MSK580

SVEISE- OG INSPEKSJONSTEKNOLOGIER
2021 HØST

THERMO-MECHANICAL MODELLING OF WELDING FOR

Title
DEFORMATION AND RESIDUAL STRESS PREVENTION

Submission Date OCTOBER 18, 2021

Supervisor PROF R.M. CHANDIMA RATNAYAKE

Submitted By IZWAN BIN AHMAD 261972

Candidate Number 4701

 Table of Contents

1. WELDING MECHANIC ................................................................................................................ 3

1.1 Process of development of ‘residual stress’ and ‘inherent strain’ in a welded joint ............... 3
2. MEASUREMENT AND PREDICTION OF RESIDUAL STRESSES USING INHERENT
STRAINS ................................................................................................................................................ 9
2.1 Measuring approaches of residual stresses in 2D models ....................................................... 9
2.2 Measuring approaches of residual stresses in 3D models ..................................................... 10
3. MECHANICAL SIMULATION OF WELDING ........................................................................ 12
3.1 Heat flow modelling: convection and radiation .................................................................... 12
3.2 Changes in material properties with temperature.................................................................. 13
3.3 An approach for solving the heat conduction problem ......................................................... 14
3.4 Solving mechanical problems in welding using 3-bar model ............................................... 15
3.5 Solving thermal visco-elasto-plastic problem in welding using 3-bar model ....................... 16
4 USE OF FINITE ELEMENT METHOD FOR MODELLING AND ANALYSIS OF WELDING
DEFORMATION AND RESIDUAL STRESSES ............................................................................... 18
4.1 Incremental Method for Nonlinear Problem ......................................................................... 19
4.2 Boundary Conditions ............................................................................................................ 19
4.3 Properties Change with Temperature .................................................................................... 19
4.4 Modelling the Heat Source ................................................................................................... 20
4.5 Flow of Analysis for Welding Deformation and Residual Stress ......................................... 20

Figure 1.1 Welding directional illustration for butt weld joint ............................................................... 3
Figure 1.2 Idealized 3 bar models ........................................................................................................... 3
Figure 1.3 Stress vs Strain development ................................................................................................. 3
Figure 1.4 Thermal and stress history of free bar ................................................................................... 4
Figure 1.5 Thermal histories with three different maximum temperatures............................................. 5
Figure 1.6 Histories of thermal stress and strain of a bar fixed at both ends .......................................... 6
Figure 1.7 Configuration difference between fixed bar and movable rigid body constraint .................. 8
Figure 2.1 Uniaxial stress models ........................................................................................................... 9
Figure 2.2 Measuring procedure of 2D residual stresses (flat-plate model) ........................................... 9
Figure 2.3 Thick-plate model ................................................................................................................ 10
Figure 2.4 Measuring procedure of 3D residual stresses (thick-plate model) ...................................... 10
Figure 3.1 Through-thickness temperature distribution ........................................................................ 12
Figure 3.2 Three-cylinder model .......................................................................................................... 12
Figure 3.3 CCT diagram ....................................................................................................................... 14
Figure 3.4 Through-thickness layers..................................................................................................... 14
Figure 3.5 Bars in through-thickness direction ..................................................................................... 14
Figure 3.6 Equivalent 3-cylinders model .............................................................................................. 14
Figure 3.7 Three-bar model subjected to external and thermal loads ................................................... 16
Figure 4.1 Example of meshed finite element model ........................................................................... 19
Figure 4.2 Flowchart of process to solve problems using simulation ................................................... 20
1. WELDING MECHANIC
1.1 Process of development of ‘residual stress’ and ‘inherent strain’ in a welded joint
For a sufficiently long welded joint, same temperature distribution is moving along the welding
direction except at the start and end of the workpiece. This situation is called quasistatic. Due to uneven
temperature distribution, the thermal expansion in the joint behave in different magnitude, and this
temperature rise are the potential source of the residual stress depending on the magnitude.

Considering butt weld joint, the high heat source will cause thermal expansion at the heated area, and
the low temperature/unexpanded area will act as the restraint. To model the thermal effect, the
sufficiently long longitudinal section is the constraint against thermal expansion and transverse
direction are presumably free. Therefore, we shall model the mechanical behaviour in the longitudinal
section.

Figure 1.1 Welding directional illustration for butt weld joint

Referring to Idealized model in Figure 1.2 Idealized 3 bar models, development of residual stresses and
inherent strain in a welded joint can be simply modelled using 3 bars model. The central bar, Bar C,
which experiencing the high temperature and the other 2 sides bar, Bar S, with high volume and low
temperature.

