Advanced Nonlinear TMG 201 400

Chapter 3: Material models and formulations
Shaded area is area

enclosed by strain
q , x1, e1P memory surface
q
e1P
x1
Time
Fig. 3.4-13: Evolution of strain memory surface in

uniaxial cyclic straining
Isotropic hardening rules: The isotropic hardening rules are

shown in Fig. 3.4-14.
Bilinear hardening:
 y  0 y  EP t t e P
t t
where 0 y and EP are material constants, and t t

e P is the
t t
accumulated effective plastic strain ( p in ref LC). e P is
t t 2 P
calculated using e P  t e P  e P , where e P  Δe : Δe P .
3
EET
For bilinear hardening, EP  where ET is the slope of
E  ET
the stress-strain curve during plasticity (this formula assumes no
kinematic hardening).
192 Advanced Nonlinear Solution  Theory and Modeling Guide

3.4: Isothermal plastic material models
sy q=q2
q=q1 R
Memory-exponential
Q(q2 ) hardening
Q(q1)
sy
0
Q Exponential hardening
sy
0
Multilinear hardening
(ePi , syi)
Bilinear hardening
0
sy EP
eP
Fig. 3.4-14: Dependence of stress radius on accumulated
effective plastic strain
Multilinear hardening:
Pairs of (e P ,  y ) are given. It is allowed for  y to either increase

or decrease as e P increases.
If the multilinear hardening curve is given in terms of the
uniaxial response (e,  ) where  is the tensile stress
corresponding to the unaxial strain e , then
 P  
 e  e  ,  y    are the corresponding points for the points
 E 
(e,  ) . This formula assumes no kinematic hardening.
Exponential hardening:
t t
 y  0 y  Q(1  exp(b t t e P ))
Advanced Nonlinear Solution  Theory and Modeling Guide 193

where Q and b are material constants. Q can be positive to

model cyclic hardening, and Q can be negative to model cyclic
softening.
Memory-exponential hardening (exponential hardening with strain

memory surface):
In this model, the yield surface size  y depends on the strain

t t
memory size t t q (see above for a description of the strain

memory surface).
The yield surface size is
 y  0 y  t t R
t t
where t t R is the change in the size of the yield surface. t t
R is
defined using the differential equation
dR  b( t Q  t R)de P
t t
and is calculated using R  t R  R ,
R  b( t t Q  t t R)e P . t t Q is the asymptotic change in
yield surface size, and is calculated as
t t
Q  Q( t t q )  QM  (QM  Q0 ) exp(2 t t q )
where t t q is the size of the strain memory surface.

The material constants for the memory-exponential material
model are 0 y , b , Q0 , QM , b ,  and the strain memory surface
parameter  .

Kinematic hardening rule: The kinematic hardening rule is:
Armstrong-Fredrick nonlinear kinematic hardening:
The back stress is expressed as a sum of partial back stresses
t t
α   t t α ( m )
where t t α ( m ) is partial back stress number m. All of the partial

back stresses are independent of each other.
Each partial back stress evolves according to the differential
rule
2 (m) P
dα ( m )  h de   ( m ) t α ( m ) de P
3
where h ( m ) and  ( m ) are material constants (these material

constants are Cl and  l for partial back stress number l in ref LC).
Assuming that the directions of plastic strain increments are
constant during a time step, this can be integrated to obtain
t t
α ( m )  A ( m )  ( A ( m )  t α ( m ) ) exp(b( m ) )
2 h( m )
where A  (m)
(directions of ΔeP ) , b( m )   ( m ) e P .
3 (m)
It is allowed to use one partial back stress, with   0 . Then

linear kinematic hardening is recovered, with E p  h (this formula
assumes no isotropic hardening).
Stress  plastic strain curve for uniaxial cycling
In uniaxial cycling, a typical stress – plastic strain curve is

shown in Fig. 3.4-15. Here only one partial back stress is used with
material constants h and  . Since a stable cycle is considered
here,  y is taken from the isotropic hardening rule assuming a
very large value of e P . The hardening modulus is given by

EP  h      y  on segment A-B. The size of the plastic

strain cycle can be related to the material constants using the
relation
 h  e P 
 tanh     y
2   2 
Thus, given several cycles for different cyclic strains, the material
constants can be estimated using the above formula.
s
B
A EP 2sy
Ds
eP
DeP
Fig. 3.4-15: Stable uniaxial plastic cycle using

one term in Armstrong-Fredrick nonlinear
kinematic hardening rule
Stress  plastic strain curve for initial loading
During initial loading, it can be shown that
 y 
h

1  exp  e 
P
The corresponding stress  plastic strain curve is shown in Fig.

3.4-16. Here only one partial back stress is used with material
constants h and  . In this formula it is assumed that there is no

isotropic hardening, and that the yield stress is  y . Hence the

hardening observed during initial loading can be modeled using
nonlinear kinematic hardening, without the use of isotropic
hardening.
s
h/z
sy
eP
Fig 3.4-16: Stress - plastic strain curve during initial loading,

using nonlinear kinematic hardening
Combination of isotropic and kinematic hardening rules: By

combining the Armstrong-Fredrick nonlinear kinematic hardening
rule with the isotropic hardening rules, a wide variety of cyclic
phenomena can be simulated, such as cyclic hardening and
softening, shakedown and ratcheting. See ref LC for details.
Stress-strain integration: When plasticity is detected, the

incremental plastic strains are solved for using an iterative solution
procedure. The maximum number of iterations can be specified,
and the tolerance used in assessing convergence can be specified.
Constitutive tensor: The constitutive tensor (stress-strain matrix)

is constructed to be tangent. However, in general, the tangent
constitutive tensor is non-symmetric, therefore the constitutive
tensor is symmetrized.
Formulations: When used with the small displacement/small

strain formulation, a materially-nonlinear-only formulation is
employed, when used with the large displacement/small strain
formulation, a TL formulation is employed, and when used with the
large displacement/large strain formulation, either a ULH or ULJ
formulation is employed.

