IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

1 od 11
4.3. RESIDUAL STRESSES AND DISTORTION IN WELDMENTS
A very trying practical problem which frequently arises in fabricating weldments is to be
certain the finished article will conform to the required dimensions and will have all portions
in proper contour. Inaccuracies in dimension or form easily develop because of the
unavoidable heat effects which accompany most of our welding processes, particularly
those involving localised fusion. Although these conditions are indeed unavoidable, they
can be minimised, and, in many cases, one condition can be used to counteract another. If
a few basic facts concerning the behaviour of metal during heating and cooling while under
restraint are understood, a little engineering skill can accomplish a great deal in securing
weldment accuracy.
There is an apparent lack of standard terminology for discussing the subject of
shrinkage and distortion with respect to welding. In literature on the subject, the terms
shrinkage, contraction and distortion, for example, are often used as if they were
synonymous. For purposes of explaining the problem with weldments, these terms should,
perhaps, carry distinct meanings. In order to establish a concept of shrinkage and distortion
control in weldments, several of these related terms will be discussed.
Although shrinkage is a term that will be used to describe the solid state phenomenon,
it should be recognised that it is often used to describe the decrease in volume that occurs
when a molten metal solidifies. Weld metal shrinks upon solidification, of course, but this
phenomenon has little to do with the distortion problem in weldments. During solidification,
as the atoms of iron in the melt assume fixed positions in the crystal lattice of growing solid
grains, the couplet of solid and liquid iron is very weak. Consequently, the weld metal
cannot exert any real stress upon adjacent base metal members. Upon completion,
solidification shrinkage has occurred in its entirety, and, although solidification shrinkage
may have resulted in some dishing or deformation of the face of the weld metal, it cannot
generate stresses capable of decreasing the overall size of the weldment or of pulling a
portion of the weldment out of shape.
Immediately following solidification, however, the cooling weld metal continues to
contract. This thermal contraction is the reverse of thermal expansion which occurs upon
heating. It is important to note that the contraction of solid metal during cooling can
generate stress, but that the maximum stress is limited to the yield strength of the weld
metal at the particular temperature. The details of the stress-producing mechanism will be
discussed shortly.
Distortion is deviation from a desired form, either permanent or temporary. Distortion
occurs as a result of welding because of stresses which develop in the weldment from
localised thermal expansion and contraction. However, it is entirely possible under some
circumstances to completely contain these stresses internally within an undistorted weldment.
Whether distortion occurs will depend upon (1) the magnitude of the welding stresses
developed (by localised thermal expansion and contraction), (2) the distribution of these
stresses in the weldment, and (3) the strength of the members upon which these stresses
act. When the final stress resulting from welding (called the contraction stress or the
shrinkage stress) is transmitted to a restraining portion of the weldment, that portion is at least
strained elastically. If the stress exceeds the member's yield strength, the yielding of the
structure can result in a permanent dimensional change and/or a distortion from its prewelded
form. Because the dimensional changes usually amount to a decrease in size, the term
shrinkage has been used to describe the net decrease in a particular dimension of a
weldment.
The ensuing discussion of distortion and shrinkage will make frequent use of the
following terms:
IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

2 od 11
Residual Stress - The internal stress which remains in the members of a weldment after
a joining operation, having been generated by localised partial yielding during the
thermal cycle of welding and the hindered contraction of these areas during cooling.
Residual stresses in a weldment represent a system of opposing stresses which
balance one another in the overall weldment, or in portions thereof.
Structure Stress - Stress arising from grain boundaries, crystal orientations, and phase
transformations, which exists in infinitely small volumes of metal.
Reaction Stress-Internal stress which could not exist if the weldment members being
joined were free to move or were isolated as free bodies without connection to other
parts of the structure.
Stress Concentration - The increase in stress level which develops at abrupt changes in
section as represented by sharp corners, notches, cracks and other surface or
subsurface irregularities. The condition which promotes the increase in stress is called a
stress raiser.

