical
tion to both sides of the dissimilar joint is tations of the heat generation term. The solution is based appears, among other
proposed, and the resulting thermal model presented here is specifically for places, in the work of Carlslaw and Jaeger
problem is solved using an analytical tubular cross sections, and the analytical (Ref. 5). The model assumptions are semi-
method. The predictions are compared to models assume constant, but tempera- infinite solid; constant flux at the free sur-
actual thermocouple data from welds ture-averaged, material properties. The face for time th, then flux is turned off;
conducted under identical conditions, only required inputs are the measured zero initial temperature; and constant ma-
and are shown to be in good agreement. decay of the rotational speed, the mo- terial properties. The solution is given by
Although the method proposed in this ment of inertia of the flywheel, the total the following equation (Refs. 1, 5):
work does not replace more accurate nu- burnoff or reduction in length due to
merical analyses, it does provide guid- flash, and an assessment (or assumption) for 0 t th :
ance in weld parameter development. of how the burnoff is divided between the at x 2
This is demonstrated through the model- two tubes. p exp 4at
based scaling of weld parameters for dis- It is not the intent of this work to sup- 2q
T = o
similar tube welds over a range of tube plant more detailed numerical analyses k
diameters. of the inertia welding process, which are x erfc x
2 at
important in determining such mechani- 2
Introduction cal aspects of the process as residual
2q x
stress and flash formation. The present = o t ierfc
Previous Thermal Models of Friction
k 2 t
and Inertia Welding for x > t h :

In industrial friction and inertia weld-
ing production situations, it is not always t ierfc x
KEY WORDS 2 t
possible to conduct extensive instru-
2qo
(t th )
mented testing during which temperature
Inertia Welding T = (1)
data are gathered. Peak joint temperature
Friction Welding k

Thermal Models
ierfc x
V. R. DAV and M. J. COLA are with the Ma- Dissimilar Tubes

2 (t th )
terials Science and Technology Div., Los Reduced Order Model
Alamos National Laboratory, Los Alamos, Weld Parameters
N.Mex. G. N. A. HUSSEN is with the Mate- where k is the thermal conductivity, is
rials Science and Engineering Dept., Stanford the thermal diffusivity, x is the distance
University, Palo Alto, Calif.

246-s | OCTOBER 2001


Dave Supplement for 10/01 9/27/01 9:33 AM Page 247

from the weld interface, and qo is the dicted temperature distribution on the have produced a series of finite differ-
magnitude of the surface heat flux. total welding time. This total welding ence models. Weiss (Ref. 19) has also in-
The next significant contribution was time is a function of all the main process vestigated the residual stresses after
made by Cheng (Refs. 6, 7), who analyzed variables: initial rotational speed, thrust welding using the FEA approach. Fu and
both similar and dissimilar tubular joint pressure, moment of inertia, etc. Their Duan (Ref. 20) have more recently used
configurations. Cheng numerically solved model has good qualitative agreement the FEA approach to model the axial
the differential equation of heat conduc- with measured temperature values. pressure distribution in addition to the
tion (see Equation 2, which follows) with Johnson, et al. (Ref. 11), have also temperature field.
appropriate boundary conditions, and also noted the power dissipation curve for in- As mentioned earlier, this work is a re-
allowed the material properties to vary ertia welding is very different than that for duced order model of the inertia welding
with temperature (Ref. 6): friction welding. They have suggested a process. It is motivated by the need to
two-part curve: Stage I, corresponding to have simple, yet realistic, models that can
T T 1 T a more concentrated initial contact and be used for in-situ process monitoring and
+ U(t ) = k Stage II, a slower (relatively slower) control and for rapid parameter develop-
t x CP x x decay. They proposed the following func- ment and validation. It differs from previ-