If geometrically nonlinear effects are to be included, the large

displacements/large strain kinematics are preferred to the large
displacement/small strain kinematics, even when the strains are
numerically small. The large displacement/small strain kinematics
should be used only when the large displacement/large strain
kinematics are not supported by the element.
Mixed displacement-pressure formulation: Plane strain,

axisymmetric or 3-D solid elements that reference this material
model should use the mixed displacement-pressure (u/p) element
formulation. This is because the plastic strains are incompressible.
This is done by setting UPFORM = 1 in the NXSTRAT command.
Thermal strains: If a thermal load is applied to the structure, the

thermal strains are taken into account but the material
characteristics are considered to be temperature independent.
Modeling of rupture: Rupture conditions can also be included

with this model.
The maximum accumulated effective plastic strain can be
specified for the rupture condition. When rupture is reached at a
given element integration point, the corresponding element is
removed from the model (see Section 10.4).
3.4.2.2 Specification of input
MATPLCY entry
Parameters ISO, KIN, RUP in this entry reference data defined in

entries PLCYISO, PLCYKIN, PLCYRUP respectively. These
entries provide the input for the isotropic, kinematic and rupture
parts of the plastic-cyclic model, see below for a detailed
description of these commands.
It is allowed to not specify a value for KIN or RUP. Then these
effects are not included in the model.
BETA is the integration factor  described above. When
BETA is blank (the default), the program chooses the value of 
as follows:  =1 for static or implicit time integration (Solution
601);  =0 for explicit time integration (Solution 701). MAXITE
is the maximum number of iterations used to solve for the

incremental plastic strains. RTOL is a tolerance used to assess

convergence of the iterations. RTOL can be thought of as a
reference incremental plastic strain.
PLCYISO entry
Parameter TYPE=BLIN, MTLIN, EXP, MEMEXP specifies the

isotropic hardening type as bilinear, multilinear, exponential or
memory-exponential respectively. The remaining parameters of
this entry specify the material constants for the isotropic hardening.
PLCYKIN entry
The parameters of this entry are used to specify the constants for
the kinematic part of the plastic-cyclic model.
PLCYRUP entry
The parametrs of this entry are used to specify the constants for the
rupture part of the plastic-cyclic model. Rupture is based on the
accumulated effective plastic strain.
Conversion formulas
Plastic-bilinear to plastic-cyclic:
In general, given a plastic-bilinear material with given yield stress

and tangent modulus ET, this material can be converted into an
equivalent plastic-cyclic material with bilinear isotropic hardening,
(E)(ET)
the same yield stress, and plastic modulus EP  .
E - ET
Plastic-multilinear to plastic-cyclic:
In general, given a plastic-multilinear material with Young's

modulus E and stress-strain points
strain(i) stress(i)
this material can be converted into an equivalent plastic-cyclic

material with multilinear isotropic hardening and multilinear points

aeps(i) sr(i)
using the formulas
aeps(i)  strain(i)  (stress(i) / E) , sr(i)  stress(i)
Note that aeps at the elastic limit will always be zero.
3.5 Temperature-dependent elastic material models

 The thermal isotropic and thermal orthotropic material models
are discussed in this section.
The thermal isotropic material is obtained with the MAT1 and
MATT1 material entries.
The thermal orthotropic material is obtained for surface
elements with the MAT2 and MATT2 material entries, or MAT8
and MATT8 material entries; for 2-D elements with MAT3 and
MATT3 material entries; and for solid elements with the MAT9
and MATT9 material entries, or MAT11 and MATT11 material
entries.
These commands allow the different elastic material constants
to vary with temperature. Thermal strains are taken into account in
these materials.
 The thermal isotropic model is available for the rod, 2-D solid,
3-D solid and shell elements.
 The thermal orthotropic model is available for the 2-D solid,

3-D solid and shell elements.
 Both models can be used with small displacement/small strain

and large displacement/small strain kinematics. The strains are
always assumed to be small.
When used with small displacement/small strain kinematics, a
materially-nonlinear-only formulation is employed.
When used with large displacement/small strain kinematics,
either the TL or UL formulation is employed. 2-D, 3-D solids and
shells use the TL formulations, and rods use a UL formulation.
 In the data input for the analysis, the nodal point temperatures
must be defined for all time steps. See Section 5.6.

3.5: Temperature-dependent elastic material models
 For these models, the elastic moduli, the shear moduli, the
Poisson's ratios and the coefficients of thermal expansion defined
in Section 3.2 are input as piecewise linear functions of the
temperature, as illustrated in Fig. 3.5-1. Linear interpolation is used
to calculate the material properties between input points.
E n
l l
tn
t
E l l
l l l l
q1 tq q2 q3 q4 q1 tq q2 q3 q4
Temperature Temperature
a
l
ta
l
l l
q1 tq q2 q3 q4
Temperature
Fig. 3.5-1: Variation of material properties for thermo-elastic model
 The calculation of thermal strains is described in Section 3.1.6.
 Note that if the material constants are all temperature

independent, and the material is isotropic, then thermal strains
could alternatively have been modeled using the elastic isotropic
material (non-thermal) of Section 3.2.
 For the evaluation of the temperatures  and 0 at the

t
integration point considered, the isoparametric interpolation

functions hi are used; e.g., in two-dimensional analysis we have
q
   hi i
i 1
where  i is the temperature at element nodal point i (see Fig.