DEVELOPMENT OF STRESSES IN WELDMENTS
In fusion welding, joints in the materials are made by raising the surfaces to the melting
point, in most cases adding molten filler material, and then allowing the joint to cool as one
piece. The basic requirement is that the material must be raised, at least locally, to its melting
point. Heat may be supplied for this purpose by one of a variety of methods including the
electric arc, electrical resistance heating, and flame heating. Heating followed by cooling, an
integral part of the welding process, produces a number of effects.
Most materials expand on heating. If a piece is heated non-uniformly as in welding,
neighbouring elements of material try to expand by different amounts. This produces thermal
stresses. If these stresses exceed the yield point of the material at the relevant temperature,
plastic flow occurs, which eventually leads both to distortion (Fig. 10) and to the formation of
residual stresses in the workpiece (Fig. 11). Further residual stresses and distortion can be
caused by the shrinkage of the added filler material during cooling.

Fig. 10 Occurrence and mechanism of distortion from welding

IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

3 od 11

Fig. 11 Formation of residual stresses and typical residual stress pattern resulting from
welding

Because shrinkage and distortion are most troublesome in weldments made by fusion
joining processes, let us concentrate on the conditions which accompany all of the known
methods or processes that involve a localised heat source that moves along the joint. The
local rise and fall in temperature at any point along the joint as the heat source first
advances toward the point, passes it and moves on, develops stresses and causes the
changes in microstructure studied in preceding chapters. If the metals welded had zero
coefficients of thermal expansion, no stress would develop during welding, and shrinkage
and distortion would not present any problem. Actually, every metal expands and contracts
when heated and cooled. Temperature changes cause a change in volume, provided the
body is not restrained or prevented from changing its dimensions as the temperature
changes. Instead of a change in volume,. we generally speak for simplicity of a change in
one dimension only. When we say that steel at room temperature has a coefficient of
expansion of 6.7 x 10-6 per degree Fahrenheit, we mean that a 1-inch length increases to
1.0000067 inches if the temperature is raised 1 degree F and the body is free to expand in
all directions. Since, in welding, the parts are never free to expand or contract in all
directions, we must study the behaviour of steel attempting to expand and contract under
conditions of restraint.
We shall start by applying the bar analogy to a simple type of weld, realising that we do
not know enough about shrinkage stresses in more complicated welds to be able to predict
their distribution. One general rule, nevertheless, seems to apply to nearly all types of
welds made at room temperature and regardless of restraint; that is, the maximum tensile
shrinkage stress is close to the yield strength. If the tensile shrinkage stresses act in two or
three directions instead of in one direction, the stress at which the metal yields is raised,
and the maximum tensile shrinkage stresses likewise rise to values that may be
considerably above the yield strength observed in the conventional uniaxial tensile test.
The simple weld we shall study is shown in Fig. 1(a). The weld is deposited rapidly so
that we can be safe in assuming that the entire welded zone is at a high temperature when
the weld is completed. At the high temperature, the metal close to the weld attempts to
expand in all directions. It is prevented (restrained) by adjacent cold metal, which replaces
the press in the bar analogy. Being prevented from elongating, the metal close to the weld
is upset, Fig. 1(b). During cooling, the upset zone attempts to contract. Again it is restrained
by adjacent cold metal. As a result, the upset zone becomes stressed in tension (tensile
IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

4 od 11
shrinkage stresses). When the welded joint has cooled to room temperature, the weld and
the upset region close to it, Fig. 1(c), are under residual tensile stresses close to the yield
strength. To balance the tensile shrinkage stress at the edge, equilibrium conditions
demand that there must be a region of tensile shrinkage stress at the opposite unwelded
edge, and a region of compressive shrinkage stress between the two tensile zones. These
deductions agree in every respect with the experimental measurement of the residual
stresses.

Fig. 1 Distribution of stresses in a simple
welded edge-joint (a) Welded edge-joint; (b)
Welded area endeavors to expand during
welding, as shown by the dotted outline, but is
restrained by adjacent cooler metal, and is
upset; (c) Welded area endeavors to contract
during cooling, as shown by the dotted outline,
but is restrained by adjacent cooler metal, and
is elongated


