MASTER’S THESIS
OLOF WIIPPOLA
Simplifi ed modelling of Fixtures in FE Welding Simulation
MASTER OF SCIENCE PROGRAMME Engineering Physics
Luleå University of Technology
Department of Mathematics
in FE welding simulation
De ember12,2006
Olof Wiippola
olof.wiippolagmail. om
Master Thesis
Institute of mathemati s
Luleå University of Te hnology
Thegoalwiththismasterthesiswork,istondasimpliedwayofmodelling
the xture in FE welding simulation. To take the entire xture into the
welding model, is very CPU-time onsuming sin e the onta t between the
xture and the work pie e is di ult to simulate. The method usedtoday,
is based on xed degree of freedom boundary ondition. This method do
not in lude the stiness of the xture, the fri tion and the thermal ee ts
between the xture and the work pie e. In order to get a more a urate
modell, these things an not be negle ted. The method that is presented
in this master thesis report, uses already existent fun tions in the FEM
program MSC.Mar . The main idea is to use non-linear springs and non-
linear elasti foundation to model fri tion and support. Sin e the onta t
for ebetweenthextureandtheworkpie edire tlyae tsthefri tionfor e,
thedistribution fun tionsthat des ribes how a known onta t for e at one
point distributedthrough thewhole xture must be used. This fun tion is
onstru ted in two ways, the rst uses the entire model with the xture,
the se ond uses Bu kingham's pi theorem. The tests that where arried
out shows that the distribution fun tions gives a good results, but there
are several problems that must be solved before the methods an be used
in welding simulations. The subroutine used in welding simulation do not
supporttheneedednewstyletableinputwhi hisneededinordertouseedge
foundation as the fri tion for e. MSC.Mar has problem with intera tion
betweenthefa elmandtheedgefoundation. Theproblemwiththefa elm
andtheedgefoundationismostliklyabug,sin ethisproblemdonota ure
inthe newrelea e ofMSC.Mar r3.
Keywords: welding simulation, non-linear spring, elasti foundation, edge
foundation,fa e foundation, onta t
Thismaster thesisworkhasbeen arried out at Volvo AeroCorporation in
Trollhättan. Iwouldliketothankmysupervisors,Seniorle torBoKjellmert
at LTUand Torbjörn Kvist at Volvo Aero. A spe ialthanksto Dr. Henrik
Alberg for his support and advi e during my work at Volvo aero. I would
alsolike to thank thesta of9634 for theresupport.
OlofWiippola
Trollhättan, De ember12, 2006
Contents
1 Introdu tion 1
2 Problem statement 2
2.1 Ba kground . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Denition of theproblem . . . . . . . . . . . . . . . . . . . . 2
2.3 Problemsetup . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3 Methods 4 3.1 Denitionsoffun tionsinMSC.Mar . . . . . . . . . . . . . . 4
3.2 Modellingof onta t . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.1 Shell onta t . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.2 Dete tion of onta t . . . . . . . . . . . . . . . . . . . 6
3.2.3 Conta t toleran e. . . . . . . . . . . . . . . . . . . . . 6
3.2.4 Conta t between deformable bodies. . . . . . . . . . . 7
3.2.5 Penetration . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2.6 Heatuxdue to onta t between deformablebodies . 7 3.3 Modellingof fri tion andsupportwithout onta t . . . . . . . 9
3.3.1 Springs . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3.2 Elasti foundation . . . . . . . . . . . . . . . . . . . . 9
3.3.3 Fa elm . . . . . . . . . . . . . . . . . . . . . . . . . 10
4 Modelling and solution 11 4.1 The models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.1 Thereferen emodel . . . . . . . . . . . . . . . . . . . 11
4.1.2 Thesimpliedmodel . . . . . . . . . . . . . . . . . . . 11
4.2 First step, fri tion test . . . . . . . . . . . . . . . . . . . . . . 12
4.2.1 Conta t for edistribution . . . . . . . . . . . . . . . . 13
4.2.2 How to usethedistributionfun tion . . . . . . . . . . 17