P
CpA ( )
T4 To4
hP
Cp A
(T To ) (2) tional forms: ous works in that it attempts to more ac-
curately capture heat generation during
Stage I :q (t ) = qmax sin( t ) welding as a function of time by using
where A is the cross-sectional area, T o is data routinely gathered by the inertia
the ambient temperature surrounding the k Cp welding machine. This data is directly
tube, P is the outer perimeter of the tube, Stage II :q (t ) = Tmax (3) used as an input to the thermal simula-
t
Cp is the specific heat, is the density, h tion, which in this case is a reduced-order
is the film coefficient of heat transfer where qmax is the maximum power dissi- analytical model of heat conduction. As
(convective cooling), U(t) is the velocity pation and Tmax is the maximum interface such, this approach could also be used
of the melt front, is the emissivity, and temperature attained Fig. 1. A more re- with more sophisticated models for heat
is the Stefan-Boltzmann constant. cent multistage thermal model for direct- transfer, and would provide a reasonable
Cheng allowed for the existence of a melt drive friction welding was developed by estimate of heat generation without hav-
layer by incorporating a moving bound- Midling and Grong (Ref. 12), who pro- ing to explicitly model the combined
ary term. Several experimental studies posed various analytical forms for the thermomechanical problem. Also, this
have refuted the notion of melting during heating stage, steady-state condition, and approach is amenable to an on-line mon-
friction welding, such as the work of cooling stage based on continuous pla- itoring strategy that flags potentially de-
Weiss and Hazlett (Ref. 8), and this topic nar disc sources at the weld interface. fective welds in critical components and
will be revisited later in the present work. There are numerous finite element has the potential to alleviate the inspec-
As Wang (Ref. 9) pointed out, it is quite and finite difference models on both con- tion burden by reducing it to inspection
likely that softened material at tempera- ventional friction and inertia welding. for cause as opposed to inspecting every
tures near the melting point will be ex- These modeling efforts account for the component.
pelled as flash before melting can occur. heat generation term by examining the
Wang and Nagappan (Ref. 10) per- coupled thermomechanical problem to- Equipment and
formed a thermal analysis similar to that gether with an interfacial friction law. Experimental Procedure
of Cheng but for the inertia welding of This interfacial friction law or constitutive
steel bars. Their predictions showed the relation must account for frictional heat- The commercially pure niobium and
peak temperature to be less than the ing. Sluzalec (Refs. 13, 14) was one of the 316L stainless steel utilized in this study
melting point and that for inertia weld- first to use the finite element analysis were in the form of 1 in.-diameter tubes.
ing, the peak temperatures are achieved (FEA) approach for friction welding, and The Nb tube wall thickness was slightly
very quickly as compared to conven- Moal, et al. (Refs. 15, 16), have devel- thicker (0.125-in.) than the 316L tube
tional friction welding. Additionally, they oped an FEA model specifically for iner- wall (0.08 in.) to provide for greater forg-
noted a strong dependence of the pre- tia welding. Sahin, et al. (Refs. 17, 18), ing action during the upset stage. Prior to

Fig. 1 Schematic heat generation term after Johnson, et


al. (Ref. 11).
Fig. 2 Schematic showing thermocouple placement.

WELDING RESEARCH SUPPLEMENT | 247-s


Dave Supplement for 10/01 9/27/01 9:34 AM Page 248

the tube and extending along its axis to ification of Wang and Nagappans treat-
0.120 in. Fig. 2. The data acquisition ment, and a second approach based
system was a National Instruments Board purely on energy considerations. The first
Model AT-MIO-16-10 connected to a approach starts with Equation 5 and ig-
Pentium-class computer running Lab- nores radial variations on account of the
view SCXI 1120 software set at a gain of thin-walled nature of the tubular sections
20. A 10-kHz filter was used to record being joined. Furthermore, heat genera-
seven channels of thermocouple data at tion at the interface is apportioned to
200 Hz and store it for off-line analysis. each material according to the expres-
Based on a thermocouple rise time analy- sion that applies to heat generation in an
sis conducted by Sawada and Nishiwaki infinite composite solid (Ref. 5):
(Ref. 21), it is expected the thermocouple
rise time could range from 10 to 50 ms, Q1 + Q 2 = Q total
but is certainly expected to be less than
100 ms. The total bond time was less than Q1 k 1 2 k1 1 C1
0.25 s for the welds made in this study, so = = (6)
Q 2 k2 1 k 2 2 C 2
the thermocouple may underestimate the
peak temperature by, at most, 6% (as- The speed curve was modeled by the
suming 50-ms rise time and peak tem- following empirical relationship:
perature of ~1000C occurring 0.1 s after t
Fig. 3 Macrosection of the as-welded nio- n
the start of the bond). Data was sampled (t ) = o exp m

