Multi-physics and non-linear modeling using

COMSOL
Venkat Perumal
Siemens Technology and Services Private Limited
Research and Technology Center, # 80, Keonics, Electronics City
Bangalore 560100, India
[email protected]

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Introduction
The keynote presentation will cover the following topics
Modeling of Friction Stir Welding (FSW) process
Non-linear modeling of polymeric structure
Application building for composite structures

Page 2

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Friction Stir Welding (FSW) is relatively advanced

metal joining technology works in the solid state
FSW is widely accepted joining process for aluminum, magnesium, titanium alloys
FSW description

Joint types

Shoulder

Pi
n

Welding by frictional and plastic/viscous dissipation

Shoulder: Generate frictional heat and prevents material

expulsion

Pin: Deform material and generate plastic/

viscous heat
Ref: http://www.twi-global.com
Perumal, V.(2010). Dissimilar Metal Friction Stir Welding of Aluminum to Magnesium Alloys. (Doctoral dissertation). Retrieved from http://scholarcommons.sc.edu/etd/2243

Page 3

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

FSW finds application in several industries due to

its technological advancements over the
conventional processes

Aerospace1

Ship building2

Mobility1

Energy1

Image courtesy: 1. www.siemens.com , 2. www.twi-global.com

Page 4

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Heat generation in FSW is governed by the friction

between the shoulder and workpiece and the
workpiece plasticity
Heat generation due to the pin and shoulder

Heat generation due to shoulder (qshoulder)

Heat generation due to the friction between

the shoulder and the workpiece is function of
the
- friction coefficient
Fn forge force (from experiments &
estimation)

Heat generation due to pin tool (qpin)

As surface area of the shoulder

- angular velocity (rad/s)
r- radial distance (x2+y2)0.5
Heat generation due to pin tool (qpin) is the the
function of the
Pin radius (rp)
- angular velocity (rad/s)
(T) temperature dependent yield
strength of the workpiece

qshoulder

qpin

www.comsol.com
P. Colegrove et al., Proceedings of the 2nd International Symposium on Friction Stir Welding, Gothenburg, Sweden, 2000

Page 5

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Temperature dependant yield strength of the

alloys is an important input model
Chemical composition (wt .%) of aluminum alloys 1
Al

Cu

Mg

Cr

Fe

Mn

Si

AA6061

95.8-98.6

0.15-0.4

0.8-1.2

0.04-0.35

0.070

0.002

0.4-0.8

AA5754

94.2-97

0.10

2.6-3.6

0.0003

0.004

0.005

AA7075

87.1-91.4

1.2-2

2.1-2.9

0.18-0.28

0.005

0.003

Ti

0.15

0.004

0.0015

0.15

0.004

0.0020 5.1-6.1

Thermal conductivity (W/m-k)

AA6061
-T6

200
150

Yield strength (MPa)

100
50
0
275

375

475

575

675

775

Residuals

0.0015 0.0025

300
250

Zn

180
160
140
120
100
80
60
40
20
0

0.002

167
125

130

875

Temperature (K)

Variation of yield strength with temperature 2

Thermal conductivity of the alloys

1. http://www.matweb.com
2. Perumal, V.(2010). Dissimilar Metal Friction Stir Welding of Aluminum to Magnesium Alloys. (Doctoral dissertation). Retrieved from
http://scholarcommons.sc.edu/etd/2243

Page 6

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

The backing plate effect was simulated by

the appropriate convective heat transfer
coefficient (h)
The backing plate effect was simulated by the appropriate convective heat transfer
coefficient (h) to save the computational time
1
2
2&3
2&3

Constant temperature

Convective surface

Radiation

Outflow

Heat source

2
1

qshoulder
Page 7

COMSOL Conference, 29-30 October, 2015, Pune

qpin

Unrestricted/ Siemens AG 2015. All rights reserved

Elasto-plastic behavior coupling of the

workiece with thermal to address the multiphysics nature of FSW
A linear isotropic hardening model is used to model the large strain
plasticity

ys0 is the initial yield stress, and

k is the isotropic hardening
modulus.
A value for Etiso isotropic tangent
modulus section for in the
plasticity region

