1

Weld Modeling with MSC.Nastran

J.Fang, C.Hoff, B.Holman, F.Mueller, D.Wallerstein
MSC.Software Corporation, Costa Mesa, Los Angeles, CA


ABSTRACT

A new general purpose connector element has been developed in MSC.Nastran. The element models
the global load transfer in spot weld and seam weld connections. The weld element has a physical
stiffness. Constraint equations are generated internally. The resulting element stiffness is always of proper
rank. The user does not have to take precautions for in plane rotations in shells. Three types of
connections can be defined with the new weld element, point to point, point to patch, and patch to
patch. The connected elements can be of different type. The meshes do not have to be congruent. The
user can request that the weld element is generated without explicit constraints, thereby avoiding the
costly elimination of m-set degrees of freedom and increasing the performance for large models.

The connectivity and property of the new element is described. A short overview of the underlying
constraint equations is given. Numerical experiments are conducted to show the accuracy, limitations,
and overall performance of the new element. MSCs Automotive Modeling System for Full Vehicle
Analysis MSC.AMS-FVA is used to show the spot weld modeling process.
2
INTRODUCTION

For a number of years, the automotive customers of MSC.Nastran have asked MSC for better tools to
model spot welds. MSC has hesitated to develop a new special purpose modeling tool mainly because
MSC.Nastran has a variety of capabilities to model connections, for example, flexible springs and bars
(CBUSH, CBAR, etc.[1]), rigid elements (RBAR, RBE2, RBE3, etc. [1]) and multi-point constraints
(MPC [1]). In recent years, criticism about the existing modeling tools has grown with the rapidly
changing requirements of todays CAE process, see for example [2]. The meshes in automotive finite
element models are getting finer to achieve higher accuracy. Users are constantly searching for ways to
cut processing time. Modeling tools must be easy to use, must be generated fast, must fit into the design
process, and must cope with demands in overall performance.

With the traditional tools of MSC.Nastran, spot welds are modeled using a point to point connection.
The shell vertex grids of an upper and lower shell are connected with a flexible bar or a rigid element,
see the bulk data entries CBAR and RBE2 [1]. A lot of automotive models exist using the point to point
technique mainly because MSC did not offer more convenient tools. There are three disadvantages using
the point to point connection.

1. Meshes of different parts have to match at the spot weld connection points. The two grids of the
spot weld have to line up. Some parts must be re-meshed, most often with time consuming manual
procedures.

2. The point to point connection underestimates the stiffness of the spot weld connection for large spot
weld diameters. With finer meshes, the diameter of a spot weld is of the same order of magnitude or
just one order of magnitude less than the element length. The diameter must be taken into account to
simulate the force transfer from the spot weld to the shell correctly.

3. The point to point connection causes numerical problems in the finite element shell model. In Mindlin
shell theory, the stiffness of the rotations normal to the shell plane does not exist. MSC.Nastrans
QUAD4, for example, is a Mindlin shell element. In the point to point connection, rotations normal
to the shell plane may be excited. No problems occur if the shell vertex grids of the upper and lower
shell element are lined up so that the vector connecting the two vertex grid points in the direction
normal to both the upper and lower shell plane. If the two shell vertex grids are not lined up,
spurious modes may occur and the force transfer becomes inaccurate. Users have tried to defeat the
pitfalls of this technique by using the parameter entry K6ROT [1] which puts a small penalty
stiffness on the rotational degree of freedom of the shell.

The first and second disadvantage are inherent to the point to point connection and can only be
overcome with other connection types. The third disadvantage can be removed by defining discrete
normals at the shell vertex grids with the bulk data entry SNORM, see [1]. The two discrete normals of
the upper and lower shell vertex grid must be in the direction of the vector connecting the two vertex
grids. Several MSC.Nastran users have written pre-processing routines to generate the normals.
Because a lot of models exist with point to point connections for spot welds, MSC has decided to offer
a tool to overcome the third disadvantage by generating shell normals automatically, see below.
3

Most efforts in spot weld modeling are now concentrating on two requirements. First, the spot weld
modeling tool must be able to connect non congruent meshes so that parts do not have to be re-meshed.
Second, the area of the spot weld must be taken into account to get a correct force transfer from the
weld into the shell.