3.5-2). Note that when higher-order elements are used the

temperatures at the integration point can be significantly different
from the values at the nodal point (for example negative although
all nodal point temperatures are greater than or equal to zero).
s
l
l
l
l
l
r
l
Integration points
1 l
Temperature t q1 l
Fig. 3.5-2: Interpolation of temperature at

integration points
 For shell elements, the temperature at an integration point is

established using the temperatures at the midsurface nodes and the
temperature gradients. Note that the temperature gradient at a shell
midsurface node is defined in the direction of the mid-surface
director vector at that node, see Section 2.3.3.
3.6 Thermal elasto-plastic and creep material models

 This section groups together thermal elasto-plastic materials and
creep materials, since a unified general solution can be applied to
these material types. The computational procedure is based on the
effective stress function algorithm, detailed in Section 3.6.4.
ref. KJB
 The thermal elasto-plastic and creep models include the effects
Section 6.6.3 of
 Isotropic elastic strains, via the MAT1 entry
t TH
 Thermal strains, ers , via the MATT1 or the MAT1 entries.
 Time-independent plastic strains, t ersp , via the MATS1 entry

 Time-dependent creep strains, t ersC , via the CREEP entry,
the CREEP and MATTC entries, or the MATCRP entry.

3.6: Thermo-elasto-plastic and creep material models
 The constitutive relation used is
t
ij  t Cijrs
E
 t ers  t ersP  t ersC  t ersTH  (3.6-1)
where t ij is the stress tensor at time t and t Cijrs

E
is the elasticity
tensor at the temperature corresponding to time t. The tensor t Cijrs
E
can be expressed in terms of Young's modulus tE and Poisson's

ratio tv both of which may be temperature-dependent.
 Note that the thermal, plastic and creep parts of these material
models are optional. If, however, the omitted strain components
result in one of the material models detailed in one of the previous
sections, then the program will select that material model.
 The formulations provided in this section are very general, and

can describe any material combining elastic, plastic, thermal and
creep strains. The combinations given in Table 3.6-1 are allowed.
 These material models can be used with the rod, 2-D solid, 3-D
solid, and shell elements.
 These models can be used with small displacement/small

strain, large displacement/small strain and large
displacement/large strain kinematics.
When used with small displacement/small strain kinematics, a
materially-nonlinear-only formulation is employed.
When used with large displacement/small strain kinematics,
either a TL or a UL formulation is employed (TL for 2-D and 3-D
solids and shells, and UL for rods).
When used with large displacement/large strain kinematics, the
ULH (updated Lagrangian Hencky) formulation is employed. This
is only supported for 2-D solid and 3-D solid elements.
 If geometrically nonlinear effects are to be included, the large

displacements/large strain kinematics are preferred to the large
displacement/small strain kinematics, even when the strains are
numerically small. The large displacement/small strain kinematics
should be used only when the large displacement/large strain
kinematics are not supported by the element.

Table 3.6-1: Combinations of elastic, plastic and creep strains
Description Elastic Plastic Creep Bulk data entries

Elastic Yes No Yes MAT1, CREEP, or MAT1, MATCRP
creep
Thermal Yes No Temp- MAT1, CREEP, MATTC, or MAT1, MATCRP
elastic dep
creep
Thermal Temp- Yes No MAT1, MATT1, MATS1, with TID in MATS1
elasto- dep pointing to a TABLES1 entry
plastic Yes Temp- No MAT1, MATS1, with TID in MATS1 pointing to
dep a TABELST entry
Temp- Temp- No MAT1, MATT1, MATS1, with TID in MATS1
dep dep pointing to a TABELST entry
Plastic- Yes Yes Yes MAT1, MATS1, CREEP, with TID in MATS1
creep pointing to a TABLES1 entry
Temp- Yes Yes MAT1, MATT1, MATS1, CREEP, with TID in
dep MATS1 pointing to a TABLES1 entry
Thermal Yes Temp- Yes MAT1, MATS1, CREEP, with TID in MATS1
plastic- dep pointing to a TABLEST entry
creep Temp- Temp- Yes MAT1, MATT1, MATS1, CREEP, with TID in
dep dep MATS1 pointing to a TABLEST entry
Yes Yes Temp- MAT1, MATS1, CREEP, MATTC, with TID in
dep MATS1 pointing to a TABLES1 entry
Temp- Yes Temp- MAT1, MATT1, MATS1, CREEP, MATTC,
dep dep with TID in MATS1 pointing to a TABLES1
entry
Yes Temp- Temp- MAT1, MATS1, CREEP, MATTC, with TID in
dep dep MATS1 pointing to a TABLEST entry
Temp- Temp- Temp- MAT1, MATT1, MATS1, CREEP, MATTC,
dep dep dep with TID in MATS1 pointing to a TABLEST
entry
Notes:
1. "No" means that this strain is not included in the material. "Yes" means that this strain is
included in the material description, and that the material constants for this strain are
temperature-independent. "Temp-dep" means that this strain is included in the material
description, and that the material constants for this strain are temperature-dependent.
2. Instead of using TID in MATS1 pointing to a TABLES1 entry, it is also allowed to specify
H and LIMIT1 in MATS1. The resulting material uses bilinear hardening, however the yield
stress and hardening modulus are temperature-independent.