Figure 1.2 Idealized 3 bar models

For simplicity, several assumptions have been made for this model:

1. The temperature is constant along each respective bar

2. Material is assumed to be elastic and perfectly plastic
3. Material is not temperature dependent, so thermal expansion coefficient α is constant

Figure 1.3 Stress vs Strain development

Thermal Elastic-Plastic Analysis
As per Figure 1.2, the thermal elastic plastic analysis was be performed on the 3-bar model under 3
different restraint configurations with several heating and cooling conditions as in Table 1.1. The
outcome of the analysis is that we figured the residual stress was produced by plastic strain induced
under thermal history, and this is called inherent strain. Therefore, the residual stress can be predicted
if the inherent strain is exerted on the stress-free joint under standard elastic analysis, provided that
inherent strain data of the weld joint.

The bars will be analysed under 3 different restraint configurations as in Table 1.1 below:
Table 1.1 Analysis of the idealized 3 bar model

No Constraint (1) Fixed at Both Ends (2) Movable Rigid Body (3)

Bar C : 0°C → T° Bar C : 0°C → T°C Bar C : 0°C → T°C

Bar S : Same as bar C Bar S : Fixed at 0°C Bar S : Fixed at 0°C
No thermal or residual stress is Thermal deformation of bar C is Thermal expansion and shrinkage
induced in bar C as the three bars completely resisted by bar S of bar C is resisted by bar S
are simultaneously subjected to through the movable rigid body. through the movable rigid body.
the same temperature change and Depending on the heating Under a given thermal history,
it can be elongated or shortened temperature of bar C, plastic strain plastic strain may be induced in
freely under any change of may be produced, which becomes bar C, which causes residual
temperature the source of residual stress. stress.
Preheating of workpiece before Welding of small workpiece Welding of large workpiece that
welding fixated by clamps are not fixated

No Constraint (1) Configuration

The displacement observed then the bars was heated u, disappeared as it cooled. This is because all the
bars are able to move freely

𝑢 = 𝛼𝑇𝐿

This can be illustrated as if the whole workpiece was heated at the same time (E.g.: during preheating
of the workpiece before welding commence)

Figure 1.4 Thermal and stress history of free bar

Fixed at Both Ends (2) Configuration

Bar C in Table 1.1 is resisted by the fixed end. Therefore, the heating and cooling cycle through Tmax
caused different elastic-plastic behaviour at each temperature reference. Beyond yielding temperature
of the bar C, the plastic strain and residual stress will develop according to equation
𝜎
𝑇𝑌 = 𝛼𝐸𝑌

Where: Ty Yield Temperature

σY Yield Stress
α Thermal Expansion Coefficient
E Young’s modulus

To model the thermal-stress-strain relations at different reference temperature, the analysis was divided
into 3 section as in Figure 1.5 and Table 1.2.

Figure 1.5 Thermal histories with three different maximum temperatures

Table 1.2 Thermal histories with three reference

Low-Temp heating (1) Medium-Temp heating (2) High-Temp heating (3)

0°C → Tmax ≤ TY → 0°C 0°C → TY ≤ Tmax ≤ 2TY → 0°C 0°C → 2TY ≤ Tmax → 0°C
Plastic strain: No Plastic strain: Yes (heating) Plastic strain: Yes (heating and
Residual stress: No Residual stress: Yes (cooling) cooling)
Residual stress: Yes (sum of both)
Low-Temp heating (1)

At low temperature cycle, the process is still in elastic range, thus the result of elastic strain and thermal
strain is zero as the bar is fixed at both end and any elongation and shrinkage are restricted

Medium-Temp heating (2)

At medium temperature cycle, bar C was heated beyond yield temperature 𝑇𝑌 but not exceeding 2𝑇𝑌 ,
and this heating process induced additional strain (compressive plastic αΔT) due to plastic flow. No
further stress increase beyond the yield stress. Therefore, total strain consists of thermal strain, plastic
strain, and elastic strain. Since the bar is fixed at both end and any elongation are restricted, the thermal
strain converts into compressive plastic strain.

During cooling by TY, the bar is shrinking elastically and no increase in plastic strain. Resulting stress
is zero. Further cooling to T = 0°C, the shrinkage strain developed due to the rigid wall, leading to
elastic tensile strain, and thus tensile stress formed as 𝜎 = 𝛼𝐸(𝑇𝑚𝑎𝑥 − 𝑇𝑌 )
But as the temperature reaches 0°C, no thermal strain in the bar, and only plastic strain remains 𝛥𝜀 𝑝 ,
thus leading to residual stress. So, this plastic strain is the inherent strain, 𝜀 ∗ = 𝛼(𝑇𝑚𝑎𝑥 − 𝑇𝑌 )

High-Temp heating (3)

This is applicable to welding analysis as the workpiece in welding will be heated beyond the yield
temperature. At high temperature heating to Tmax ≥ 2TY, the thermal strain changes into compressive
plastic strain. During the cooling by TY, the compressive yield stress in the bar diminished and the
process is elastic. Further cooling by 2TY causing more shrinkage, but the constraint from the wall causes
this shrinkage into tensile yield stress. While still within tensile yielding, further cooling to T = 0°C, no
increase in stress, and the shrinkage strain turns to plastic strain.
Thus, we can see that the residual stress was developed in two occasion, during heating from TY to Tmax,
and during cooling by Tmax – 2TY to T = 0°C.