bium-to-stainless steel joint. (7)
from the start of the weld cycle until the
weld had cooled below 260C. A macro-
section of a completed weld between where m and n are constants, and the
inertia welding, the tubes were sectioned niobium and stainless is shown in Fig. 3. characteristic time is defined as the time
into 3-in. lengths and the faying surfaces required for the rotational speed to decay
were machined while flood cooled in The Heat Generation Term to 10% of its original value at the start of
isopropyl alcohol. Derived Directly from the weld. The data in Fig. 4 shows the ac-
Inertia welds were produced using an Machine Measurements tual speed curve from a weld and the em-
MTI Model 90B inertia welding system. pirical form chosen to represent it. For the
Initial emphasis was placed on determin- Rykalin, et al. (Ref. 1), have deter- case shown here, normalized speed
ing parameters capable of producing joints mined the heat input during friction [(t)/o] is shown vs. time. The weld con-
that could sustain bending through 90 deg. welding is given by the following: ditions were o = 1488 rpm; Io = 5.19
Once suitable starting parameters were de- q(t ) = const. p r (t ) (4) lbm-ft2; P = 8330 lbf. The nonlinear, least-
veloped, welds were made at a surface ve- squares-fit parameters are m = 2.931 and
locity of 393 ft/min and axial force of Wang and Nagappan (Ref. 10) as- n = 1.718. The heat input is then assumed
8330 lbf while maintaining a constant mo- sumed the product of the friction coeffi- to have the following form:
ment of inertia of 5.19 lbm-ft2. After weld- cient and the normal pressure p re- Q( t) = A o (t ) (8)
ing, and prior to testing, each joint was He mains a constant. By using this
leak checked, achieving a minimum leak assumption an empirical fit to the an- where Q is the total heat input in watts,
rate of 1x1010 atm.-cc/s. gular speed curve and a heat balance is the speed curve as represented by
Thermal cycle data were recorded by they were able to derive an expression for Equation 7, and Ao is a constant. The con-
attaching 0.010-in.-diameter Type K the heat generation of the following form: stant is evaluated based on conservation
(Chromel-Alumel) thermocouples to the
q(t ) = const.r (t ) (5) of energy. Before this can be done, it is
outside of the stainless steel tube. Seven recognized that some of the energy avail-
thermocouples were attached in 0.020- The present work takes two ap- able to the joint from the flywheel will be
in. increments starting at the joint end of proaches to modeling heat input: a mod- used to heat and expel flash. Therefore,

Fig. 4 The angular speed curve as modeled empirically by Equa- Fig. 5 Comparison between predicted and actual thermal pro-
tion 7. files on the stainless side of the weld.

248-s | OCTOBER 2001


Dave Supplement for 10/01 9/27/01 9:34 AM Page 249

Fig. 6 Error as defined by Equation 14 shown as a function of Fig. 7 The effect of variations in thermomechanical properties
time for Models 1 and 2. on model predictions.