4.2.3 Implementation andresultofthersttest usingpoly- nomialtting . . . . . . . . . . . . . . . . . . . . . . . 18
4.2.4 ImplementationandresultofthersttestusingBu k- ingham'sPi theorem . . . . . . . . . . . . . . . . . . . 20
4.2.5 How to useresultsof therst test . . . . . . . . . . . 21
4.3 Se ondstep . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.4 Welding simulations . . . . . . . . . . . . . . . . . . . . . . . 24
4.4.1 Welding simulation setup . . . . . . . . . . . . . . . . 24
4.4.2 Therst welding simulation . . . . . . . . . . . . . . . 24
4.4.3 Fa elm togetherwithelasti foundation . . . . . . . 24
5 Con lusions 26 5.1 Errors inthesimpliedmodel . . . . . . . . . . . . . . . . . . 26
5.2 Re ommendations . . . . . . . . . . . . . . . . . . . . . . . . 27
A Full size plots a B Element type 75 f B.1 Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . f B.2 Bilinearinterpolation . . . . . . . . . . . . . . . . . . . . . . . f C Element type 7 g D Matlab ode h D.1 Distributionfun tion onstru tionusingBu kingham'sPiThe- orem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . h D.2 Distribution fun tion onstru tion usingnumeri al polyt . . i
Chapter 1
Introdu tion
Welding ofdetailsto theaeroindustryisavery omplextask. There areno
roomfor errors during the welding pro ess. That is why, itis important to
know as mu h as possible about how the material is going to be deformed
duringthewelding pro ess. Thedeformationandstressinthematerial dur-
ingandafter welding areof interest. To make physi al testsofdeformation
and stress during the development step of a new details is both ost and
time onsuming. That iswhy,simulation toolsare oftenused. Onemethod
that is often used is the Finite Element Method (FEM). FEM is a numer-
i al method of solving mathemati al problems. In this Master Thesis, the
possibilityof analternative way ofxture representation isinvestigated.
Chapter 2
Problem statement
This hapter givestheba kground anddenitionof theproblemon whi hthe
master thesis Simplied modelling of xtures in FE welding simulation is
based.
2.1 Ba kground
Inthe aerospa eindustries, itisvery important tohave goodknowledge of
all problemsthat mayo urduring themanufa turing pro ess. To identify
theseproblemsbyexperimentisbothtimeand ost onsuming. Thatiswhy,
simulationtoolsare used.
In welding simulations, deformations and stresses are often interesting.
The purpose of the welding simulations is to nd optimal manufa turing
pro essesand manufa turing on epts. Theweldingpro essesarevery om-
pli ated and CPU-time onsuming to model. E ient methods of xture
simpli ationdoesnot exist at Volvo AeroCorporation today. If thewhole
xture would be a part of the simulation model, the CPU-time that would
be requiredinorder to getgoodsimulation results wouldin rease to alevel
thatisimpossible towork with. ThatiswhyVolvoAeroCorporationtoday
doesnotin lude thextureintothelargeweldingmodels. Insteadxdegree
offreedom(D.O.F)atthe boundariesareused. Thesesimpli ationsdonot
in lude the ee ts ofthe xture su h asthestiness of the xture, fri tion
andthermal ee tsbetween thexture and thework pie e.
2.2 Denition of the problem
As mention earlier, the fri tion between the xture and welding obje t is
today negle ted inthe welding simulations. The goal of this master thesis
workistodevelopamethodthatin ludesthestinessofthexture,fri tion
and thermal ee ts between the xture and the work pie e into the FE
welding simulations, without signi antly in reasing the CPU-time. The
xture and plate on wit h thesimulations arebased on areshown ingure
2.1and 2.2.
2.3 Problem setup
Thexture in2.1isthe referen emodel, onwhi hthesimpli ationsareto
be made. Figure 2.2shows theplate thatis to bewelded. The thi kness of
theplate inthe gure 2.2is
1.65mm
. Asshown in gure 2.1, the left partofthe plate isxed inthe lamped area. The right part oftheplate,on the
otherhand, an move upfrom thexture. The lamping for e of theplate,
omes fromtheboltfor eand the weight oftheupperpartofthe xture.