Figure 1.6 Histories of thermal stress and strain of a bar fixed at both ends

Therefore, we can conclude the production of residual stress depends on the magnitude of heating. The
resulting plastic strain from thermal strain induced the residual stress in heating and cooling stage
depending on the temperature. For welding analysis, the residual stress can be calculated using standard
elastic analysis by imposing inherent strain on stress free workpiece.
Plastic Strain, Inherent Strain, Inherent Deformation and Residual Stress of bar C summarized in Table
1.3 below.
Table 1.3 Strain, Deformation, and Stress of Bar C Under Three Different Heating Processes

Heating Process Low Temp Medium Temp Limit Temp High Temp

Maximum temp, Tmax Tmax ≤ TY TY ≤ Tmax ≤ 2TY Tmax ≤ 2TY 2TY ≤ Tmax

Plastic strain in heating

– – α(Tmax – TY) – εY – α(Tmax – TY)
process, 𝛥𝜀1𝑃

Plastic strain in cooling

– – – α(Tmax – 2TY)
process, 𝛥𝜀2𝑃

Inherent strain, 𝜀 ∗ – – α(Tmax – TY) – εY – αTY = – εY

Inherent deformation,
– – α(Tmax – TY)L – εYL – αTYL = – εY
𝛥𝐿∗ = 𝜀 ∗ 𝐿

Residual stress, 𝜎 – αE(Tmax – TY) ≤ σY σY σY

It should be noted that if base temperature, T°C is not 0°C, the yield temperature TY shall be modified
to TY = (σY / αE) + T°C

Movable Rigid Body (3)

For the analysis of the 3 bars restricted by movable rigid body. The situation is basically similar to
restraint configuration (2), but the important takeaway here is that the unheated bar S presumably having
large in cross sectional area, A0, gives the additional coefficient (A+A0)/A0. This coefficient represents
the degree of restraint exerted by the large body of bar S to bar C. Mathematically, as A0 approaches
infinity, the coefficient becomes 1.
We will see how the movable rigid body effect the stress strain on bar C under 3 thermal process, Tmax
as in Table 1.2 in previous configuration.

Low-Temp heating (1) 0°C → Tmax ≤ TY → 0°C

As we heat bar C to Tmax, the bar elongates, but restricted by bar S though the movable rigid body
wall. Therefore, compressive stress and tensile stress developed in Bar C and Bar S, and the movable
rigid body moves by u. Since only Bar C is heated, thermal strain exists in Bar C. Since the Tmax ≤ TY,
the bars react elastically. Here, TY will be expressed as:
𝐴 + 𝐴0 𝜎𝑌
𝑇𝑌 =
𝐴0 𝐸𝛼
Where: Ty Yield Temperature
σY Yield Stress
α Thermal Expansion Coefficient
E Young’s modulus
A Cross sectional area of Bar C
A0 Cross sectional area of Bar S combined, A0/2 +A0/2
 All the bars are subjected to same displacement, u. Since only Bar C is heated, only Bar C develop
thermal strain, and these strains are function of u. As the heating temperature reached T = TY, Bar C
approach yield stress, thus u = uY. The uY and TY here are affected by the cross-sectional area ratio
coefficient, (A+A0)/A0. Note that how it is different than the configuration where all the bars are fixed
at both ends. Figure 1.7 below explains.

Figure 1.7 Configuration difference between fixed bar and movable rigid body constraint

During cooling to 0°C, no residual stresses developed since the process is within elastic range.

Medium-Temp heating (2) 0°C → TY ≤ Tmax ≤ 2TY → 0°C

As the temperature reached beyond yield temperature, compressive stress in bar C approaches yield
temperature and additional thermal strain induced. As the bar yielded, no more increase in stress in the
bar and the thermal strain as compressive plastic strain in Bar C remains.
During cooling by TY from Tmax, the bars S unloaded elastically, and the rigid body shrink by Δu, which
Δu determined from equilibrium where all net forces are zero. Some plastic strain is cancelled by
thermal strain thus the total strain and stress in the bar are zero.
Further cooling to T = 0°C, the bar unloaded elastically. However, the remining plastic strain the bar C
and at 0°C, no more thermal strain exists. Thus, plastic strain, produced during heating, is the inherent
strain, and this is the source of residual stress. If the maximum heating temperature goes beyond 2TY,
tensile plastic strain will start to develop during cooling stage, which we will see in the next thermal
case.

