Review of Solid Element

Formulations in LS-DYNA
Properties, Limits, Advantages, Disadvantages

LS-DYNA Forum 2011 > Entwicklerforum

Stuttgart, 12. Oktober 2011
tobias.erhart @ dynamore.de

Motivation
Solid elements are three-dimensional finite elements that can model solid bodies
and structures without any a priori geometric simplification.
No geometric, constitutive and loading
assumptions required.
Boundary conditions treated more realistically
(compared to shells or beams).
FE mesh visually looks like the physical system.
b t
but
g
effort: mesh p
preparation,
p
, CPU time,,
Higher
post-processing,
Expensive mesh refinement: Curse of dimensionality.
Often poor performance for thin-walled structures (locking problems).

Motivation
Applications

Foam structures
Rubber components
Cast iron parts
S lid b
Solid
barriers
i
Plastic parts
Bulk forming
Thick metal sheets
Elastic tools
Impact analysis

Overview
LS-DYNA Users manual: *SECTION_SOLID, parameter ELFORM
EQ
EQ.

-2:
2 ffully
ll integrated
i t
t d S/R solid
lid intended
i t d d ffor elements
l
t with
ith poor aspectt ratio,
ti
accurate formulation

EQ.

-1: fully integrated S/R solid intended for elements with poor aspect ratio,
efficient formulation

EQ.

1: constant stress solid element (default)

EQ
EQ.

2: fully integrated S/R solid

EQ.

3: fully integrated quadratic 8 node element with nodal rotations

EQ.

4: S/R q
quadratic tetrahedron element with nodal rotations

EQ. 10: 1 point tetrahedron

EQ. 13: 1 point nodal pressure tetrahedron
EQ. 15: 2 point pentahedron element
EQ. 16: 4 or 5 point 10-noded tetrahedron
EQ 17: 10-noded composite tetrahedron
EQ.
EQ. 115: 1 point pentahedron element with hourglass control

Hexahedra elements in LS-DYNA

ELFORM = 1

underintegrated constant stress

efficient and accurate
even works for severe deformations
needs hourglass stabilization:
choice of hourglass formulation
and values remains an issue

ELFORM = 2
selective reduced integrated brick element
(volumetric locking alleviated)
no hourglass stabilization needed
slower than ELFORM=1
too stiff in many situations,
especially for poor aspect ratios (shear locking)
more unstable in large deformation applications

Hourglass control for ELFORM=1

*HOURGLASS: IHQ = 15
viscous form (1,2,3) for higher velocities
stiffness form (4,5) for lower velocities
exact volume integration recommended (3,5)

*HOURGLASS: IHQ = 6
the QBI (Quintessential Bending Incompressible)
h
hourglass
l
control
t l by
b B
Belytschko
l t hk and
d Bi
Bindeman
d
hourglass stiffness uses elastic constants
recommended in most situations
sometimes
ti
modified
difi d QM makes
k sense
(watch hourglass energy)

*HOURGLASS: IHQ = 7/9

similar to type 6, but less experience
type
t
7 uses total
t t l deformation
d f
ti instead
i t d off updated
d t d
type 9 should provide more accurate results for distorted meshes

Property of ELFORM=2
Shear locking
pure bending modes trigger spurious shear energy
getting
tti worse for
f poor aspectt ratios
ti
y

0
x

xx 2y / lx , yy 0, xyy x / ly
Alleviation possibility 1: under-integration ELFORM = 1
Alleviation possibility 2: enhanced strain formulations
Jacobian matrix

xx 2y / lx , yy 0, xy ... x / lx

modified

ELFORM = -1 / -2

Solid element types -1 and -2

NEW: ELFORM = -1 / -2
Thomas Borrvall: "A heuristic attempt to reduce transverse shear locking in
fully integrated hexahedra with poor aspect ratio", Salzburg 2009
Modification of the Jacobian matrix: reduction of spurious stiffness
without affecting the true physical behavior of the element

J ijorig

xi
1
I
xIi jI jkI k jlI l 123
kl
j
8
aspect ratios between dimensions

mod
ij

1 I
I
xIi j jkI k jk jlI l jl 123
k jkl jl
8

Type -2: accurate formulation, but higher computational cost in explicit

Type -1: efficient formulation
CPU cost compared to type 2: ~1.2 (type -1), ~4 (type -2)

New hexahedra elements in LS-DYNA

ELFORM = -1
identical with type 2, but accounted for
poor aspect ratio on order to reduce
shear locking
efficient formulation
sometimes hourglass tendencies

ELFORM = -2
identical with type 2, but accounted for
poor aspect ratio on order to reduce
shear locking
accurate formulation
higher computational cost than type -1