the effective energy conducted into the The second method of determining The thermal profiles can now be eval-
workpiece will be less than the initial en- the heat flux involves direct considera- uated with the aid of Duhamels theorem,
ergy of the flywheel. The approach taken tion of the dissipation of rotational ki- as applied to the case of 1-D heat con-
in this work was to examine the joint after netic energy by the weld. For any given duction in a semi-infinite solid with a
the weld and make a determination as to time during the weld, it can be generally time-varying flux at the free surface, as
the amount of material expelled from said the following expression relates the follows:
1 t
T(x,t ) = q(t )
each side. For example, in the case of dis- energy dissipated by the weld to the loss
similar welds between Nb and 316L in kinetic energy of the flywheel: k 0
stainless, the flash was expelled almost
E weld (t ) =
1
[ ]
I I (t )
1 2
entirely in the Nb. The inertia welding
2
1(1) x 2 d
2O O 2O exp (15)
machine tracks the reduction in length 4
during the weld, and, therefore, it is pos- Therefore, it is immediately observed that
sible to know the burnoff directly from the power dissipation in the weld is given where q(t) is the time-varying surface
machine measurements. Then it is as- by the following: flux.
sumed the flash carries off an amount of The heat flux q is easily derived from
Q( t) = IO(t )
d
energy equal to (12) the expressions for total power shown in
dt Equations 8 and 13 by considering the
E flash = B AC C TMAX (9)
This ignores stored elastic energy in the cross-sectional area and the fraction of
where B is the total reduction in length, tooling or inertia welding machine and energy entering into a particular side of
or burnoff; A C is the cross-sectional area also ignores other energy losses such as the joint, as follows:
of the tube; is the density; C is the av- machine friction and grip slippage, i.e.,
erage specific heat over the temperature energy loss at the workpiece/tool inter- Q total (t )
q1(t ) =
1
range TMAX; and TMAX is the maximum face. It turns out this method was inde- (16)
temperature rise. pendently discovered by Dr. H. A. Nied,
AC k 2 2 C2
1+
Since the maximum temperature at- Sr., of General Electric Co., and was k 1 1 C1
tained is not known a priori, an iterative brought to the authors attention through
procedure must be used. The peak joint Ref. 22. Using the same model for the ro- For the two proposed heat inputs, the
temperature is first estimated, the energy tational speed as shown in Equation 7, thermal profiles will now be compared
lost to flash is then evaluated, the result- the total power dissipation in the weld is for a stainless steel-to-niobium weld. The
ing thermal profile is calculated, and the assumed to be the following: weld parameter data and assumed ther-
m n
Q(t ) = CO
process is repeated until the maximum mal properties of the materials are shown
predicted temperature matches the esti- in Table 1. The resulting thermal profiles
mate. The constant AO can now be eval- are shown in Figure 5 and are compared
t t
n n 1
uated: to the actual data at the position of the
exp 2 m (13)
weld interface, i.e., x = 0. Model 1 refers
Q( t) = A O (t ) to the heat flux as specified by Equation
As before, the constant Co takes into 8, whereas Model 2 refers to the heat flux
0 Q(t )dt = (EO Eflash ) account the fact that some of the energy as given by Equation 13.
dissipated in the weld must go into heat- A measure of how well the two mod-
so
ing and expelling the flash. Co is calcu- els represent the data can be deduced by
AO =
(EO E flash ) lated in an entirely analogous manner to considering the error as a function of
Ao and is given by the following: time:
0 (t )dt Error(t) M(t) Y(t) (17)
1 E
where EO = IO 2O (10) CO = 1 flash (14) where M(t) is the predicted temperature at
2 EO
t and Y(t) is the measured temperature at t.

WELDING RESEARCH SUPPLEMENT | 249-s


Dave Supplement for 10/01 9/27/01 9:34 AM Page 250

Table 1 Weld Parameter Data and Assumed Thermal Properties Table 2 Comparison Between Subscale
and Fullscale Inertia Weld Parameters
Quantity Value
(all thermophysical data from Ref. 20) Parameter Value
Initial Rotational Speed 1487 rpm (1500 nominal) Full-scale Inertia 5.19 lb-ft2
Moment of Inertia 5.19 lb-ft2 Subscale Inertia 3.32 lb-ft2
Axial Welding Force 8330 lb Full-scale Initial Speed 1570 rpm
Average Specific Heat for Nb over 2731273 K 0.29 J/g/K Subscale Initial Speed 1675 rpm
Average Thermal Conductivity for Nb over 2731273 K 0.59 W/cm/K Full-scale Area 0.231 in.2
Average Specific Heat for 316 SS over 2731273 K 0.56 J/g/K Subscale Area 0.168 in.2
Average Thermal Conductivity for 316 SS over temperature 0.21 W/cm/K Bond Pressure (held constant 25854 lb/in.2
range 2731200 K as area changes)
Fraction of total power going to 316 SS side (equation 16) 0.445
Fraction of total power going to Nb side 0.555