 Plane strain, axisymmetric or 3-D solid elements that reference

these material models should preferably employ the mixed u/p
element formulation. This is done by setting UPFORM=1 in the
NXSTRAT entry.
 Note that the constitutive relations for the thermal, plastic and
creep strains are independent of each other; hence the only
interaction between the strains comes from the fact that all strains
affect the stresses according to Eq. 3.6-1. Fig. 3.6-1 summarizes the
constitutive description for a one-dimensional stress situation and a
bilinear stress-strain curve.
 Since there is no direct coupling in the evaluation of the

different strain components, we can discuss the calculation of each
strain component independently.
ref. M.D. Snyder and K.J. Bathe, "A Solution Procedure for
Thermo-Elastic-Plastic and Creep Problems," J. Nuclear
Eng. and Design, Vol. 64, pp. 49-80, 1981.
ref. M. Kojić and K.J. Bathe, "The Effective-Stress-Function

Algorithm for Thermo-Elasto-Plasticity and Creep," Int.
J. Numer. Meth. Engng., Vol. 24, No. 8, pp. 1509-1532,
1987.
 In multilinear plasticity, the rupture plastic strain corresponds to

the effective plastic strain at the last point input for the stress-strain
curve.
 When rupture is reached at a given element integration point,

the corresponding element is removed from the model (see Section
10.4).

Area A
l l
tq tP
Temperature
(a) Model problem of rod element under constant load
Stress
ts
ts (t q)
T( q)
tE t
yv
tE(tq)
Strain
teP (time independent)
teE (time independent)
Creep
strain
teC (time dependent)
t Time
(b) Strains considered in the model
Fig. 3.6-1: Thermo-elasto-plasticity and creep constitutive

description in one-dimensional analysis

3.6.1 Evaluation of thermal strains
 The thermal strains are calculated as described in Section 3.1.6.
3.6.2 Evaluation of plastic strains
 Plasticity effects are included in the thermal elasto-plastic

material model and is based on the von Mises yield criterion, an
associated flow rule, isotropic or kinematic hardening (no mixed
hardening), and bilinear or multilinear stress-strain curves (based
on the H, LIMIT1 and TID fields in MATS1).

 The plastic strains are calculated using the von Mises plasticity
model (see Section 3.4) with temperature-dependent material
parameters (Young's modulus, Poisson's ratio, stress-strain curves,
...).
 The yield function is, for the case of isotropic hardening
1t t 1t 2
t
fy  s  s   yv
2 3
and for the case of kinematic hardening
1 t 1
t
fy 
2
 s  t α    t s  t α   t y2v
3
where t s is the deviatoric stress tensor, t yv is the virgin yield

stress corresponding to temperature t  and t α is the shift of the
stress tensor due to kinematic hardening.
 The expressions for plastic strain increments resulting from the

flow theory are deijP  d  t sij for isotropic hardening and
deijP  d   t sij  t ij  for kinematic hardening, in which d  is the
plastic multiplier (positive scalar) which can be determined from
the yield condition t f y  0 . In the case of kinematic hardening, we
express the change of the yield surface position in the form

d ij  t CdeijP
where t C is the modulus
2 t E t ET 2
t
C  t EP
3 E  ET 3
t t
In the case of multilinear yield curves, t ET represents the tangent

modulus of the segment on the stress - total strain yield curve
corresponding to the accumulated effective plastic strain t e P , and
t
EP represents the tangent modulus of the segment on the stress -
plastic strain yield curve corresponding to the accumulated
effective plastic strain t e P .
 When H and LIMIT1 are specified in MATS1, bilinear

hardening plasticity is assumed, in which only the elastic material
parameters can be temperature dependent (Young’s modulus,
Poisson’s ratio and coefficient of thermal expansion). The yield
stress and hardening modulus are temperature independent.
 When TID in MATS1 points to a TABLES1 entry, multilinear

hardening plasticity is assumed, in which only the elastic material
parameters can be temperature dependent. The yield curve is
temperature independent.
 When TID in MATS1 points to a TABLEST entry, multilinear

hardening plasticity is assumed, in which the elastic material
parameters, and also the yield curves, can be temperature
dependent
The yield curves are interpolated as shown in Fig. 3.6-2.

Stress 3 4
2
1
qi
e 1i e i2 ei3 e4i Strain
3 4
1 2
qi+1
3
1
e i+1 2
e i+1 e i+1 e 4i+1
Temperature
a) Stress-strain curves input data
Stress 4q
3 i
2
Interpolated
yield curve 1
tq
4 Plastic
3
1 2 qi+1 strain
P2 P3
e i+1 e i+1
Temperature
b) Yield curves
Fig. 3.6-2: Interpolation of multilinear yield curves with temperature