Although it is not difficult to predict the distribution and magnitude of residual stresses
in a joint of the type shown in Fig. 1(a), the problem becomes more difficult with many other
types of joint for the reason that the shrinkage stresses are important, not only parallel to
the weld, but perpendicular to the weld as well. Consequently, we must rely heavily on
experimental measurements of stresses, remembering only the general rule that the
residual stresses are likely to be close to the yield strength if the majority of the welded part
during welding was several hundred degrees below the temperature at which the yield
strength virtually vanishes; that is, if the base metal originally was close to room
temperature.
Measurement and Calculation of Residual Stresses
One of the earliest methods of measuring residual stresses in a simple section, like a
bar or a beam, involved the machining away of successive layers of metal from one
surface, and noting the amount of bowing which developed. Computations of the curvature
of the section could produce close estimates of the residual stresses originally present.
Briefly, the removal of metal containing an internal stress would unbalance the system of
opposing stresses and the section would bow or distort until a new balance of stresses was
achieved. Many investigators have attempted to calculate the intensity and distribution of
residual stresses at welded joints because an accurate method would be of inestimatable
value in many phases of weldment design and fabrication. Distortion could be anticipated
and the residual stresses counteracted in the most efficient manner. Residual stresses
could be more favourably distributed, and in some cases might actually add to the useful
strength of certain members of the weldment which were required to bear the higher
service loads. Although the stress analyst has not shied from the complex calculations
which appear necessary to even start the development of a method, his attempts have
been seriously hindered by lack of detailed, accurate information on temperature
distribution during the welding of a joint. Our knowledge of the actual heat input is, of
IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

5 od 11
course, only approximate, and this, along with the variation in thermal conductivity and
changes in mechanical properties at high temperatures makes a reasonably accurate
computation of residual stresses quite difficult. Only in the simpler joints, where the thermal
conditions during welding have been very carefully assessed, has the analyst been able to
calculate values which are in reasonable agreement with measured values. One
comparatively simple system of residual stress calculation, known as Tall's "mirror method"
takes into account the incremental thermal stresses which arise in the course of making a
weld, and through a series of equations arrives at predicted stress values for specific
boundaries. However, much additional work is needed in this field before reliable
computations can be made to predict the residual stress in joints of actual weldments.

Fig. 2 Relationship between shrinkage stresses in arc-welded plates joined
without restraint. Line b-b' is centreline of weld

The residual stresses measured in an arc-welded butt joint between two plates are
shown in Fig. 2. In this case, one measurement method known as the "subdivision method"
is illustrated in Fig. 3. The residual stresses parallel to the weld are shown in Fig. 2 for the
central region of the plates. In agreement with predictions, the residual stresses parallel to
the butt weld at the centre are tensile and are close to the yield strength.


Fig. 3 Subdivision method of measuring shrinkage stresses created by
a single-vee groove weld between highly restrained plates
IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

6 od 11

At right angles to the weld the residual stresses are maximum at the weld. Regions
some distance from the weld are stressed less highly perpendicular to the weld than the
regions close to the weld. Figure 2 shows that in the centre of the plate the shrinkage stress
at right angles to the weld is tension; the edges are in compression to maintain equilibrium.
It is always fortunate to have compressive stress at the edges of any part because the
edges are the most sensitive regions from the standpoint of notch effects. The distribution
of stress shown in Fig. 2 is not always observed; for back-step procedures or frequent and
long interruptions to change electrodes may alter the distribution.
It should be remembered that the plates of the butt weld in Fig. 2 were under no external
restraint during welding. Had the plates been clamped rigidly during welding, additional tensile
stresses (the reaction stresses) would have arisen throughout the plates perpendicular to the
weld (Fig. 4). When the butt-welded plates are unclamped, the reaction stresses disappear.
Often, however, it is not so simple to avoid or remove reaction stresses, particularly if the
weld is the last to be made in a complicated rigid structure.
Fig. 4 Development of reaction
stress in a restrained weldment. Reaction
stresses are added to the shrinkage
stresses that would have appeared in the
plates without restraint, provided the sum
of the two does not exceed the yield
strength. Otherwise permanent
deformation occurs.

SHRINKAGE OF WELDMENTS
The overall or total dimensional decrease of a weldment member caused by
contraction of the solid weld metal and the adjacent base metal during cooling is difficult to
accurately predict. Yet, this information on shrinkage would be very valuable in scaling-up
preweld dimensions in order to have the finished weidment conform to size tolerances.
Often, a fabricator facing this problem of anticipating the amount of shrinkage in a
weldment will standardise a welding procedure, weld a trial joint, and measure the
shrinkage. This practical approach has many pitfalls if the fabricator does not understand
and control the many variables involved. Let us examine shrinkage of welded members in
different directions with respect to the weld axes, and study the techniques for measuring
and study the techniques for measuring and controlling this physical change.