In recent years, spot weld generators have been developed by several third party software vendors using
existing finite element tools and generating input for solvers like MSC.Nastran. All these products start
with an existing finite element model without spot welds. The information of the spot weld locations is
added from a separate file containing a list of spot weld grid points with xyz locations. The spot weld
information comes from a CAD model or a special spot weld modeling system. The projection of the
spot weld grid to two or more shell elements is found. Then, the spot weld itself and its connection to
the shell elements is generated. For example, in [3], the spot weld itself is modeled with a HEXA
element which has the actual size of the weld nugget. The HEXA grids are connected to the three
translational degrees of freedom of the shell vertex grids with RBE3 type elements. This modeling
technique fulfills the two main requirements, non congruent meshes and area consideration. MSCs
Automotive Modeling System - Full Vehicle Analysis MSC.AMS-FVA [4] offers an integrated spot
weld modeling package including MSC.Nastran and the spot weld generator described in [3]. The
package is used by several MSC clients in the automotive industry.

Although the solution in [3] and other products on the market fulfill all requirements for spot weld
modeling today, MSC has decided to offer an in-house solution with a general purpose connector
element in MSC.Nastran. MSC wants to be able to offer a complete package including graphical user
interface, spot weld generator, and solver.

The German Volkswagen AG has approached MSC about two years ago and has proposed an input
definition of a spot weld in MSC.Nastran which is close to the definition used in PAMCRASH [5] .
MSC has written a proto-type which handles non congruent meshes. The results have been presented a
year ago [6].

After a successful launch of the proto-type, MSC has improved and extended the weld element. The
area of the weld is taken into account. In addition, the weld element has been extended to a general
purpose connector element which can model other types of connections and can connect arbitrary
element types. Besides spot welds, the element can be used to model seam welds, see Figure 1. The
element can also be used to model bolts, screws, and engine mounts. In this paper, we will only discuss
the spot weld connection.

4











We start with a description of the connectivity and property of the new element. The underlying
constraint equations are explained. Then we show some numerical experiments to demonstrate the
accuracy and the limitations of the new element. Finally, we show MSC.Nastrans spot weld capability
integrated into MSC.AMS-FVA [4].


SPOT WELD CONNECTIVITY AND PROPERTY DEFINITION

With the new CWELD element, the user can define three types of connections.

1. For a point to point connection, an upper and lower shell vertex grid, GA and GB, are connected,
see Figure 2. The weld axis is the vector from GA to GB. Two shell normals are automatically
generated for the upper and lower shell vertex grid, both normals point in the direction of the weld
axis. The user does not have to specify PARAM, SNORM to activate the generation of normals.

2. For a point to patch connection, a vertex grid point GS of a shell element is connected to a surface
patch, see Figure 3. The patch is either defined by grid points GAI or a shell element SHIDA. A
shell normal in the direction normal to the patch is automatically generated and put on GS.

3. For a patch to patch connection, a spot weld grid GS is connected to an upper and lower surface
patch. GS may or may not lie on the patches, see Figure 4. The upper patch is defined by grid
points GAI or by a shell element SHIDA, the lower patch by GBI or SHIDB.

The second and third type are defining the most general connectivity. In addition, the third type meets the
requirements of non congruent meshes and weld area consideration. The spot weld grids GS are usually
not associated to a finite element mesh, they are generated in a CAD system or a special purpose
automotive modeling system like MSC.AMS-FVA [4]. The spot weld grid GS may or may not lie on
the finite element geometry. The weld element geometry is determined by projecting the spot weld grid
GS on the finite element faces of the upper and lower shell element. First, GS is projected normal to the
upper shell, the piercing point is called grid GA. Along the direction GS-GA, the piercing point GB on
the lower shell is determined. The projected grids, GA on the upper shell and GB on the lower shell, are
defining the element length and direction of the spot weld itself, see Figure 4.

Figure 1. Weld

Spot B-Seam T-Seam
5










Figure 2. Spot Weld with Point to Point Connection.



Figure 4. Spot Weld with Patch to Patch Connection.

GA
GB
n
n
upper shell mid surface
lower shell mid surface
Figure 3. Spot Weld with Point to Patch Connection.
GA1
GA2
GA
GS
GA4 GA3
n
GB1
GB3
GA1
GB
GB2
GA2
GA
GS
GA4
GA3
SHIDB
SHIDA
6
If more than two layers need to be connected, a second CWELD element is defined which refers to the
same spot weld grid GS as the first CWELD. The projection vector of the first CWELD determines the
weld element axis and is maintained for all following layers. Up to 10 layers can be connected. Elements
of different types can be connected, for example, a CQUAD8 and a CTRIA3. The spot weld grid GS
can be at an arbitrary location as long as the projected grids GA and GB lie in the upper and lower
surface patch, respectively. After GA and GB are generated, the CWELD element connectivity is
stored in a generalized form, the grids GS, GA, GB, grids GAI of the upper surface patch and grids GBI
of the lower surface patch. A surface patch must have at least 3 grids and can not exceed 8 grids. A
surface patch does not have to correspond to an element.