 The formula d ij  Cdeij used in kinematic hardening

t P
requires modification to avoid nonphysical effects when t EP , and

hence t C , is a function of temperature.
 Here is an example showing a nonphysical effect that occurs

when the formula d ij  t CdeijP is used without modification.
Consider the following material description, in which kinematic
hardening is used:
Temperature E   yv ET EP
0 1E6 0 100 1996.008 2000
200 1E6 0 100 999.001 1000
The yield curves for this material are shown in Fig. 3.6-3. Take a
uniaxial specimen, set the temperature to 0 and load with
prescribed force until the uniaxial stress is 130. The corresponding
plastic strain is e P  0.015 (since 100  2000e P  130 ). This
point is labeled A in Fig. 3.6.3.
Now, without changing the prescribed force, change the
temperature to 200. For the plastic strain e P  0.015 , the yield
stress is 100  1000e P  115 , which is less than the stress of 130.
So we would expect that the material would plastically deform
further until the (plastic strain, stress) point is on the yield curve
corresponding to temperature 200 (point B in Fig. 3.6-4). However
the plastic strain does not change and the current stress remains
above the yield curve for temperature 200 (point A).
The reason for this unexpected behavior is as follows: The yield
condition involves the initial yield stress, which is unchanged, the
current stress, which is unchanged and the back stress, which is
also unchanged (if the back stress had changed, then the yield
condition would not be satisfied). Since the change in plastic strain
is related to the change in back stress through d ij  t CdeijP ,
because the back stress does not change, the plastic strain does not
change either.

Stress
q 1 =0
EP =2000
q 2 =200
sy =130 A B EP =1000
syv =100
eP=0.015 eP=0.030 Plastic strain
Fig. 3.6-3: Example showing nonphysical results in kinematic

hardening when EP is a function of temperature
 Chaboche has discussed the formulation of kinematic hardening

under varying temperature conditions, and has pointed out other
nonphysical effects when the above back stress evolution rule is
used in varying temperature conditions, see Section 4.4 of the
following reference:
ref. J.L. Chaboche, “Cyclic viscoplastic constitutive

equations, part I: A thermodynamically consistent
formulation”, J. Appl. Mech, December 1993, Vol 60, pp
813-821.
When the yield curve is a straight line for each temperature (that is,
when t C is a function of temperature but not of plastic strain),
Chaboche suggests a rule equivalent to
 ij  tC t eijP
t
which in differential form becomes
d ij  t C deijP  t eijP dC
It is seen that the first term on the right-hand-side corresponds to

the formulation presented above. The second term on the right-
hand side corresponds to a “back stress temperature correction”,

since this term is nonzero only if the temperature is varying.

When the yield curve is not a straight line for each temperature
(that is, when t C is a function of temperature and of plastic strain),
we again include a temperature correction corresponding to the
second term on the right-hand side, but in which dC in the second
term is evaluated only due to changes in temperature.
When the temperature correction is included in the above
example, the plastic strain in the above example increases to point
B when the temperature is increased to 200, so that the current
stress remains on the yield curve. This response is the expected
behavior.
 The back stress temperature correction can be selected using the

BSTC parameter of the NXSTRAT entry. BSTC=0 means to not
use the back stress temperature correction (the default). BSTC=1
means to use the back stress temperature correction.
3.6.3 Evaluation of creep strains
 Two creep laws are currently available in Solution 601. The first
called the Power creep law is obtained by setting TYPE = 300 in
the CREEP material entry, or by setting TYPE = 301 in the
MATCRP material entry. The second creep law called the
Exponential creep law is obtained by setting TYPE = 222 in the
CREEP material entry. The Power creep law is currently supported
for the elastic-creep, thermal elastic-creep, plastic-creep and
thermal plastic-creep material models. The Exponential creep law
is currently supported only for the elastic-creep and plastic-creep
material models.
 The effective creep strain is calculated as follows:
Power creep law (creep law 1) :
t
e C  a  t b  t d
in which σ is the effective stress, t is the time, and a, b, d are

material constants from the CREEP or MATCRP material
entries. These three constants can be set to be temperature
dependent via the MATTC entry.

Exponential creep law (creep law 2) :
t
e C  F    1  e   R    t
  G    t
with
 
F    a  eb   ; R    c   t   ; G    e  e f  
t d
 t

in which a through f are material constants from the CREEP
material entry.
 The creep strains are evaluated using the strain hardening
procedure for load and temperature variations, and the O.R.N.L.
rules for cyclic loading conditions.
ref. C.E. Pugh, J.M. Corum, K.C. Liu and W.L. Greenstreet,
"Currently Recommended Constitutive Equations for
Inelastic Design of FFTF Components," Report No. TM-
3602, Oak Ridge National Laboratory, Oak Ridge,
Tennessee, 1972.
The procedure used to evaluate the incremental creep strains is

summarized in the following: Given the total creep strains t eijC and
t t
the deviatoric stresses sij ,
1) Calculate the effective stress
1
3 2
t t
   t t sij t t sij 
2 
2) Calculate the pseudo-effective creep strain
1
2 2
t
e    t eijC  eijorig  t eijC  eijorig  
C
3 
3a) For power creep with temperature-independent material

constants, calculate the effective creep strain and effective creep

strain rate at time t  t using
 eC   teC    a0 t t a1 
1/ a2 1/ a2 1/ a2
t t
t
t t
e C t e C
t t
e C 
t
3b) For other creep laws (including power creep with

temperature dependent constants), calculate the pseudo-time t
satisfying
e C  t t  t t  t  t e C  e C  t t  t t  t  t
where e C  t t
  t t  t  is the generalized uniaxial creep law
d e C  t t  t t  t  . Then calculate
and e C  t t  t t  t  
dt
the effective creep strain and effective creep strain rate at time
t  t using
t t
e C  e C  t t  t t  t  , t t
e C  e C  t t  t t  t  .
4) Calculate t t
 using
 t t
e C
t t

 t t

5) Calculate the incremental creep strains using
eijC  t t t
 t t
sij
The use of the pseudo-time in step 3b corresponds to a strain

hardening procedure. See ref. KJB, pp 607-608 for a discussion of
strain hardening for calculation of creep strains.