High-Temp heating (3) 0°C → 2TY ≤ Tmax → 0°C

Similar as before, as we heat the bar C beyond the TY, the Bar C will yield and no more increase in
stress from 𝜎𝑌 . And the additional thermal energy will convert into compressive plastic strain stored in
the bar.
As we cooled it. By TY, the compressive yield stress vanishes. Further cooling by amount of TY, the
tensile stress begins to develop as the shrinkage of bar C is resisted by bar S through the rigid wall.
Then continues cooling to T = 0°C , no more thermal strain remains. But as the tensile stress maxed out
and started yielding, thus no increase in stress, and tensile strain developed in the bar C.
We can see, the plastic strain developed in two occasion, cooling (tensile strain) and heating
(compressive strain). And these two strains are the inherent strains, and thus leading to residual stress
from elastic-plastic behaviour.
2. MEASUREMENT AND PREDICTION OF RESIDUAL
STRESSES USING INHERENT STRAINS
Same as previous, by using the three-bar model, we will analyse the residual stress in 2D and 3D. We
assume that effective inherent strain, which is produced by the differences in inherent strain among the
3 bars, exists in bar 1 and bar 2.
According to inherent strain method, in general, residual stress in a component can be measure by using
the following steps
a) Cut/Slice the component
b) Measure the inherent strain
c) Compute the residual stress distribution by elastic analysis using the inherent strain data

2.1 Measuring approaches of residual stresses in 2D models

Referring to Figure 2.1 and Figure 2.2 below which correspond to butt weld joint of a thin plate analysis,
the method are as follows:
1. Place strain gauges on all three bars
2. Cut all bars at the same time to relax the stresses
3. Measure the released strain, 𝑚𝜀1𝑒 , 𝑚𝜀2𝑒 , 𝑚𝜀3𝑒

Figure 2.1 Uniaxial stress models Figure 2.2 Measuring procedure of 2D residual stresses
(flat-plate model)

The measured stresses can then be solved by using equation below, where we assume that effective
inherent in bar 3, 𝜀3∗ , as the base value. So, we have the effective inherent strain as;
𝜀1∗ 𝜀3∗ 𝜀1∗ − 𝜀3∗ 𝜀1∗
{𝜀 ∗ } − {𝜀3∗ } = {𝜀2∗ } − {𝜀3∗ } = {𝜀2∗ − 𝜀3∗ } ≡ {𝜀2∗ }
𝜀3∗ 𝜀3∗ 𝜀3∗ − 𝜀3∗ 0

Then we can use the equation;

𝑒 ∗
𝑚𝜀1 1 −2 1 𝜀1∗
{ 𝑚𝜀2𝑒 } = [ 1 −2] {𝜀2 }
𝑒 3
𝑚𝜀3 1 1 0
And solve for the effective inherent strain using the measured strain;
𝜀∗ −2 −1 𝑚𝜀1
𝑒
{ 1∗ } = [ ] { 𝑒}
𝜀2 −1 −2 𝑚𝜀2
Then, we can calculate the residual stresses by;
{𝜎} = [𝐷]{𝜀 𝑒 } = [𝐷][𝐻 ∗ ]{𝜀 ∗ }

This is the for the best care where the measured strain doesn’t contain any error. In case of measurement
error, we need to get the most probable value of the inherent strain, {𝜀̂ ∗ } = [𝜀̂1∗ 𝜀̂2∗ ]𝑇 by;
𝑒
𝑚𝜀1
𝜀̂1∗ −1 0 1 𝜀𝑒 − 𝜀𝑒
{ ∗} = [ ] { 𝑚𝜀2𝑒 } = { 𝑚 3𝑒 𝑚 1𝑒 }
𝜀̂2 0 −1 1 𝑒 𝑚𝜀3 − 𝑚𝜀2
𝑚𝜀3

Consequently, the residual stresses can be calculated by;

{𝜎̂} = [𝐷]{𝜀̂𝑒 } = [𝐷][𝐻 ∗ ]{𝜀̂ ∗ }

2.2 Measuring approaches of residual stresses in 3D models

3D analysing can be used for thick plate workpiece such as 3 bar model in Figure 2.3.