Elasto plastic behavior of the workpiece

https://www.comsol.co.in/
Page 8

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Elements around the tool/workpiece

interface are smaller than rest of the
workpiece
An auto meshing option was used to mesh the geometry. Backing plate is
excluded to reduce the computational time, however, taken care by the heat
transfer coefficient

Element
type
Tetrahedral

48219

Triangular

10033

Edge

Page 9

No. of
elements

847

Vertex

38

Quality

0.70

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Shear layer around the pin tool has been

modeled to represent the typical high
strain rate in FSW
Velocity and strain components
In the sticking condition the velocity
component vectors
inside the shear layer are given as function of the
i) tangential velocity caused by the rotational
motion &
ii) translational velocity of the material in x
direction1

Shear layer

Shear layer around the pin2

where,
u -x component of velocity
v- y component of velocity

-angular velocity of tool in (rad/s)

uweld- welding velocity (mm/s)
- ramping factor for the shear layer

Strain rate and effective strain rate can be

expressed as follows
Velocity inside the shear layer around the pin: Source: Schmidt, H et al.,
Modelling and Simulation in Materials Science and Engineering, Vol. 12,
2004, pp. 143 157.

[1] H.B. Schmidt, J. Hattel, Science and Technology of Welding and Joining Vol:10(2005),pp:176-186.
[2] Basil M. Darras, Experimental and analytical study of friction stir processing, (2005). University of Kentucky Master's Thesis.

10

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

FSW simulation set up used for AA6061: Initial

parameters have been arrived based on some
prior data
Geometric dimensions for thermo-mechanical analysis.
Tool pin upper diameter

10.32 mm

Tool pin lower diameter

6.38 mm

Tool pin height

2.40 mm

Tool shoulder diameter

25.40 mm

Plate size

150 mm x100 mm x 6.1 mm

Simulation Parameters
RPM 500, 575, 600, 700, 800, 900, 1000,
and 1100.
Feed rate (in mm/min) 10, 15, 20, 30, 50,
and 80.
Experimental Parameters
RPM 575, 800, and 1100.
Feed rate (in mm/min) 10, 15, 20, 30, and
50.

Page 11

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Results and discussion

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

The weld surface with free from flash was

obtained at optimal heat input conditions
Image of a AA6061 butt weld joint after 800 rpm, 20 mm/min feed
rate

Successful weld
Exit

start

25 mm

Unsuccessful welds at low

rpm and high feed rates
E.g., 575 rpm and >10
mm/min
feed rates
Unsuccessful welds at high
rpm
levels regardless of the
feed Unsuccessful weld
rate

Direction of tool traverse

Image of a AA6061 butt weld joint after 575 rpm, 30 mm/min feed
rate
(bottom right)

The weld surface shows improper (low) heat

input at 575 rpm and 30 mm/min resulting in
no weld formation
Page 13

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Isothermal temperature contours show peak

temperature inside the weld nugget
AA 6061 alloy, 800 RPM and 20 mm/min

Temperature isotherm of the Al plate

Page 14

COMSOL
29-30October,
October,
2015,
Pune
COMSOL Conference,
Conference, 29-30
2015,
Pune

Temperature at the cross section

Unrestricted/ Siemens AG 2015. All rights reserved

The calibration of the model was performed using

thermocouple measurements at the HAZ
The maximum deviation between the experimental and the simulated temperatures in HAZ is
about 1%

Comparison of the surface temperature from the IR pyrometer

and simulation
Side
Advancing
Retreating
Page 15

Parameters
1100 rpm, 30
mm/min

Thermocouple measurement locations in HAZ

Experimental
Temperature (K)

Simulation
Temperature (K)

634

628

0.94

616

615

0.16

COMSOL Conference, 29-30 October, 2015, Pune

% Error

Unrestricted/ Siemens AG 2015. All rights reserved

Mid plane of the nugget remains hot and less

hot at distances ahead of the tool

Page 16

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

The increment in the weld nugget

temperature is proportional with the
strain rate
900

80

800
70

700

Strain rate (1/s)

600

60

500
50

Strain rate (1/s)

Temperature (K)

400
300
900

40

500

600

700

800

Strain rate is obtained for different tool

rotational speeds 500 to 800 rpm; the
tool traverse speed was kept constant
Temperature (K)
of 15 mm/min.
It is observed that the effective strain
rate increases with increase in the tool
rotational velocity.