Implicit elastic bending

clamped plate of dimensions 10x5x1 mm3

subjected to 1 Nm torque at the free end
E = 210 GPa
y
solution for end tip deflection:
analytical
0.57143 mm
convergence study
with aspect ratio 5:1 kept constant

Plastic bending

explicit plastic 3 point bending (prescribed motion)

plate of dimensions 300x60x5 mm3
*MAT_024 (aluminum)
convergence study - aspect ratio 4:1 kept constant

120x24x8 elements
60x12x4 elements
30x6x2 elements
15x3x1 elements

Plastic bending
Results
maximum energy
gy ((internal + hourglass)
g
)
1

elements through thickness

maximu
um energ
gy

300
250

bad
b
d convergence
of type 2
(stiff behavior)

200
150

ELFORM=2
100

shells #16
ELFORM=1

50

ELFORM=-1/-2

0
10

100

1000

10000

number of elements

100000

good convergence
with types 1, -1, -2

Plastic bending
CPU times

ELFORM = 1:
ELFORM = 2:
ELFORM = -1:
ELFORM = -2:
2

56
116
136
542

minutes
minutes
minutes
minutes
i t

ELFORM = 2

ELFORM = -2

ELFORM = -1
1

ELFORM = 1

ELFORM = -2 not efficient, ELFORM = -1 comparable to 2

Tube crash
element size: 3.5 mm
thickness: 2 mm

shells
type 16

solids
type 1
(tCPU=1.0)

solids
type 2
(tCPU=5.5)

solids
type -1
1
(tCPU=5.2)

solids
type -2
2
(tCPU=8.3)

Tube crash
Results
contact force

internal energy

Tetrahedra elements in LS-DYNA

ELFORM = 10

ELFORM = 13

Theoretical background
manual: "1 point nodal pressure tetrahedron for bulk forming"
paper: J
J. B
Bonett & A.J.
A J Burton.
B t
A simple
i l average nodal
d l pressure
tetrahedral element for incompressible dynamic explicit
applications.
pp
Comm. Num. Meth. Engrg.
g g 14: 437-449,, 1998
"... the element prevents volumetric locking by defining nodal
volumes
l
and
d evaluating
l ti average nodal
d l pressures iin tterms off
these volumes ...
... it can be used in explicit dynamic applications involving
(nearly) incompressible material behavior
(e.g. rubber, ductile elastoplastic metals) ..."

TET #13 = TET #10 + averaging nodal pressures

= TET #10 volumetric locking

Notched steel specimen

discretized quarter system:

5 mm

TET #13

10 mm

30 mm

u (t)
TET #10

*MAT_PIECEWISE_LINEAR_PLASTICITY
E

y
Et

=
=
=
=

206.9 kN/mm2
0.29
0.45 kN/mm2
0.02 kN/mm2 (nearly ideal plastic)

isochoric plastic flow

HEX #2

HEX #1
(IHQ 6)
(IHQ=6)

Notched steel specimen

von Mises
stresses
t

load-displacement curve:
should show a limit force

TET #13
volumetric
locking

forcce in kN

0 - 480
N/mm2

TET #10

TET #10

HEX #1
HEX #2

TET #13

HEX #2

0
0

0,1

0,2

0,3

displacement in mm

0,4

HEX #1
(IHQ=6)

Rubber block compression

HEX #1
(IHQ=6)

HEX #2

TET #10

TET #13

*MAT_MOONEY-RIVLIN_RUBBER
A
B

=
=
=
=

4.0
4
0 N/
N/mm2
2.4 N/mm2
0.499
1.5E-06 kg/mm3

nearly incompressible material

Rubber block compression

deformation

HEX #1
(IHQ=6)

HEX #2

von Mises stresses (0

( 1.2 N/mm2)

TET #10

TET #13

Taylor bar impact

*MAT_PIECEWISE_LINEAR_PLASTICITY:
= 8930 kg/m3, E = 117 kN/mm2, = 0.35, y = 0.4 kN/mm2, Et = 0.1 kN/mm2

deformation
HEX #1 (IHQ=6)

12.6 mm (QM=0.10)
13.8 mm (QM=0.01)

HEX #2

TET #10

14.1
14 1 mm

10.4
10 4 mm

TET #13

13.9
13 9 mm

Taylor bar impact

pressure (-300 300 N/mm2)
HEX #1 (IHQ=6)