Figure 2.3 Thick-plate model Figure 2.4 Measuring procedure of 3D residual stresses (thick-plate model)

Assuming that there will be no error in the measured value, the residual stress in the component can be
measure by using the following steps
1. Place strain gauge in Bar 1
2. Cut Bar 1 and measure the strain development 𝑚𝜀1𝑒 ,
1 𝜀∗
a) where 𝑚𝜀1𝑒 = 3 [−2 1] { 1∗ }
𝜀2
3. Place strain gauge in Bar 2
4. Cur bar 2 and measure the strain development
𝜀𝑒 1 −1
a) From elastic strain { 2𝑒 } = 2 { } 𝜀2∗ ,
𝜀3 1
1 ∗ 1
b) We know that 𝑚𝜀2 = − 2 {𝜀2 }, and 𝑚𝜀3𝑒 = 2 {𝜀2∗ }
𝑒

c) Which the true inherent strain can then be solved by {𝜀 ∗ } = [𝐻 ∗ ]−1 { 𝑚𝜀 𝑒 }

𝜀∗ −3/2 −1 𝑚𝜀1𝑒
d) Thus, sets of equation 2(a) and 4(b) can be solved by { 1∗ } = [ ]{ }
𝜀2 0 −2 𝑚𝜀2𝑒
5. The calculated true inherent strain {𝜀 ∗ }, can then be used to calculate residual stress by;

{𝜎} = [𝐷]{𝜀 𝑒 } = [𝐷][𝐻 ∗ ]{𝜀 ∗ }

𝐸 0 0 −2 1 1
1
Where [𝐷] = [ 0 𝐸 0 ] and [𝐻 ∗ ] = 3 [ 1 −2 1 ]
0 0 𝐸 1 1 −2
In the case where measurement error present, we shall estimate the most probable value of
effective inherent strain, 𝜀̂ ∗ where {𝜀̂ ∗ } = [𝜀̂1∗ 𝜀̂2∗ ]𝑇 .

{𝜀̂ ∗ } = ([𝐻 ∗ ]𝑇 [𝐻 ∗ ])−1 [𝐻∗ ]𝑇 { 𝑚𝜀 𝑒 }

Thus
𝜀𝑒
𝜀̂1∗ 1 −3 −1 1 𝑚 1𝑒
{ ∗} = [ ]{ 𝜀 }
𝜀̂2 2 0 −2 2 𝑚 2𝑒
𝑚𝜀3
∗
So 𝜀̂ can be solve for most probable value of residual stresses, {𝜎̂} by
{𝜎̂} = [𝐷]{𝜀̂𝑒 } = [𝐷][𝐻 ∗ ]{𝜀̂ ∗ }
3. MECHANICAL SIMULATION OF WELDING
3.1 Heat flow modelling: convection and radiation
Modelling of heat convection and radiation in welding can be visualize easily using water flow in three
cylinders connected by tube. The flow of heat in thickness direction can be shown by the water flow
between the cylinder through the connected tube. For simplicity, the flow in the length of the cylinder
will be neglected.

Figure 3.1 Through-thickness temperature distribution Figure 3.2 Three-cylinder model

Referring to Figure 3.2 above, the water flow downward at cylinder A and C symbolizes the heat loos
due to transfer. While heat flow between the cylinder symbolizes the heat flow due to convection. The
TA, TB, and TC indicate the magnitude of the temperature at the virtual top, mid and bottom later
respectively.

From this basic model, we can visualize:

Table 3.1 Heat flow phenomena and the respective 3-cylinder model

Heat Flow Phenomena Illustration

Change of Temperature Distribution

with time without heat transfer.

Insulated, heat is not allowed to flow

out

Change of Temperature Distribution

with time with heat transfer.

Allowed to cool down to room

temperature. Mid layer temperature
appears to be high just before complete
cool-down
 Heat Flow Phenomena Illustration

Influence of heat input.

Temperature high when heat input is

high.

Influence of speed of heat input.

High power over short period of time

creates steep temperature distribution.
Relevant in welding understanding.

Influence of plate thickness.

Influence of heat transfer.

Hight heat transfer coefficient causes

the heat loss to the surrounding to be
faster, causing faster temperature drop.

Influence of heat conduction.

The high heat conductivity causes the

through thickness temperature
distribution to be small.

3.2 Changes in material properties with temperature

Material properties namely heat conductivity, specific heat, and density effect the heat flow. These
values vary depending on the type or composition of the metal. The heat flow will in turn effect the heat
expansion coefficient, thus influencing the distortion and residual stresses during welding.
During welding, the heat conductivity, specific heat, density, heat transfer coefficient, thermal
expansion ratio, Young’s modulus, tensile strength and yield strength, changes throughout the
temperature cycle. Heating beyond 800°C (phase transformation to austenite complete, depends on
carbon content), the specific heat becomes large due the latent heat, same goes for thermal expansion
ratio. Austenite is the soft phase with low Young’s modulus and yield stress. At around 600°C and
700°C (mechanical melting temperature), or above A1 the material loses it resistance to deformation,
thus the yield strength reaching zero around these temperatures.
 During welding, region close the heat
source the base metal may be heated
above A1 (HAZ), causing the base
metal to transform into austenite,
which is stable at high temperature.
During cooling, depending on the
cooling rate (refer CCT diagram in
Figure 3.3), hard and brittle martensite
may form which results in welding
crack and low toughness point.
Common practice to measure cooling
rate in post-welding is to measure the
time taken for the temperature to drop
from 800°C to 500°C.