The property of a CWELD element is defined on a PWELD entry. The property parameters are the
material identification number and the diameter D of the spot weld. The property entry has additional
parameters to describe seam welds. In addition, the user can define flags for advanced features. For
example, the user can ask to include all constraint equations of the element in the stiffness matrix and
thereby avoid the creation of m-set degrees of freedom. Without explicit m-set degrees of freedom, the
resulting system of equations is more robust and faster to solve.


FINITE ELEMENT REPRESENTATION OF THE CONNECTOR

The weld itself is modeled with a special shear flexible beam type element with length L and a circular cross
section with diameter D. The length L is the distance of GA to GB. The element has 2x6 degrees of
freedom. The element has been developed to behave well for very small ratios of length L to diameter D,
see Figure 5. The range of the ratio L /D is restricted to 0 . 5 / 2 . 0 D L
by default. The program overwrites the length L if it is outside the range to protect against spot welds
with zero length and overly long spot welds. The user can change the range on the property entry.
















Figure 5. Spot Weld Element

L
D
GB1
GB2
GB3
GB
GA1
GA4
GA3
GA2
GA
7
For the point to point connection, two normals are generated at grid GA and GB in the direction of the
weld axis. For the point to patch and patch to patch connection, the degrees of fredom of the spot weld
end points GA and GB are constrained as follows. The translational and rotational degrees of freedom of
spot weld end grid GA are connected to the translational degrees of freedom of grids GAI with
constraints from Kirchhoff shell theory,



( )
I
A A I
A
w
v
u
N
w
v
u

'

'

,
(1)



( )

,
_

I y I I x I
A
z
I x I
A
y
I y I
A
x
u N v N
y
u
x
v
w N
x
w
w N
y
w
, ,
,
,
2
1
2
1

(2)



The 6 equations are written in the local tangent system of the surface patch at point GA. The two tangent
directions are x and y, the normal direction is z.
I
N are the shape functions of the surface patch,
A
and
A
are the normalized coordinates of GA, u, v, w are the displacements,
x
,
y
,
z
are the rotations in
the local tangent system at GA. For grid GB, another set of 6 equations similar to (1) and (2) are written
resulting in 12 constraint equations.

In summary, the spot weld element consists of a two node element with 12 degrees of freedom and 12
constraint equations. The user has two options to process the constraint equations. In the first option,
the constraints are generated as explicit multi-point constraints, the 2x6 degrees of freedom of GA and
GB are put into the dependent set (m-set). In the second option, the 2x6 constraint equations are
worked into the 12x12 stiffness matrix of the spot weld element. The resulting element is a 3xN degrees
of freedom element where N is the total number of grids GAI plus GBI. The second option does not
generate m-set degrees of freedom in the problem. The costly m-set elimination is avoided. Furthermore,
occasional problems with singular constraint matrices are avoided.

8
EXAMPLES

SPOT WELDS MODELED WITH THE POINT TO POINT CONNECTION

We investigate a simple model of a two layered plate to demonstrate the numerical problems with
traditional modeling techniques of point to point connections. Two sheets of 1x30x90 mm are connected
with spot welds, see Figure 6. Each sheet is modeled with a mesh of 3x9 CQUAD4 elements, the
meshes are connected with two rows of 8 spot welds using the point to point connection. The spot
welds are modeled in two ways.

1. The traditional modeling technique using rigid bars (RBARs) and K6ROT.
2. The new CWELD element with the point to point connection.

We compare the results of three different meshes of the upper layer. First, the upper mesh is perfectly
aligned with the lower mesh. This mesh is the baseline. For the second and third mesh, the inner grids of
the upper layer are slightly perturbed with an amplitude of 0.2 mm in the y-direction. The second mesh
has a wavy pattern, the third mesh has a uniform shift.






wave perturbation y shift perturbation y
n
n
y
x
z
x
2 x ( 3 x 9) Quad4s t = 1.0 mm
D = 2.0 mm
y = 0.2 mm
E = 2.06 e+5 N/ mm
2
= 0.3
Perturbation
Figure 6. Two Plates Connected with 16 Spot Welds
mm 0 . 1 t
2 x 8 Spot Welds
mm 90
3
0

m
m
2
kg/mm 8 - e 785 . 0
n
9

The eigenfrequencies of the model are calculated in MSC.Nastran Sol 103. For the perfectly aligned
mesh, the results of the traditional technique using RBARs and K6ROT are nearly identical to the new
CWELD element. For the two perturbed meshes, we do not expect a significant change in the results
because the outer grids are still perfectly aligned. With the traditional modeling technique the
eigenfrequencies are underestimated by 6 to 12 %, see Figure 7. With the CWELD element, the results
of the two perturbed meshes are nearly identical to the perfectly aligned mesh, the errors are so small
that they are not visible in Figure 7.