HEX #2
TET #10

checkerboard mode

TET #13

Structural component
von Mises stresses

load-displacement curve
500

force in
n kN

400

x - stresses

TET #10

300
200

TET #13

100
0
0

25

50

75

100

time in ms

125

150

Higher order tets in LS-DYNA

ELFORM = 16

4(5)
( ) point 10-noded tetrahedron
good accuracy for moderate strains
high cpu cost
observe the node numbering
use *CONTACT_AUTOMATIC_... With PID
easy conversion of 4-noded tets via
*ELEMENT_SOLID_TET4TOTET10
full output with TET10=1 on *CONTROL_OUTPUT

ELFORM = 17
4(5) point 10-noded composite tetrahedron
(12 linear sub-tetrahedrons)
properties similar to type 16
correct external force distribution

Plastic bending
Explicit plastic 3 point bending (prescribed motion)
plate of dimensions 300x60x5 mm3
p
*MAT_024 (aluminum)

256300 elements
33700 elements
4300 elements

Plastic bending
Results
maximum energy (internal)

400

maximu
um energ
gy

ELFORM=10
300

200

ELFORM=13

bad convergence
of type 10
(stiff behavior)

ELFORM=16 ELFORM=2

g
good
convergence
g
with types 13,16,17

100

0
2000

20000

ELFORM=17

200000

number of elements

2000000

Plastic bending
Results
maximum energy (internal)
140
ELFORM=10

maximu
um enerrgy

120
100

ELFORM=13

80

O
ELFORM=17

ELFORM=16

ELFORM=2

Tet type 13
comparable to
types 16 / 17

60
40
20
0
0

50

100

150

CPU time [min]

200

Hex and Tet with nodal rotations

ELFORM = 3
quadratic 8 node hexahedron with
nodal rotations, i.e. 6 DOF per node
derived from 20 node hexahedron
u integration
teg at o ((12-point)
po t)
full
well suited for connections to shells
good accuracy for small strains
tendencyy to volumetric locking
g

ELFORM = 4
quadratic 4 node tetrahedron with
nodal rotations, i.e. 6 DOF per node
derived from 10 node tetrahedron
S/R integration (5-point)
well suited for connections to shells
good accuracy for small strains
tendency to volumetric locking

Structural component
3 different discretizations
ELFORM =

10

13

16

17

18

39

66

2-10 mm, 7500 elements

CPU times in minutes

1.5-7 mm, 27000 elements

97

14

22

212

344

403

64

98

945

1529

1-5 mm, 77000 elements

Structural component
load-displacement curve
250
TET #4 (2-10 mm)

force in kN
N

225

TET #10 (1-5 mm)

200

TET #17 (2-10

(2 10 mm)
TET #16 (2-10 mm)
TET #13 (1.5-7 mm)

175
150
125
0

20

40

60

time in ms

80

100

Structural component
Maximum principal stress (-50.0 450.0 N/mm2)

424 N/mm2

TET #4
(2-10 mm, 18 min.)

437 N/mm2

TET #13
(1.5-7 mm, 22 min.)

509 N/mm2

TET #10
(1-5 mm, 64 min.)

450 N/mm2

TET #16/17
(2-10 mm, 39/66 min.)

Pentahedra elements in LS-DYNA

ELFORM = 15
2 point selective reduced integration
needs hourglass stabilization for twist mode
(recent improvement next official versions)
often used as transition element
(ESORT=1)

ELFORM = 115 (new in next official versions)

1 point reduced integration
needs hourglass stabilization
(analogue to hexahedron element type 1
with Flanagan
Flanagan-Belytschko
Belytschko hourglass
formulation)

Time step control

critical time step:

adiabatic sound speed:

characteristic element length

Time step control

Example 1: Time step for solid elements with same volume
1.67
1.42
1.0
0 68
0.68

1/2/3/
-1/-2

10,13

15

0 65
0.65
16

0.56
17

Example 2: Time step for solid elements with same edge length
10
1.0
0.69
1/2/3/
-1/-2

0.82
0.52
10,13

15

0.32
16

0.27
17

Conclusions & Remarks

Always set ESORT = 1 on *CONTROL_SOLID
Use hexahedron elements if possible (regular solid bodies)
- ELFORM = 1 with IHQ = 6 or ELFORM = 2, 3
- ELFORM = -1 or -2 for flat hexas
For complex solid structures, use tetrahedrons type 4, 13, 16, or 17
- ELFORM = 16/17 are the most accurate tets, but not suited for large strains
- ELFORM = 13 needs finer mesh,
mesh well
ell suited
s ited even
e en for large strains
(check if your material is supported)
For metals or plastics (moderate strains)
strains), use tet type 4
4, 13
13, 16
16, or 17
For rubber materials (incompressible, large strains) use tet type 13
For bulk forming problems, use ELFORM = 13 and r-adaptivity
Pentahedrons 15/115 should only be used as transition elements