Fixture top
Plate Fixture
300 mm
300 mm Bolts
60mm
Figure2.1: Fixturesetup
000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000 000000
111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111 111111
00000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000
11111 11111 11111 11111 11111 11111 11111 11111 11111 11111 11111 0000000
0000000 0000000 0000000 0000000 0000000 0000000
1111111 1111111 1111111 1111111 1111111 1111111 1111111
00000000000 00000000000 00000000000 00000000000 00000000000 00000000000 00000000000
11111111111 11111111111 11111111111 11111111111 11111111111 11111111111 11111111111
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 0000 0000 1111 1111
0 1 0 100
11 0000 0000 1111 1111 0000
0000 00 1111 1111 11
Clamp area
Plate 100 mm
190 mm Weld path
Figure 2.2: Plate
Chapter 3
Methods
In this hapter, the methods that are going to be used in order to onstru t
a simpli ation ofthe xture problemin FEwelding simulation are going to
be presented.
3.1 Denitions of fun tions in MSC.Mar
•
MSC.Mar : FEMprogram thatisused inthismaster thesis.•
Table driven input: Almostall inputs into MSC.Mar an be tabledriven. Thatmeansthattheinputvalue(referen evalue)ismultiplied
bythevalueofthetable. Thetable anbeafun tionofseveraldierent
variables, su h as time, displa ement, urrent oordinates and so on.
Forexample,ifaboundary onditionofthetypepointloadisusedwith
an input value in the
x − direction = 1
. That means that the pointloadis 1 inx-dire tion. Ifthe table 3.1is usedas an inputtable, the
pointloadwillbethepointloadreferen e value(whi hisoftensetto1)
multipliedbythe multipli ation value at ea h timestep. That means
thatthe pointload will hange it'svaluewithtimea ording totable
3.1 and gure 3.1. For example after 1 timeunit, the multipli ation
value is 50 a ording to table 3.1 and gure 3.1, this multiplied with
the referen evalue1 givesthe point load50N.
Table 3.1: Exampleof atable for thetabledriven input
Time Multipli ation value
0 0
1 50
2 150
3 300
4 500
5 650
0 3
Time Multiplication
value
100
0 300
200
1 2 4
400 500 600
5
Figure3.1: The gureshows thetable plotfrom thetable driven input
Table 3.2: The point load at ea h time step with the point load referen e
vaule setto 1
Time Point load
0 0
1 50
2 150
3 300
4 500
5 650
3.2 Modelling of onta t
Theanalyzeof onta tis omplex,due tothefa tthatthepositionandthe
motionof the onta t bodymust be determined with a very high a ura y
to avoid thatthe onta tbodies move throughea h other.
There aretwo typesof onta t bodies inMSC.MARC,
•
Deformable onta t bodyis asetof elementsthat a tlikea body.Thedeformable onta t body an bein onta twithotherdeformable
bodiesor ridged bodies.
•
Ridged onta t body is a set of urves or surfa es that a ts as abody. Theridged bodydoesn't deform.
3.2.1 Shell onta t
Shell elements an bein onta t withridged bodies anddeformable bodies.
Anode ofashellelementissaidto be in onta twithasegement,whenthe
position of the shellelement node
±
halfthethi knessof theshell element,normalto the shellarewithin anothersegment.
3.2.2 Dete tion of onta t
Ea h potential onta t node is he ked, if near onta t is possible with a
segment. In3-D,fa esofelementsarepossible onta tsegment. MSC.Mar
usesanear onta talgorithmtodetermineif onta tnodesareneara onta t
segment. Ifanodeisinnear onta t, amoresophisti atedalgorithmisused
to dete t onta t.
3.2.3 Conta t toleran e
Itisunlikely,thatexa t onta t o urs,sin e thepositionofa nodeisa nu-
meri alvalueandtherearealwaysroundo-andtrun ationerrorsinvolved.
That iswhy onta t toleran eis needed. Conta t toleran eis themeasure-
mentofwhen onta t anbesaidtoo ur. Bydefault,the onta ttoleran e
issetto 25%ofthe shellthi kness. Ifavalueofthe onta ttoleran e istoo
large,it will ausetoo manynodesto bein onta t. Iftoo smallvalues are
hosen, it willlead to more omputation time.
3.2.4 Conta t between deformable bodies
When a 3-D deformable onta t body is reated, the program reates a
surfa e to outline the boundary of the onta t body. When a node of a
deformable onta t body omes in onta twitha segment,a tieis reated.