Figure 3.3 CCT diagram

3.3 An approach for solving the heat conduction problem

Three-cylinders model can be used again for solving the heat conduction problem, and the heat flow
will be visualize using fluid flow. Referring to Figure 3.5, thickness of each cell is 1/3 of the plate
thickness h.

Figure 3.4 Through-thickness layers Figure 3.5 Bars in through-thickness Figure 3.6 Equivalent 3-cylinders
direction model

Using finite element method in matrix form and simplification of the derivation, temperature
distribution can be calculated using equation:

𝑑𝑇
[𝐶] { } + [𝐾]{𝑇} = {𝑄}
𝑑𝑡
𝑑𝑇𝐴
1 0 0 𝑑𝑡
𝑐𝜌𝑠ℎ 𝑑𝑇 𝑑𝑇𝐵
Where [𝐶] = ( ) [0 1 0] =
3 𝑑𝑡 𝑑𝑡
0 0 1 𝑑𝑇𝐶
{ 𝑑𝑡 }
1 −1 0 1 0 0
3𝜆𝑠
[𝐾] = ( ) [−1 2 −1] + 𝛽𝑠 [0 0 0]
ℎ
0 −1 1 0 0 1
Heat conduction heat transfer
 𝑇𝐴 1
{𝑇} = {𝑇𝐵 } [𝑄] = 𝛽𝑠 𝑇0 {0}
𝑇𝐶 1
Where c specific heat capacity
s cross sectional area
ρ density
λ thermal conductivity
β heat transfer coefficient (convection and radiation)
T0 Initial temperature

The unknown variable in {dT/dt} and {T} can be calculated numerically using Crank-Nicolson or
central difference method. Finite element calculation for temperature distribution in welding, in a case
for central difference method, works by splitting the time into small time increments and solve for {T(t
+ Δt)}. Smaller time increments mean faster convergence time. However, there’s a limit to the size of
time increment Δt.

𝑐𝜌 ℎ 2
Δ𝑡 < ( ) ( )
𝜆 2
From the equation above, upper limit of the time increment Δt is inversely proportional to thermal
diffusivity (λ/cρ) and proportional to (h/2) size of the cell.
𝑐𝜌 ℎ
Δ𝑡 < ( )( )
𝛽 2
From equation above, upper limit of the time increment is proportional to the (cρ/β) and thickness h.
To sum it up, time increment upper limit is determined by the heat conduction and heat transfer. There
is no lower limit for central differencing. However, too small of time increment will only results in
longer computation time with negligible accuracy increment.

3.4 Solving mechanical problems in welding using 3-bar model

As we previously discussed, the thermal cycle during welding induces deformation, stress induced by
expansion and shrinkage, residual stress, and creep deformation. Most significant change are the
thermal expansion and shrinkage, temperature-dependent material properties, and plastic and creep
deformations.
For practicality and simplification in the FEM, specific mechanical problem in welding simulation will
be classified according to dimension. Either in 1D, 2D, or 3D. This is done to reduce the computation
time. The 3-bars model that we’ve been using is classified into 1D problem. Though it’s only useful to
get the understanding, but not for quantitative analysis.
Other real world welding problems can also be idealized into 2D problems such as plane stress, plane
strain, and plane deformations. For an example in the case of butt welding of 2 thin plate, stress in
thickness direction can be neglected. Anther properties of a problem that can be specified into 2D
modelling is those that are axis symmetric.
Table 3.2 Idealization of welding problem into 2D

Problems Axis in 2D Picture

Girth weld of pipes Neglect circumferential deformation

Spot welding Neglect thickness direction

Fillet weld of thick plate Neglect strain in welding direction

3.5 Solving thermal visco-elasto-plastic problem in welding using 3-bar model

Similarly, 3 bar model is used to simplify the calculation for stress and deformation of a structure. This
can be done by solving the strain-displacement (1), stress-strain (2), and equilibrium equation (3) under
suitable boundary condition.

Figure 3.7 Three-bar model subjected to external and thermal loads

Table 3.3 Relations of stress, strain, and equilibrium equation

Relations Bar a Bar b Bar c

(1) Stress-Displacement Not heated, remain elastic, Heated, thus Δεb Not heated, remain
thus experience elastic contains thermal, elastic, thus experience
strain Δεa elastic, plastic, and elastic strain Δεc
creep strain increment

∆𝑢
∆𝜀𝑎 = ∆𝜀𝑏 = ∆𝜀𝑐 =
𝐿
(2) Stress-Strain ∆𝜎𝑎 = 𝐸∆𝜀𝑐 = ∆𝜎𝑐 ∆𝜎𝑏 = 𝐷 𝑠𝑡𝑟𝑎𝑖𝑛 ∆𝜀𝑏 + ∆𝜎𝑐 = 𝐸∆𝜀𝑎 = ∆𝜎𝑎
𝐷𝑡𝑒𝑚𝑝 ∆𝑇 +
𝐷𝑡𝑖𝑚𝑒 ∆𝑡