Tool rotational speed (rpm)

102

Strain rate (1/s)

Temperature (K)

100

Temperature (K)Strain rate is obtained for different

weld speeds of 10,15,30,60,120

mm/min and the tool rotational speed
was kept constant of 800rpm.

98
96

Strain rate (1/s)

94

The strain rate decreases as the

traverse speed increases.

92
90
10

20

30

40

50

60

70

80

90

Tool traverse speed (mm/min)

17

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

AA5754 alloy showed highest nugget

temperature than AA6061 and 7075 alloys due to
yield strength K differences
The nugget temperature of the AA5754 alloy is the highest compared to that of the AA 6061
and AA 7075 alloys, which can be attributed to the yield strength and thermal conductivity
differences.

800

700

AA 5754
AA 6061
AA 7075

Temperature (K)

Nugget temperature
Adv. side

Ret. side

AA 5754

851

847

AA 7075

744

743

AA 6061

730

727

600

HAZ temperature

500

Adv. side

Ret. side

AA 5754

705

682

AA 7075

630

612

AA 6061

628

615

400
-100 -80 -60 -40 -20 0
20 40 60 80 100
Distance from the weld center (mm)
Page 18

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

In aluminum alloys, the heat generation is

governed
by both the flow stress and thermal conductivity
900
850

The rotational velocity of the tool

was
swept from low to high rpm for a
constant feed rate of 20 mm/min

Peak Temperature (K)

For AA5754 & AA 7075, 300-575 rpm

800
750
700

resulted in T<Tsolidus

650

For AA6061, the Tsolidus was reached

at 1100 rpm, where no welding was
formed. Also at low rpm levels 575
rpm

600
300

AA 5757
AA 7075
AA 6061

500

700

900

1100

1300

1500

Tool rotational speed (rpm)

Solidus temperature
Alloy
AA5754
AA6061
AA7075

(TemperatureC)
603
583
477

Temperature (K)
876
856
750

http://asm.matweb.com
http://www.diva-portal.org/smash/get/diva2:515154/FULLTEXT01.pdf

Page 19

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Modeling of the non-linear behavior using

COMSOL

Elastomers
Loose cross linking of amorphous
Loose fixation of polymer chain by the chemical
bonds results in highly elastic behavior above the
glass transition temperature [Tg]
Elastomers exhibit viscoelastic behavior

Creep : If the stress is held constant, the strain

increases
with time.

Creep and recovery behavior of elastomer

Stress and strain versus time plot

Relaxation : If the strain is held constant, the

stress
decreases with time.
Viscoelastic nature of elastomer

Viscoelastic material properties have strong

Page
20
dependence
on
COMSOL
Conference, 29-30 October, 2015, Pune

Image ref. Lakes R.S , Viscoelastic materials

Unrestricted/ Siemens AG 2015. All rights reserved

Viscoelastic models in COMSOL

Viscoelastic models

Pure Maxwell model

Kelvin Voigt model

Standard linear solid

model

In pure maxwell model only

one spring and dashpot is
connected in series
Here, spring = dashpot
= spring + dashpot

Only one spring-dashpot

branch is connected in
parallel
s = s spring + s dashpot
e = e spring = e dashpot

Zener model, or 3
parameter model
Placing a spring in parallel
with the Maxwell model

Generalized Maxwell
model

More spring dashpot

pair gives better
solution

In the present simulation, Pure Maxwell model with shear modulus and relaxation time is used.

Page 21

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Non linear hyperelastic simulation of

damper elastomer

A hyper or Green elastic material is an ideally elastic

material.
Stress-strain relationship derived from a
Strain Energy Density Function

A strain energy density function relates strain energy

density
of a material to the deformation gradient.

Unlike other materials such as metal and ceramics there is

no
linear relationship between stress and strain in Elastomers

inNeo- Hookean
Mooney
Rivlin state.
2 parameter
hyperelastic

<30%
100% in tension 30% in
compression

Mooney Rivlin 5 and 9 parameters 30-200%

Ploynomial function of order N
300%

Feasible to model up to

Arruda Boyce

<300%

Ogden

<700%

Page 22

Neo-Hookean
Mooney-Rivlin
Yeoh
Arruda-Boyce
Mooney-Rivlin is a commonly used model
for polymers
Blatz-Ko

COMSOL Conference, 29-30 October, 2015, Pune

polymers,
moderate stretch
levels

Ogden

Polyurethane
foams
High stretch
levels

Unrestricted/ Siemens AG 2015. All rights reserved

Hyperelastic model used in simulation

Mooney Rivlin two parameter model :

Only a nearly incompressible version is available.