Shrinkage Transverse to a Butt Weld
The shrinkage of a member, whether a plate, sheet, bar or structural section,
perpendicular to the long axis of an included weld usually is called transverse shrinkage. It
is primarily dependent upon the cross-section of the weld metal in the joint as illustrated in
Fig. 5.
The larger the cross-section, the greater is the transverse shrinkage. As indicated in
Fig. 5, the smaller cross-section weld area incorporated in a double-V or double-U
preparation, as compared with a single-V, decreases shrinkage. One investigator of arc-
welded joints found the transverse shrinkage to be approximately 10% of the average width
of the weld metal section.


IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

7 od 11
Shrinkage Longitudinal to a Butt Weld

Shrinkage parallel to a butt weld tends to reduce the width of the plate at the weld, Fig.
6. It seems reasonable to believe that this longitudinal shrinkage will be proportional to the
length of the weld. For example, a weld 2 feet long will shrink twice as much parallel to the
weld as a similar weld only 1 foot long, whereas both will produce the same amount of
shrinkage perpendicular to the weld. Longitudinal shrinkage is also a function of the cross-
section of the weld and the cross-section of the colder surrounding base metal which
resists the expansion and contraction forces of heated weld and base metal in and adjacent
to the joint.
Because the stress is not equally applied, not all of the area in the colder portion is
equally effective in resisting the shrinkage and upsetting action. Therefore, it is difficult to
calculate how much is effective
Where the area of the restraining plates is not more than 20 times the cross-sectional
area of the total weld, the following formula has been employed for predicting longitudinal
shrinkage:
Longitudinal Shrinkage = 0.025 (A
w
/A
p
), where A
w
= area of weld; A
p
= area of
restraining plates.

Fig. 5 Shrinkage transverse (perpendicular) to butt-joint in metal-arc welded plate.
Joints have different bevel preparations and have a root spacing of either 0.08-in.
or 0.12-in. prior to welding

IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

8 od 11

Fig. 6 Shrinkage longitudinal to the joint in welded butt joints


Fig. 7 Longitudinal shrinkage of butt welds as affected by cross-sectional area of
weld metal in joint

Where the area of the restraining plates is more than 20 times the area of the weld
metal, the chart provided in Fig. 7 may give a closer estimation of longitudinal shrinkage.
Again, proper peening can be used to counteract longitudinal shrinkage.
Each curve represents the variation of unit longitudinal shrinkage as a function of the
transverse cross-sectional area of the welded assembly for a given weld cross section. The
shrinkage tends to become stabilised when the sectional area of the assembly exceeds a
certain value, which is indicated by the dotted line.

DISTORTION OF WELDMENTS

Distortion in weldments is an exceedingly complex subject, but the factors which tend to
distort the form of a weldment are reasonably well understood. We know that the localised
area along which the arc or heat source passes is the starting point of our distortion problem.
IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

9 od 11
It is appreciated that the temperature differential between weld zone and unaffected base
metal is great and much localised expansion and plastic flow takes place here. We know that
restraint from external clamping and restraint from mass in the base metal itself will have an
influence upon the extent of plastic flow. Yet these are just a few of the many mechanical
engineering aspects of the physical changes in a weldment during and following the joining
operations. The relationship between the many factors involved in a common type of welded
joint is so complex that no method of analysis has been devised which closely predicts the
degree of distortion to be expected. Distortion occurs in all weldments, but can vary widely in
degree. Attaching a small clip to a massive block of steel by electron beam welding may be
claimed to have been accomplished without distortion. Yet, measuring instruments of great
precision invariable can find a minute change in form. In the extreme opposite, distortion from
welding with an oxyacetylene torch can make a thin sheet-metal cylinder look like a crumpled
trash can. Engineers recognise distortion when the degree of change from welding makes the
weldment form unsightly or unsuitable for its intended purpose. Distortion can be roughly
classified in three common kinds: angular distortion (Fig. 8), longitudinal bowing (Fig. 9), and
buckling.