3.2.5 Penetration
To avoid penetration between elements, MSC.Mar has three pro edures
to dete t and avoid penetration. The default and theused pro edure, uses
in rementsplitting. Inthispro edure,thetimestepwhi h ausespenetration
∆t
isdevidedintotwosubin rements. Intherstsubin rement,anoden
1 isnot onstrainedandinthese ondsubin rementthenodeis onstrained,then
MSC.Mar has to nd thetimewhen onta t rst o urs, seeg. 3.2. The
timestep
∆t
aand∆t
b is hoosenbylinearizingthedispla ementin rement.∆t
a= a a + b ∆t
∆t
b= b
a + b ∆t
(3.1)Thatmeans that
∆t
a isthe timewhen onta trst o urs.3.2.6 Heat ux due to onta t between deformable bodies
a
b
Figure 3.2: Illustration that shows how the in rement splitting pro edure
works
When deformable bodies are reated, heat ux lms are automati ally
reatedat allthe boundaries. TheuxesusedinMSC.Mar dependsonthe
distan e
d
betweenthe onta tbodies. Therearetwodierentdistan esthatare used to al ulate the heat ux. The rst is the onta t distan e
d
con,whenthebodiesaresaidtobein onta ta ordingtothe onta ttoleran e.
The se ond is the near onta t distan e
d
near, when the distan e betweenelements is smalleror equal to the smallest element. For distan e
d < d
contheheatuxisdened as[4℄
q = H
T C(T
2− T
1)
(3.2)where
q = heatux
H
T C = lm oe ientT
1 = thesurfa etemperatureT
2 = theinterpolated nodaltemperature at onta t lo ation on the onta ted body.Inthe near onta t ase, theheatuxis dened by
q = H
CV(T
2− T
1) + H
N C(T
2− T
1)
BN C+ σǫf T
A24− T
A14+
(3.3)+
H
CT1 − d d
near+ H
BLd d
near(T
2− T
1)
where
q = heatux
H
CV = onve tion oe ient for neareld behaviourH
N C = natural onve tion for neareldbehaviourB
CN = exponent asso iated withnatural onve tionσ
= Stefan-Boltzman onstantǫ
= emissivityH
BL = distan e dependent heattransfer oe ient.If the distan e
d > d
near, the heat onve tion to the environment is dened byq = H
CT V E(T
2− T
SIN K) + σǫf T
A24− T
A14(3.4)
where
q = Heatux
H
CT V E = heattransfer oe ient to theenvironmentT
SIN K = environment sinktemperature.Iftheneardistan e
d
nearisnotdenedinthethermalpartofthe onta toption,and
d > d
contact,eq.3.4is still valid.3.3 Modelling of fri tion and support without on-
ta t
There aretwopossibleways tomodelfri tionand supportwithout onta t.
The rst is the spring option, the se ond is the elasti foundation option.
Bothoptions an bemodelled withboth linearand non-linear properties.
3.3.1 Springs
The spring option an be used in both me hani al and thermal analysis.
During oupled thermo-me hani al analysis, thethermal part of thespring
a tsasalink. Thethermallinkis ontrolledwithaheattransfer oe ient.
The springs inMSC.MARC an be dened in three ways. All these spring
options need a beginning and a end node. It is important to remember to
apply xed boundary ondition to the begin node, in other ase the spring
will justoatinspa e.
•
FixeddegreeoffreedomThespringfor eisonlya tiveinonedegreeof freedom between the beginning and the end node that denes the
spring.
•
To groundThespring for eis dire tedto origo.•
True dire tion The spring for e isa tive inthedire tion of thelinebetween the two nodeswhi hdene thespring.
Thespring for e
F
s isdened byF
s= k(U
1− U
2)
(3.5)where
k
is the spring stinessandU
isthe displa ement.3.3.2 Elasti foundation
Elasti foundations workslike springs, butthey don't operate on nodeslike
springs.
Edge foundation
Theedgefoundations,asthenamesays,operateontheedgesoftheelement.
Thefoundation for eis dened by
F
f= k (U
1− U
2)
(3.6)where
F
f isthe foundation for eandk
is the foundation stiness.U
1− U
2isthe displa ement of the edge.
Fa e foundation
Thefa efoundationworksliketheedgefoundation,butoperatesonthefa es
ofthe elementsinsteadof ontheedges. Thefoundation for eisdened like
eq. 3.6, but then
U
1− U
2 isthedispla ement ofthefa esof theelement.3.3.3 Fa e lm
In order to simulate the heat transfer to the xture and the environment,
thefun tionfa e lm an beused. Fa elmwaspresentedinse tion3.2.6.