If K6ROT is not used in the traditional modeling technique with RBARs, then the error is much larger
and spurious modes may occur. With the new CWELD element K6ROT is not necessary.



A large scale model of a body in white (BIW) with point to point connection of spot welds has been
converted from the traditional modeling technique with RBE2s and K6ROT to the new CWELD
elements. For the first 40 eigenfrequencies, the RBE2 and K6ROT model is about 1.0 to 2.5 % softer
than the CWELD model.





Figure 7. Errors in the first 6 Eigenfrequencies
Error in first 6 Eigenvalues
-14
-12
-10
-8
-6
-4
-2
0
2
1 2 3 4 5 6
Eigenfrequency No.
E
r
r
o
r

[
%
]
CWELD, shifted grids
RBAR, shifted grids
CWELD, missaligned grids
RBAR, missaligned grids
(not visible)
(not visible)
10


SPOT WELDS MODELED WITH THE PATCH TO PATCH CONNECTION

The stiffness of the patch to patch connection is investigated with a simple model of two QUAD4
elements connected with one CWELD element. The diameter D of the spot weld is varied from

0 . 1 01 . 0
S
D


where S is the element length. We look at deformations for the following loads and boundary conditions:
in plane shear, in plane twist, transverse shear, and pure bending, see Figures 8 to 11. We compare the
displacements of the extreme case D/S=1.0 with two other models. First, we use a HEXA element to
connect the two QUAD4 elements. The HEXA element has the same dimensions as the QUAD4
elements in the x-y plane and a height equal to the thickness of one shell in the z-direction. Second, we
use one shell element with double thickness.

All displacements are normalized with the displacements of the HEXA model with D/S=1.0. The
normalized displacements are plotted against the ratio D/S in a double logarithmic scale to show the
dependency of the displacements with respect to the diameter D. The plots show a clear dependency of
the deformations with respect to the spot weld diameter that is consistent with continuum theory. For the
extreme case of D/S= 1.0, the CWELD is softer than the CHEXA in shear and twist. However,
compared to the shell with double thickness, the CWELD is stiffer in shear and twist. Overall, the new
CWELD element has a reasonable stiffness for diameters which do not exceed the size of an element.











11

Figure 8. Spot weld deformation for various spot weld diameters, in plane shear



Figure 9. Spot weld deformation for various spot weld diameters, in plane twist.

In Plane Shear
0.1
1
10
100
1000
0.001 0.01 0.1 1
D/S Ratio of Diameter to Element Length
U
X
/
U
X
H

R
e
l
a
t
i
v
e

D
i
s
p
l
a
c
e
m
e
n
t
~2
1.591 1.693
10.0
y
x
S = 20.0
2QUAD4 t =1.0
HEXA8 20. x 10. x 1.
2QUAD4 t =1.0
D
z
x
1.0
2t t QUAD4
In Plane Twist
1.00E+00
1.00E+01
1.00E+02
1.00E+03
1.00E+04
0.001 0.01 0.1 1
U
X
/
U
X
H

R
e
l
a
t
i
v
e

D
i
s
p
l
a
c
e
m
e
n
t
~4
1.393
D/S Ratio of Diameter to Element Length
2.635
2QUAD4 t =1.0
HEXA8 20. x 10. x 1.
2t t QUAD4
2QUAD4 t =1.0
D
z
x
1.0
10.0
y
x
S = 20.0
12

Figure 10. Spot weld deformation for various spot weld diameters, transverse shear.