3.6.4 Computational procedures
ref. KJB  The stresses and strains at the integration points are evaluated
Section 6.6.3 using the effective-stress-function algorithm.
ref. M. Kojić and K.J. Bathe, "The Effective-Stress-Function

Algorithm for Thermo-Elasto-Plasticity and Creep," Int.
J. Numer. Meth. Engng., Vol. 24, No. 8, pp. 1509-1532,
1987.
Briefly, the procedure used consists of the following calculations.
The general constitutive equation
t t
 (i )  t t C E  t t e(i )  t t e P (i )  t t eC (i )  t t eTH  (3.6-1)
is solved separately for the mean stress and for the deviatoric
stresses. In this equation the index (i) denotes the iteration counter
in the iteration for nodal point equilibrium. For easier writing this
index will be dropped in the discussion to follow. The mean stress
is calculated as
t t
E
t t
m 
1 2 t t

 t t
em  t t eTH  (3.6-2)
The deviatoric stresses t t s depend on the inelastic strains and

they can be expressed as
1
t t
s 
 t t e  1    t    t s  (3.6-3)
t t
aE   t   
t t
t t E t
where aE  , s = deviatoric stress at the start of the
t t
1 
time step and  is the integration parameter used for stress
evaluation  0    1 . The creep and plastic multipliers   and
 are functions of the effective stress t t
 only, and they
account for creep and plasticity; also

t t
e  t t e  t e P  t eC
is known since the deviatoric strains t t e , plastic strains t e P and

creep strains t eC are known from the current displacements and
the stress/strain state at the start of the current time step.
The following scalar function f  t t
  is obtained from Eq.
(3.6-3)
f  t t   a 2 t t   b    c 2     d 2  0 (3.6-4)
The zero of Eq. (3.6-4) provides the solution for the effective stress
t t
 , where
a  t t aE   t    
b  3 1   t t t eij t sij
c  1    t t
3 t t
d2  eij t t
eij
2
with summation on the indices i, j.

Once the solution for t t  has been determined from Eq.
(3.6-4), simultaneously with the scalars   and  from the creep
and plasticity conditions, the deviatoric stress t t s is calculated
from Eq. (3.6-3), and the plastic and creep strains at the end of the
time step are obtained as
e  t e P   t t s
t t P
e  t eC  (1    t s  
t t C t t
s  t  
The above equations correspond to isotropic hardening

conditions and a general 3-D analysis. The solution details for
kinematic hardening conditions and for special problems (for the
plane stress and shell elements) are given in the above cited
references, and also in the following reference:

ref. M. Kojić and K.J. Bathe, "Thermo-Elastic-Plastic and

Creep Analysis of Shell Structures", Computers &
Structures, Vol. 26, No 1/2, pp. 135-143, 1987.
3.7 Hyperelastic material models

 The hyperelastic material models available in Advanced
Nonlinear Solution are the Mooney-Rivlin, Ogden, Arruda-Boyce,
Hyperfoam, and Sussman-Bathe material models. They are all
defined using the MATHE command. In addition MATHP can be
used to define a hyperelastic Mooney-Rivlin material.
 This material model can be employed with the 2-D solid, 3-D
solid and 3D-shell elements.
 This material model uses large displacement/large strain

kinematics. A total Lagrangian (TL) formulation is employed. The
same formulation is used if a large displacement/small strain
kinematics is selected.
 Viscoelastic effects and Mullins effects can be included using

the MATHEV and MATHEM entries.
 Thermal strains can be included via a constant thermal

expansion coefficient. Section 3.7.6 shows how thermal strains are
computed for hyperelastic materials.
 In Solution 701 only the Mooney-Rivlin, Ogden and Sussman-

Bathe material models can be used, and only for 3-D solid
elements.
 The isotropic hyperelastic effects are mathematically described

by specifying the dependence of the strain energy density (per unit
original volume) W on the Green-Lagrange strain tensor  ij .
 We now give a brief summary of the quantities and concepts

used. For more information, refer to ref KJB, section 6.6.2. Here
and below, we omit the usual left superscripts and subscripts for
ease of writing. Unless otherwise stated, all quantities are evaluated

at time t and referred to reference time 0 .
 Useful quantities are the Cauchy-Green deformation tensor Cij ,

given by
Cij  2 ij   ij (3.7-1)
where  ij is the Kronecker delta; the principal invariants of the

Cauchy-Green deformation tensor,
1 2
I1  Ckk , I2 
2
 I1  Cij Cij  , I 3  det C (3.7-2a,b,c)
the reduced invariants:
1 2 1
 
I1  I1 I 3 3 , I 2  I 2 I 3 3 , J  I 3 2 , (3.7-3a,b,c)
the stretches i where the i ’s are the square roots of the principal
stretches of the Cauchy-Green deformation tensor; and the reduced
stretches:
1
i  i  12 3 

3 (3.7-4)
Note that
J  12 3 (3.7-5)
is the volume ratio (ratio of the deformed volume to the

undeformed volume).
 The strain energy density W is written either in terms of the

invariants or in terms of the stretches. In many cases, the strain
energy density is conveniently written as the sum of the deviatoric
strain energy density WD and the volumetric strain energy density
WV .