The nominal bond as the part size changes, then this sug-
pressure is usually not gests an additional scaling law (in addi-
altered as the size is in- tion to Equation 18):
creased, although the
bond force is adjusted to rpPq = (const.) (21)
make the bond pressure
constant. In the present In the present work, it was assumed
treatment, the effect of that p = q = 1 in Equation 21 for purposes
pressure is considered of testing the newly proposed scaling
and modifications to the law. First, welds were produced between
bond pressure with size titanium and niobium using only Equa-
are proposed. The effect tion 18, i.e., keeping the initial bond en-
of increasing pressure is ergy per unit area a constant. The inter-
tracked by a close ex- facial bond pressure was also held
amination of the power constant. The subscale tube diameter
Fig. 8 A comparison between the power dissipation during the dissipation curves, and was 0.75 in., and the full scale was 1 in.
bond for the full-scale bond and the subscale bond conducted at joints made in compo- The bond parameters for both bonds
various bond forces. nents of different sizes based purely on Equation 18 and con-
are considered equiva- stant interfacial bond pressure are shown
lent when their respec- in Table 2.
The evolution of the error is shown in tive power dissipation curves resemble The full-scale bonds made under
Fig. 6 for the two models under consid- one another. these conditions easily passed a destruc-
eration. It is clear Model 2 better repre- The behavior of the hot, highly tive bend test (samples were bent after
sents the data at short times, whereas worked interfacial layer that forms during flash was removed and a bend angle of
Model 1 seems to be more suitable at the inertia weld can generally be mod- greater than 45 deg was achieved before
long times. Another important factor that eled using the following phenomenolog- failure at the joint), whereas the subscale
has been ignored in the treatment thus far ical form (Ref. 15): samples failed immediately upon being
is the effect of temperature-dependent subjected to bending loads (essentially
material properties. This effect is shown = (const.)(V)pPq (19) zero bend angle). Clearly the assumed
in Fig. 7 by changing the assumed mate- scaling law based purely on Equation 18
rial properties for conduction through the where is the shear stress, V is the rel- and a constant interfacial bond pressure
stainless steel. ative slip velocity, and P is the interfacial did not work. The bond pressure was
bond pressure. then modified based on Equation 21 with
Heat Input Model as a Basis The slip velocity for thin-walled tubes exponents p and q equal to 1. The origi-
for Selection of Weld Parameters is given by v = r, where r is the average nal bond pressure based on constant
radius. The expression for the tangential bond pressure was 5978 lbf. The bond
It will now be shown that the pro- shear (radial shear effects ignored) stress pressure predicted by Equation 21 with p
posed heat generation term can be used then becomes the following: = q = 1 is 7963 lbf. The full-scale bond
to select weld parameters as part size is was conducted at a bond pressure of
[ ]
(t ) = (const.) (t ) r P q
p
changed, i.e., the parameter-scaling 5978 lbf. Based on the data in Fig. 8, as
problem. The basic assumption in inertia the subscale and full-scale power curves
welding is that the energy per unit area of or start to resemble one another, the result-
the joint must be kept constant as the (t ) ing bonds are expected to have compa-
weld parameters are scaled for larger or = (const.) r p P q (20) rable bond quality. The sample bend tests
[ ] ()
p
smaller diameters. This means the fol- t also suggest this is the case. This means
lowing, with E specified by the initial fly- the power dissipation curve can provide
wheel kinetic energy: The quantities on the left-hand side of valuable guidance in the selection of
Equation 20 are directly related to the bond parameters as the size of the joint
E1 E heat input and power dissipation during changes. The assumption of p = q = 1 is
= 2 (18)
A1 A 2 the weld. Therefore, if it is assumed the somewhat arbitrary, but importance
power dissipation is to be held constant should not be attached to the specific