Figure 11. Spot weld deformation for various spot weld diameters, pure bending


Transverse Shear
0.1
1
10
100
1000
10000
0.001 0.01 0.1
1
U
X
/
U
Z
H

R
e
l
a
t
i
v
e

D
i
s
p
l
a
c
e
m
e
n
t
~2
0.637
D/S Ratio of Diameter to Element Length
10.0 y
x
S = 20.0
2QUAD4 t =1.0
HEXA8 20. x 10. x 1.
2QUAD4 t =1.0
D
z
x
1.0
2t t QUAD4
Pure Bending
0.001
0.01
0.1
1
10
0.001 0.01 0.1 1
D/S Diameter versus Element Length
U
Z
/
U
Z
H

R
e
l
a
t
i
v
e

D
i
s
p
l
a
c
e
m
e
n
t
1.134
10.0
y
x
S = 20.0
2QUAD4 t =1.0
HEXA8 20. x 10. x 1.
2QUAD4 t =1.0
D
z
x
1.0
2t t QUAD4
1.07
13
SPOT WELD MODELING PROCESS WITH MSC.AMS-FVA AND
MSC.NASTRAN

Two automotive parts are connected with spot welds. The parts have non congruent meshes. The finite
element model of both parts exist in form of a MSC.Nastran input file. In MSC.AMS-FVA the user
specifies the spot weld locations GS or reads the location from a file generated by a CAD system.
MSC.AMS-FVA finds the connection of the spot weld grid GS to the upper and lower shell element,
generates CWELD elements and merges them into the existing MSC.Nastran input file. The model of
the connected parts is then analyzed in MSC.Nastran. Several Body in White FE models from
automotive customers have been processed successfully with MSC.AMS-FVA and MSC.Nastran. The
results of the new CWELD match experiments within sufficient accuracy and furthermore, agree with
results from other simulation techniques, for example with CDH-Spot [3].



CONCLUSIONS

MSC has developed a general purpose connector element which can connect two vertex shell points, a
vertex shell point and a surface patch or two surface patches. The element fulfills the two basic
requirements for spot welds, it can join non coincident meshes and it takes the area of the spot weld into
account. The new spot weld element simulates the force transfer between two surface patches
accurately for 0 . 1 / S D where D is the spot weld diameter and S is the mesh size. If the diameter is
larger than the surface patch, the spot weld element may underestimate the stiffness of the connection.
The weld element is always of proper rank. For high performance analysis, the user has the option to
avoid the generation of explicit constraints. The accuracy and efficiency of the new element has been
proven in the analysis of several large automotive models.


OUTLOOK

Fatigue analysis of spot welds requires high quality stresses. However, with the modeling technique
described here, the stresses in the shell elements around the spot weld grid points GA and GB are not
accurate enough. There are two ways to get more accurate stresses.

1. A new detailed model around the spot weld is created. A couple of elements of the coarse model
around the spot weld are cut out, a finer mesh is created, grid point forces and boundary conditions
of the coarse model are transferred to the fine model. Then a stress analysis is done on the fine
model, see for example [7].
2. For a quick estimate of the maximum stress in the shell, stress intensity factors have been developed,
see for example [8]. The stress values from the finite element patch are multiplied by stress intensity
factors to estimate stress peaks around the weld nugget.

Fatigue is just one of many opportunities to develop more tools for the spot weld modeling process.
14


REFERENCES

[1] MSC.Nastran Quick Reference Guide, Version 70.5, MSC.Software, Costa Mesa, Los Angeles,
CA, 1998.
[2] A.Jonscher, M.Lewerenz, C.Hoff, MSC.Nastrans New Spot Weld Element in the CAE
Process, Proceedings of the 2
nd
MSC Worldwide Automotive User Conference, Dearborn, MI,
2000.
[3] D.Heiserer, M.Chargin, J.Sielaff, High Performance, Process Oriented Weld Spot Approach,1
st

MSC Worldwide Automotive User Conference, Munich, Germany, 1999.
[4] MSC.AMS- FVA Automotive Modeling System, Full Vehicle Analysis, MSC.Software, Costa
Mesa, Los Angeles, CA, 1999.
[5] PAMCRASH Reference Manual, PAM-SCL Version 1998, ESI/PSI Wiesbaden, Germany.
[6] A.Jonscher, M.Lewerenz, G.Luehrs, The New Spot Weld Element in the CAE Process,
Proceedings of the 1
st
MSC Worldwide Automotive User Conference, Munich, Germany, 1999.
[7] X.Deng, W.Chen, G.Shi, Three-Dimensional Finite Element Analysis of the Mechanical Behavior
of Spot Welds, Finite Elements in Analysis and Design, Vol.35, 2000, pp.17-39.
[8] Shicheng Zhang, Recovery of Notch Stress and Stress Intensity Factors in Finite Element
Modeling of Spot Welds, Proceedings of NAFEMS World Congress 99, Newport, RI, USA,
April 1999.