(3) Equilibrium 𝐴(∆𝜎𝑎 + ∆𝜎𝑏 + ∆𝜎𝑐 ) = ∆𝐹

All the stress increments in the bars shall balance with the force applied at the top
By eliminating the stress and strain from equilibrium equation, we get:
∆𝑢
𝐴 {(2𝐸 + 𝐷 𝑠𝑡𝑟𝑎𝑖𝑛 ) ( ) + 𝐷 𝑡𝑒𝑚𝑝 ∆𝑇 + 𝐷 𝑡𝑖𝑚𝑒 ∆𝑡} = ∆𝐹
𝐿
𝐴
(2𝐸 + 𝐷 𝑠𝑡𝑟𝑎𝑖𝑛 ) ( ) ∆𝑢 = ∆𝐹 + 𝐴𝐷 𝑡𝑒𝑚𝑝 ∆𝑇 + 𝐴𝐷 𝑡𝑖𝑚𝑒 ∆𝑡
𝐿
𝐾∆𝑢 = ∆𝐹 ∗
Where;

𝐿𝑜𝑎𝑑 𝐼𝑛𝑐𝑟𝑒𝑚𝑒𝑛𝑡, ∆𝐹 ∗ = ∆𝐹 + 𝐴𝐷 𝑡𝑒𝑚𝑝 ∆𝑇 + 𝐴𝐷 𝑡𝑖𝑚𝑒 ∆𝑡

𝐴
𝑆𝑡𝑖𝑓𝑓𝑛𝑒𝑠𝑠, 𝐾 = (2𝐸 + 𝐷 𝑠𝑡𝑟𝑎𝑖𝑛 ) ( )
𝐿
𝑇𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑖𝑛𝑐𝑟𝑒𝑚𝑒𝑛𝑡, ∆𝑇
𝑇𝑖𝑚𝑒 𝑖𝑛𝑐𝑟𝑒𝑚𝑒𝑛𝑡, ∆𝑡
Therefore, displacement ∆𝑢, Strain ∆𝜀, and Stress ∆𝜎 can be calculated by rearranging the equations
above.
Since we consider the time gradual increment of stress and deformation as time increase incrementally
∆𝑡, total displacement, strain, and stress can be computed as below:
𝑢(𝑡 + ∆𝑡) = 𝑢(𝑡) + ∆𝑢
𝜀𝑎 (𝑡 + ∆𝑡) = 𝜀𝑎 (𝑡) + ∆𝜀𝑎 , 𝜀𝑏 (𝑡 + ∆𝑡) = 𝜀𝑏 (𝑡) + ∆𝜀𝑏 , 𝜀𝑐 (𝑡 + ∆𝑡) = 𝜀𝑐 (𝑡) + ∆𝜀𝑐
𝜎𝑎 (𝑡 + ∆𝑡) = 𝜎𝑎 (𝑡) + ∆𝜎𝑎 , 𝜎𝑏 (𝑡 + ∆𝑡) = 𝜎𝑏 (𝑡) + ∆𝜎𝑏 , 𝜎𝑐 (𝑡 + ∆𝑡) = 𝜎𝑐 (𝑡) + ∆𝜎𝑐

Where (𝑡 + ∆𝑡) term indicate the total changes over time.

4 USE OF FINITE ELEMENT METHOD FOR MODELLING
AND ANALYSIS OF WELDING DEFORMATION AND
RESIDUAL STRESSES
Simulating welding problem by Finite Element Method is only sensible to be use with a computerized
calculation by computer. Thus, the accuracy of the simulation is only as accurate as the data we feed to
the model. This is because, FEM itself is a method of discretizing/meshing a large or complex body
into a smaller cell and calculating target parameters at the nodes and centre of the cells by applying
appropriate formula and boundary conditions. Therefore, hand calculation is almost impossible and of
no benefit.

As all other solution using FEM, certain type of cells must be chosen for specific problem. Such as bar
element (1D), quadrilateral and triangular (2D, tetrahedron and hexagonal (3D). Meshing technique
also must be monitor by using smaller cells at the target area and bigger cells at the less important area
in order to minimize computation load and modelling time. Once the complex structure has been broken
down to simple shaped cells, the basic governing formula and equations that was explained in the
preceding chapter can be apply in the simulation.