250-s | OCTOBER 2001


Dave Supplement for 10/01 9/27/01 9:34 AM Page 251

values of these parameters. The central 0.2 m. Therefore, the Reynolds Number possible to establish a process window
message of this study is that the power of this flow is the following: based solely on in-situ measurements,
dissipation characteristic as a function of which would be the first step toward a
time is a good means of transferring weld U x (0.93 m / s) (0.2 m) 100% quality-assured methodology for
Re =
( )
parameters from one part size to a inertia welding (no postprocess inspection
smaller or larger size.
v 1 106 m 2 / s required). Accounting for temperature-de-
5 pendent material properties can enhance
= 1.86 10 (23) the thermal predictions in this work, and
Inertia Welding Heat Generation
which is within the laminar regime. The such simulations are in progress.
and the Possible Existence
corresponding film thickness then be-
of a Liquid Interlayer
comes Acknowledgments
There has been and continues to be v x
some debate about the possible exis- This research was funded under DOE
U
tence of a liquid layer during inertia or contract No. W-7405-ENG-36. The au-
friction welding. Cheng (Refs. 6, 7) al-
lowed for the existence of such a molten =
(1 10 6
)
m 2 / s (0.2 m )
thors also wish to thank Ann M. Kelly for
her valuable assistance in metallography,
layer in his modeling work. Several ex- (0.93 m / s) Dr. David F. Teter for his valuable sug-
gestions and electron microscopy work,
perimental studies by Squires (Ref. 23),
Weiss, and Hazlett (Ref. 8), and Hasui, et = 464m (24) and Dr. R. D. Dixon for his critical review
al. (Ref. 24), did not find evidence of of the manuscript.
From microstructural investigation,
melting. Wang and Nagappan (Ref. 10)
there is no evidence of a 400+ micron-
predicted the peak temperatures would References
wide melt layer that, if it did exist, would
be below the melting point based on their
be most conspicuous and easily de-
modeling work of inertia-welded steel 1. Rykalin, N. N., Pugin, A. I., and
tectable. Furthermore, even on submi-
bars. Wang (Ref. 9) points out available Vasileva, V. A. 1959. The heating and cooling
cron-length scales, no evidence of melt -
metallurgical investigation does not sup- of rods butt-welded by the friction process.
ing was detected. At the interface, there
port the existence of a liquid film at the Svarochnoe Proizvodstvo (Welding Produc-
was an intermetallic reaction layer con-
interface, and torque measurements do tion): 4252.
sisting of a Nb-Fe-Cr intermetallic com-
not show a disruption or sudden drop 2. Vill, V. I. 1959. Energy distribution in
pound that was approximately 200 nm
that may be expected on account of a liq- the friction welding of steel bars. Svarochnoe
thick. If we assume this layer was formed
uid interlayer. Proizvodstvo (Welding Production): 3151.
by liquid-phase reaction, we can ap-
To further analyze the possibility of 3. Geldman, A. S., and Sander, M. P. 1959.
proximate the thickness of the laminar
melting during the inertia bonds made in Power and heating in the friction welding of
boundary layer to be the width of this re-
this work, the possibility of a fluid layer t h i ck - walled steel pipes. S va ro c h n o e
action zone. In that case, using Equation
subjected to shear is considered. A pre- Proizvodstvo (Welding Production): 5361.
22, the equivalent metal viscosity would