Elastic strain energy density is written in terms of the two invariants of

the deformation tensors I1(Cel) and I2(Cel) and the elastic volume ratio
Jel

Material properties required for top and base

body:
Youngs modulus
Poisson's ratio
Density

Material Properties required for damper:

Mostly used model during finite element analysis.

Mooney Revlin model parameter C10

Mooney Revlin model parameter C01
Youngs modulus
Poisson's ratio
Initial bulk modulus
Density

Relation between the Mooney material parameters, shear and Youngs

modulus are,
G0 = 2(C10+C01)
E0 = 6(C10+C01)

Page 23

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Contact formulations in COMSOL

To solve the contact problems two types of approach is used

Augmented Lagrangian method

Penalty Contact Method

Augmented Lagrangian method

By default COMSOL solves contact problem using ALM, in a Segregated Way.
Tn = Contact pressure variable
dg = Gap distance between destination and source boundary
Pn = User defined penalty factor

Penalty Contact Method

hmin

To
G

= Minimum element size on the contact pair.

= Constant penalty factor
= Estimated contact pressure
= Clearance between the parts

Tnp = Penalized contact pressure , Value of penalty factor in ALM does not affect
the accuracy but it influences the convergence.
Penalty contact method does not require extra variable for the contact
pressure. Penalty contact method speeds up the solving process but it is
less exact. Contact pressure estimated from penalty contact formulation
COMSOL Conference, 29-30 October,
2015,
Pune
Unrestricted/
Siemensformulation.
AG 2015. All rights reserved
is not as
precise
as the Augmented
Lagrangian

Page 24

Hyperelastic model results showed

increased deflection level than
viscoelastic model
Viscoelastic model

Page 25

COMSOL Conference, 29-30 October, 2015, Pune

Hyperelastic model

Unrestricted/ Siemens AG 2015. All rights reserved

Composite structures application building

using COMSOL

COMSOL API (Application Programming Interface) which is an interface

based on Java, is used to develop custom application based on COMSOL
An initial GUI is built based on a model of a simple Rectangular Block,
having a point force acting on the edge
Final GUI is built based on a model of Mechanical Part, having a
combination of forces and moments acting on it
Halpin-Tsai Model, a semi-empirical model chosen based on its accuracy
and appropriateness, is used to develop algorithm to calculate Composite
Material properties

Ref. Venkateswaran et al., 2014 COMSOL Conference, Bangalore

Page 26

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Haplin-Tsai Model to calculate the

composite properties using rule of
mixture
Halpin-Tsai equations are the handy forms of Hills generalized selfconsistent model results with engineering approximations to make them
suitable for the designing of composite materials.

where,
M = composite material modulus E22, G12 or 23
= fiber material modulus Ef, Gf or f
= matrix material modulus Em, Gm, or m

Page 27

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

User interface for pre-processing

INPUT INTERFACE

Page 28

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

User interface for post processing

INTERFACE FOR POSTPROCESSING OF

RESULTS

Page 29

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Stress plot in a GUI

SIMULATION ON A SIMPLE BLOCK

Page 30

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Mesh visualization in GUI

VISUALIZING MESH
Page 31

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Contour stress plot in GUI

Page 32

CONTOUR PLOT

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Concluding remarks

The modeling of friction stir welding process has enhanced

the understanding of the process, and resulted in reducing
the experimental runs

The selection of appropriate model and parameters is the

key in modeling the non-linear response of the polymeric
structure

JAVA API option in COMSOL enabled the pre and post

processing GUI for composite structures modeling

Page 33

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved

Acknowledgment

Author thanks:
Hande V (Research Group Head)
Siemens Tech. and Services Pvt. Ltd., Bangalore
for his inputs & approval
Chethan Ravi, BR
Siemens Tech. and Services Pvt. Ltd., Bangalore
for his contribution on viscoelastic/hyperelastic
modeling results

Page 34

COMSOL Conference, 29-30 October, 2015, Pune

Unrestricted/ Siemens AG 2015. All rights reserved