For modelling of welding, essentially there are two parts that need to be solved.
1. Computation of temperature distribution with respect to time (heat conduction analysis)
2. Computation of deformation, strain, and stress under transient heat (plastic stress analysis)

Therefore, two important concepts for FEM will be used for welding simulations:

Table 4.1 Basic Form of Equations applicable for Welding

Mechanical Components Applicable Problems Equations

Spring • Stress analysis
• Steady heat conduction 𝑲𝒖 = 𝒇
• Static electric field
Spring & damper

• Transient heat conduction 𝒅𝒖

𝑪 + 𝑲𝒖 = 𝒇
𝒅𝒕

Stress analysis will be done according to Variational Principle. The term 𝒇𝒖 indicate potential energy
𝟏
of the external force 𝒇, while the term 𝑲𝒖𝟐 indicate the stored strain energy in the spring. Thus, this
𝟐
equation indicates the total potential energy of the mechanical system. We can also notice that the
equation is governed by the displacement 𝒖 value. The potential energy is the lowest when 𝒖 is
minimum. So, from understanding of solid mechanic, this concept is also known as principle of
minimum potential energy.

𝟏
𝚷(𝒖) = 𝑲𝒖𝟐 − 𝒇𝒖
𝟐
 Figure 4.1 Example of meshed finite element model

As previously mention, the object subjected for FEM will be meshed. So, there will always be multiple
elements involved. The basic concept of FEM is superposition of the stiffness matrix (consisting the
strain-displacement, stress-strain, and equilibrium equation) between the elements. Then, appropriate
boundary condition will be applied (geometrical and force)

4.1 Incremental Method for Nonlinear Problem

Welding are nonlinear process. Temperature, displacement, stress, and strain changes in complex
manner throughout the time of welding. Therefore, incremental method will be used, in which each
properties will be calculated at each time step 𝚫𝒕 based on the initial/previous time 𝒕.

Therefore, we will have: 𝑻(𝒕 + 𝚫𝒕) = 𝑻(𝒕) + 𝚫𝒕

𝒖(𝒕 + 𝚫𝒕) = 𝒖(𝒕) + 𝚫𝒖
𝜺(𝒕 + 𝚫𝒕) = 𝜺(𝒕) + 𝚫𝜺
𝝈(𝒕 + 𝚫𝒕) = 𝝈(𝒕) + 𝚫𝝈

Since there are relations between temperature and stress, it will be solved simultaneously. For doing
that in the FEM software, thermal analysis throughout the cycle (heating and cooling) will first be done
and the temperature distribution during each time step will be stored in a separate file. Then at stress
analysis can be done by extracting the temperature distribution file and to calculate the stress and strain
at each temperature and each time step from heating until cooling.

4.2 Boundary Conditions

As what we can see from previous chapter, each model has its own boundary condition, such as bar
fixed at both ends, constricted by movable rigid body, and so on. So, in the FEM model, same thing
must be applied. Otherwise, displacement cannot be determined, and rigid body motion will occur.

4.3 Properties Change with Temperature

Then we have to consider the change in material properties with temperature. In which we have:
1. Elastic behaviour
2. Elastic-plastic behaviour
3. Elastic-plastic behaviour

All of this behaviour has been discussed in previous chapter where we can see material properties
change effect on the yield stress and consequently residual stress development during cooling after
certain level and duration of heat applied.
4.4 Modelling the Heat Source
When modelling the heat source, we must make sure to understand how the heat source is applied.
When this is clear, the model can be simplified into 2D problem. There are 3 typical type which is
instantaneous point source, instantaneous line source, and instantaneous plane source.

4.5 Flow of Analysis for Welding Deformation and Residual Stress

Simulation Planning

Preparation

Simulation Condition

Run Simulation

Evaluate Results

Yes
Improvise? Evaluate Results

No

Feedback to Design
and Fabrication

Figure 4.2 Flowchart of process to solve problems using simulation

Figure 4.2 summarizes the flow of welding analysis using simulation. Moving from top to bottom.
Firstly, we as the engineer shall be imprinted with the analytical mind to understand the physics behind
the welding. Then we can determine the objective of the simulation based on the problems that we are
expecting to encounter. Once that is clear, necessary preparation of input data and model shall be done.
At this stage, we can determine whether if it’s possible to simplify the 3D model into 2D or 1D to reduce
the computation load. Then we proceed to simulation condition. This stage is almost similar to referring
to WPS in actual welding, and we will modify certain parameter in this stage if the simulation result is
unreasonable and doesn’t satisfy our objective, such as the geometry and welding condition. Once done,
we can run the simulation. This is the stage where thermal analysis will be done to get the time-
temperature file, which will be used in stress analysis. The computed result will be evaluated by
comparing maximum and distribution of temperature, stress and strain with yield stress and melting
temperature (refer to discussion in chapter 1). We can detect errors in input data during this stage by
doing so. Once the evaluated result satisfies our need, we can proceed to use the simulated result to
improve the actual product and improvise fabrication process.