dicted melt layer thickness will now be 4. Zadegson, R. I., and Voznessenskii, V. D.
have been approximately 2 x 10-10 cP,
derived for such a layer by invoking sim- 1959. Power and heat parameters of friction
which is a physically unrealistic number
ple hydrodynamic reasoning. The lami- welding. Svarochnoe Proizvodstvo (Welding
for molten metals. It is, therefore, rea-
nar boundary layer thickness is specified Production): 6370.
sonable to assume there was no melting
by the following: 5. Carlslaw, H. S., and Jaeger, J. C. 1959. Con-
during the welding process.
duction of Heat in Solids. Oxford, U.K.:
v x Clarendon Press.
(22) Conclusions and Future Work
U 6. Cheng, C. J. 1962. Transient temperature
distribution during friction welding of two
where x is the position along the wall in In this study of the heat generation similar materials in tubular form. Welding
the direction of flow, v is the kinematic term during inertia welding of dissimilar Journal 41(12): 542-s to 550-s.
viscosity, and U is the free-stream tubes, the following was demonstrated: 7. Cheng, C. J. 1963. Transient temperature
velocity. 1) The temperature profile during in- distribution during friction welding of two dis-
The average velocity during the speed ertia welding can be well represented by similar materials in tubular form. Welding
curve decay is obtained by finding the av- simple analytical solutions that directly Journal 42(5): 233-s to 240-s.
erage value of the speed curve shown in use machine-generated data. 8. Weiss, H. D., and Hazlett, T. H. 1966.
Equation 7 using fit parameters as de- 2) The proposed forms of the heat gen- The role of material properties and interface
scribed in Fig. 4. This results in an aver- eration term can be used to provide guid- temperatures in friction welding dissimilar
age velocity of 0.93 m/s. The dynamic ance for parameter development when metals. ASME Paper 66-MET-8.
viscosity of molten iron ranges from 5 to attempting to transfer successful weld pa- 9. Wang, K. K. 1975. Friction welding.
10 centipoise (a centipoise, cP, is equal rameters to varying part diameters. WRC Bulletin no. 204.
to 103 kgm1s1) over a range of temper- 3) It is unlikely a fluid layer was gen- 10. Wang, K. K., and Nagappan, P. 1970.
atures (Ref. 25), and this equates to a erated during the inertia welds discussed Transient temperature distribution in inertia
kinematic viscosity of 6 X 107 1.2 x 106 in this work. welding of steels. Welding Journal 49(9): 419-s
m2/s. The distance x is the maximum dis- The heat generation terms represented to 426-s.
tance the part may rotate during the iner- by Equations 8 and 13 could also be used 11. Johnson, P. C., Stein, B. A., and Davis,
tia weld, which in this case is the total cir- as part of an in-situ process monitoring R. S. 1966. Inertia Welding. Unpublished
cumferential travel during the decay in methodology. If the heat generation term, technical report, Caterpillar Tractor Co.
rotational velocity. Again using Equation the effective bond time (time required for 12. Midling, O. T., and Grong, O. 1994. A
7 for the conditions described in Fig. 3, angular speed to drop to 10% of its initial process model for friction welding of Al-Mg-
the part makes a total of 2.5 rotations, so value), the material burnoff, and the Si alloys and Al-SiC metal matrix composites
the maximum possible travel distance is burnoff rate are all monitored, it may be I: HAZ temperature and strain rate distrib-