Fig. 8 Angular distortion of a fillet welded corner joint and a single-V groove weld
in a butt joint

Angular Distortion

Angular distortion is the angular change in relative positions of members extending
from a weld area. This kind of distortion in a fillet welded corner joint and a butt joint is
illustrated in Fig. 8. When viewed as a transverse section, there is a shorter width of
contracting weld metal at the root of each weld than at the face. This difference in width
which must contract upon cooling is a major factor in angular distortion.
For single-bevel groove welds in butt joints and T sections unclamped during welding,
angular distortion is nearly proportional to the number of beads or layers, assuming all
layers are approximately the same size. That is, the greater the number of beads or layers
deposited in a joint, the larger will be the angular distortion.
A formula sometimes used to roughly compute angular distortion in terms of deflection
at the outer edge of flanges fillet welded to the web of plate girders is:
=
2
3 . 1
t
) w ( W 02 . 0

Where = deflection at edge of flange (in.); W = width of flange (in.); w = size of fillet
welds (in.); t = thickness of flange (in.). Angular distortion in general, is decreased (1) by
using the minimum amount of weld metal required to gain the desired joint strength, (2) by
depositing the weld metal in the fewest possible number of layers, (3) by avoiding as much
IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

10 od 11
as possible the weld profile with the very narrow root and the wide face, (4) by balancing
the amount of weld metal about the neutral axis of the joint, and (5) by pre-setting the
members at a slight angle opposite to that expected to develop as a result of welding.

Longitudinal Bowing

Longitudinal distortion or bowing of long members is caused by shrinkage stress which
develops at some distance from the neutral axis of the member. As shown in Fig. 9, the
amount of bowing distortion is determined by the magnitude of the shrinkage stress and the
resistance of the member to the bending indicated by its moment of inertia. A formula used
to calculate the bending resulting from longitudinal welding on a long, slim member is:
=
2
3 . 1
t
) w ( W 02 . 0

where: = resulting vertical movement (in.) A
w
= total cross-sectional area within the
fusion line of all welds (sq in.) d = distance between the centre of gravity of the weld group
and the neutral axis of the member (in.) L = length of member (in.) Assumed to be welded full
length, I=moment of inertia of member.
Fig. 9 Longitudinal distortion or bowing in
column member caused by
welding.
(A) During welding of
longitudinal butt joint on top
surface, this side expands and box
beam bows upward at centre.
(B) During cooling, upward
bow at centre disappears at some
elevated temperature.
(C) Upon reaching room temperature,
beam bows downward at centre
because top side undergoes
longitudinal shrinkage at weld.
Approximate vertical movement of
free end () can be calculated from
equation shown in text.


Control of longitudinal distortion is done by balancing welds about the neutral axis of
the member. When welds of unequal size must be made at uneven distances from the
neutral axis, the opposing stresses should be balanced by first depositing welds closest to
the neutral axis, and by making the welds that are further away from the neutral axis
smaller in size. Prebending or cambering the member in a bow opposite to that which will
develop from welding often is a practical recourse. Remember, the hot side will finally
become the short side.

IWE course Moduo 4- Fabrication, applications engineering

4.3. Residual Stresses and Distortion

11 od 11
Buckling

Thin sheet often buckles during welding whereas heavy plate does not. The buckling is
caused by the well-known inability of laterally unsupported sheet to resist compressive stress
without buckling (column failure).
Compressive shrinkage stress in regions near the weld accounts for the buckling.
Wandering sequence is a corrective for distortion. Instead of making all the welds in a strict
succession, welding alternately on opposite sides of a part often will reduce distortion. The
distortion of the weld on the opposite side may neutralise the distortion created by the first
weld. When welding is resumed on the first side, furthermore, the temperature has fallen
and less of the metal will enter the plastic range during subsequent welding.

References:

1.D. Buevac, Z. Markovi, D. uki, D. Toi, Metalne konstrukcije, Osnove prorauna i
konstruisanja, Graevinska knjiga, Beograd, 2007.

2.J. Tuek, Tehnika spajanja, Univerza u Ljubljani, Fakulteta za strojnitvo, Ljubljana, 2006.

3. J. Rigler, Varjene konstrukcije, Univerza u Ljubljani, Fakulteta za strojnitvo, Ljubljana,
1989.