3.7: Hyperelastic material models
 With knowledge of how the strain energy density W depends

on the Green-Lagrange strain tensor (through the invariants or
stretches), the 2nd Piola-Kirchhoff stress tensor is evaluated using
1  W W 
Sij     (3.7-6)
2  ij  ji 
and the incremental material tensor is evaluated using
1  Sij Sij 
Cijrs     (3.7-7)
2   rs  sr 
3.7.1 Mooney-Rivlin material model


 The Mooney-Rivlin material model is obtained by setting
Model=Mooney in the MATHE material entry. It can also be
obtained using the MATHP material entry. It is based on the
following expression of the strain energy density:
WD  C10  I1  3  C01  I 2  3  C20  I1  3  C11  I1  3 I 2  3 

2
C02  I 2  3  C30  I1  3  C21  I1  3  I 2  3 

2 3 2
C12  I1  3 I 2  3  C03  I 2  3

2 3
(3.7-8)
where Cij are material constants , and I1 and I 2 are the first and
second strain invariants at time t, referring to the original
configuration (see ref. KJB, Section 6.6.2 for the definitions of the
strain invariants).
Note that constants Aij used in the MATHP material entry are
identical to Cij constants used in MATHE and in the equation
above.
 This strain energy density expression assumes a totally

ref. KJB
Section 6.6.2 incompressible material  I 3  1 . It is modified as explained below
for plane strain, axisymmetric or 3-D analysis.

Plane stress analysis: In plane stress analysis, the material is

assumed to be totally incompressible. Therefore WV is zero and
W  WD . A displacement-based finite element formulation is used,
in which the incompressibility condition of the material is imposed
by calculating the appropriate thickness of the material.
Plane strain, axisymmetric and 3-D analysis: In plane strain,

ref. KJB axisymmetric and 3-D analysis, the material is modeled as
Section 6.6.2 compressible (that is, the bulk modulus is not infinite), but the bulk
modulus can be set high so that the material is “almost
incompressible”.
The Mooney-Rivlin strain energy density equation is modified
by:
1) substituting for the invariants I1 , I 2 the reduced invariants

I1 , I 2 ,
2) removing the condition I 3  1 , and
3) adding the volumetric strain energy density
1
WV    J  1
2
(3.7-9)
2
where κ is the bulk modulus given by K in the MATHE material

entry (or two times D1 in the MATHP material entry). This
expression for the volumetric strain energy density yields the
following relationship between the pressure and the volume ratio:
p    J  1 (3.7-10)
The mixed u/p formulation (u/p formulation) is always used for

these elements, to avoid volumetric locking. The material stress-
strain descriptions are obtained by differentiation of W to obtain
stresses due to the element displacements and then taking into
account the effect of the separately interpolated pressure.
ref. T. Sussman and K.J. Bathe, "A Finite Element

Formulation for Nonlinear Incompressible Elastic and
Inelastic Analysis," J. Computers and Structures, Vol.

26, No. 1/2, pp. 357-409, 1987.
Selection of material constants: The Mooney-Rivlin material

description used here has 9 Cij constants and the bulk modulus κ.
Strictly speaking, this material law is termed a higher-order or
generalized Mooney-Rivlin material law. Choosing only C10 ≠ 0
yields the neo-Hookean material law, and choosing only C10 ≠ 0,
C01 ≠ 0 yields the standard two-term Mooney-Rivlin material law.
The small strain shear modulus and small strain Young’s
modulus can be written in terms of these constants as (assuming
 )
G  2  C10  C01  (3.7-11)

E  6  C10  C01  (3.7-12)
These moduli must be greater than zero.
 The bulk modulus κ is used to model the compressibility of the

material for plane strain, axisymmetric and 3-D analysis.
 Solution 601 assumes a default for the bulk modulus based on

small strain near-incompressibility, i.e.,
E
 with    (3.7-13)
3 1  2 
where E is the small strain Young's modulus or, in terms of the

small strain shear modulus G,
2G 1   
  500G for    (3.7-14)
3 1  2 
This rule of thumb can be used to estimate the bulk modulus in the
absence of experimental data. However, lower values of the bulk
modulus can be used to model compressible materials.
 Solution 701 assumes the same bulk modulus based on small

strain near-incompressibility. However, this can significantly

reduce the stable time step. In such cases, is better to use a bulk
modulus that results in ν=0.49.
 When automatic time step calculation is used for a Mooney-

Rivlin material in Solution 701, the critical time step is governed
by the dilatational wave speed. This is most frequently an
acceptable assumption since the material is almost incompressible.
 As the material deforms, the bulk to shear modulus ratio may

change, because the instantaneous shear modulus is dependent on
the amount of deformation. A value of the bulk modulus that
corresponds to near incompressibility for small strains may not be
large enough to correspond to near incompressibility for large
strains.
3.7.2 Ogden material model
 The Ogden material model is obtained by setting Model=Ogden

in the MATHE material entry. It is based on the following
expression:
9
 
WD    n 1 n  2 n  3 n  3  (3.7-15)
n 1   n 
where  n and  n are Ogden material constants.

ref. KJB
Section 6.6.2 incompressible material  I 3  1 . As in the Mooney-Rivlin
material, the strain energy density expression of the Ogden material
is used unmodified for plane stress analysis, and is modified for
plane strain, axisymmetric and 3-D analysis. The modification is
made by:
1) substituting for the stretches 1 , 2 , 3 the reduced stretches

1 , 2 , 3 ,
2) removing the condition 12 3  1 , and
3) adding the volumetric strain energy density

1 1
WV    12 3  1    J  1
2 2
(3.7-16)
2 2
where κ is the bulk modulus. The relationship between the pressure

and the volumetric ratio is the same as for the Mooney-Rivlin
material.
The u/p formulation is always used for plane strain,
axisymmetric and 3-D elements. For comments about the u/p
formulation, see the corresponding comments in the Mooney-
Rivlin material description.
Selection of material constants: The Ogden material description

used here has 19 constants:  n ,  n , n  1,...,9 and the bulk
modulus. Choosing only  n ,  n  0, n  1, 2,3 the standard three-
term Ogden material description is recovered.
The small strain shear modulus and small strain Young’s
modulus can be written as (assuming    )
1 9
G  n n
2 n 1
(3.7-17)
3 9
E  n n
2 n 1
(3.7-18)
These moduli must be greater than zero.