WELDING RESEARCH SUPPLEMENT | 251-s


Dave Supplement for 10/01 9/27/01 9:34 AM Page 252

ution. Acta Metallugica et Materialia 42(5): A. Z. 1996. Friction welding of Al-Al, Al-steel, Report SR9821, Inertia Welding of IN718 and
15951609. and steel-steel samples. J. of Mat Eng. and Per. 1045 Steel: Numerical Modeling and Experi-
13. Sluzalec, A. 1985. Thermoplastic effects 5(1): 8999. ments.
in friction welding analyzed by finite element 18. Sahin, A. Z., Yilbas, B. S., Ahmed, M., 23. Squires, I. F. 1966. Thermal and me-
method. The Effects of Fabrication Related and Nickel, J. 1998. Analysis of the friction chanical characteristics of friction welding
Stresses on Product Manufacture and Perfor- welding process in relation to the welding of mild steel. British Welding Journal 13(11):
mance, TWI Conference Proceedings. The copper and steel bars. J. of Mat. Proc. And 652657.
Welding Institute, Abington, U.K. Paper 8. Tech. 82: 127136. 24. Hasui, et al. 1968. Experimental stud-
14. Sluzalec, A. 1990. Thermal effects in 19. Weiss, R. 1998. Residual stresses and ies on friction welding phenomena. Trans.
friction welding. Int. J. of Mech. Sci. 32(6): strength of friction welded ceramic/metal Nat. Res. Inst. Metals Japan 10(4): 5371.
467478. joints. Welding Journal 77(3): 115-s to 122-s. 25. Geiger, G. H., and Poirier, D. R. 1973.
15. Moal, A., Massoni, E., and Chenot, J-L. 20. Fu, L., and Duan, L. 1998. The coupled Transport Phenomena in Metallurgy. Reading,
1990. A finite element modeling for the iner- deformation and heat flow analysis by finite Mass.: Addison Wesley Publishing Co.
tia welding process. Conference on Computa- element method during friction welding.
tional Plasticity: Fundamentals and Applica- Welding Journal 77(5): 202-s to 207-s.
tions I, pp. 289300. Swansea, U. K.: 21. Sawada, T., and Nishiwaki, N. 1991.
Pineridge Press. Response of a thermocouple to transient tem-
16. Moal, A., and Massoni, E. 1995. Finite perature changes in a metal to which it is at-
element modeling of the inertia welding of tached. Int. J. Mech. Sci 33(7): 551561.
two similar parts. Engineering Computations 22. Communication with Tom McGaughy
12(6): 497512. of the Edison Welding Institute (EWI),
17. Sahin, A. Z., Yilbas, B. S., and Al-Garni, 10/20/99, who distributed a copy of EWI CRP

Preparation of Manuscripts for Submission


to the Welding Journal Research Supplement
All authors should address themselves to the and symbols are important considerations in processing
following questions when writing papers for submission a manuscript for publication. For welding terminology,
to the Welding Research Supplement: the Welding Journal adheres to ANSI/AWS A3.0-94,
Why was the work done? Standard Welding Terms and Definitions.
What was done? Papers submitted for consideration in the Welding
What was found? Research Supplement are required to undergo Peer
What is the significance of your results? Review before acceptance for publication. Submit an
original and one copy (double-spaced, with 1-in. margins
What are your most important conclusions?
on 8 12 x 11-in. or A4 paper) of the manuscript. Submit
With those questions in mind, most authors can the abstract only on a computer disk. The preferred
logically organize their material along the following format is from any Macintosh word processor on a 3.5-
lines, using suitable headings and subheadings to in. double- or high-density disk. Other acceptable
divide the paper. formats include ASCII text, Windows or DOS. A
1) Abstract. A concise summary of the major manuscript submission form should accompany the
elements of the presentation, not exceeding 200 words, to manuscript.
help the reader decide if the information is for him or her. Tables and figures should be separate from the
2) Introduction. A short statement giving relevant manuscript copy and only high-quality figures will be
background, purpose and scope to help orient the published. Figures should be original line art or glossy
reader. Do not duplicate the abstract. photos. Special instructions are required if figures are
3) Experimental Procedure, Materials, Equipment. submitted by electronic means. To receive complete
4) Results, Discussion. The facts or data obtained instructions and the manuscript submission form,
and their evaluation. please contact the Peer Review Coordinator, Doreen
5) Conclusion. An evaluation and interpretation of Kubish, at (305) 443-9353, ext. 275; FAX 305-443-
your results. Most often, this is what the readers 7404; or write to the American Welding Society, 550 NW
remember. LeJeune Rd., Miami, FL 33126.
6) Acknowledgment, References, and Appendix.
Keep in mind that proper use of terms, abbreviations

252-s | OCTOBER 2001