 When automatic time step calculation is used for an Ogden
material in Solution 701, the critical time step is governed by the
dilatational wave speed. This is most frequently an acceptable
assumption since the material is almost incompressible.
 For comments about the bulk modulus, see the corresponding

comments about the bulk modulus in the Mooney-Rivlin material
description.
3.7.3 Arruda-Boyce material model
 The Arruda-Boyce model is obtained by setting MODEL =

ABOYCE in the MATHE material entry. It is based on the

following expression:
1 1 11
WD  N KT [
2
 I1  3 
20 N
 I12  9  
1050 N 2  1
I 3  27 
(3.7-19)
19 519

7000 N 3
 I1  81  673750 N 4  I1  243]
4 5
where NKT is a material constant and N is a material parameter

representing the number of statistical links of the material chain.
 The Arruda-Boyce material model is described in the following

reference:
ref. E. M. Arruda and M. C. Boyce, “A three-dimensional

constitutive model for the large stretch behavior of
rubber elastic materials”, J. Mech. Phys. Solids, Vol,. 41
(2), pp 389-412 (1993).

ref. KJB
Section 6.6.2 incompressible material  I 3  1 . As in the Mooney-Rivlin
material, the strain energy density expression of the Arruda-Boyce
material is used unmodified for plane stress analysis, and is
modified for plane strain, axisymmetric and 3-D analysis. The
modification is made by:
1) substituting for the strain invariant I1 the reduced strain

invariant I1 ,
2) removing the condition I 3  1 , and
3) adding the volumetric energy term
WV    J ln J  ( J  1)  (3.7-20)
where κ is the small-strain bulk modulus. The relationship between

the pressure and the volume ratio is
p   ln J (3.7-21)

The u/p formulation is always used for plane strain,

axisymmetric and 3-D elements. For comments about the u/p
formulation, see the corresponding comments in the Mooney-
Rivlin material description.
 When plane strain, axisymmetric or 3-D elements are used,

there should be at least one solution unknown. This is because the
constraint equation used in the u/p formulation is nonlinear in the
unknown pressures. Therefore equilibrium iterations are required
for convergence, even when all of the displacements in the model
are prescribed.
 Material constant N KT is only an approximation to the initial

shear modulus G . The relationship between N KT and G is
 3 99 513 42039 
G  N KT 1   2
 3
 
 5 N 175 N 875 N 67375 N 4 
3.7.4 Hyperfoam material model


 The hyperfoam material model is obtained by setting
Model=Foam in the MATHE material entry. It is based on the
following expression:
N
n  n 1 
W  n n
1  2  3  3  
J  n  n  1   (3.7-22)
n 1  n  n 
in which there are the material constants  n ,  n ,  n , n  1,..., N .

The maximum value of N is 9.
 A material model similar to the hyperfoam material model is

described in the following reference:
ref. B. Storåkers, “On material representation and

constitutive branching in finite compressible elasticity”,
J. Mech. Phys. Solids, Vol,. 34(2), pp 125-145 (1986).
In this reference,  n is the same for all values of n.

 The strain energy density can be split into deviatoric and

volumetric parts
N
n 
WD   1  2  3  3 J 3 / 3 
n n n n
(3.7-23)
n 1  n
N
n  1 
WV  
n 1  n
 n / 3
3 J 3  1 
n
 
J 3 n n  1   (3.7-24)
 
Notice that W  WD  WV . This decomposition of the strain energy

density has the advantage that the stresses obtained from the
deviatoric and volumetric parts separately are zero when there are
no deformations:
1  W WD 
SijD   D   0,
ij  0 2  ij  ji 
ij  0
1  W W 
SijV   V  V 0
ij  0 2  ij  ji 
ij  0
Notice that WD contains the volumetric part of the motion through

 /3
the term 3 J 3 n . Therefore WD is not entirely deviatoric.
 The material is not assumed to be totally incompressible.

Because both WD and WV contain the volumetric part of the
motion, the mixed u/p formulation cannot be used with the
hyperfoam material. A displacement-based formulation is used.
Selection of material constants: The hyperfoam material

description used here has 27 constants:  n ,  n ,  n , n  1,...,9 .
The small strain shear modulus and small strain bulk modulus
can be written as
1 9
G  n n
2 n 1
(3.7-25)

9
 1
     n   n n (3.7-26)
n 1  3 
These moduli must be greater than zero, hence we note that  n

should be greater than –1/3.
When all of the  n are equal to each other   , then the
Poisson’s ratio is related to  using

 (3.7-27)
1  2
 The hyperfoam material model is generally used for highly

compressible elastomers.
If the ratio of the bulk modulus to shear modulus is high
(greater than about 10), the material is almost incompressible and
we recommend that one of the other hyperelastic materials be used.
3.7.5 Sussman-Bathe material model
 The Sussman-Bathe model is obtained by setting MODEL =

SUSSBAT in the MATHE material entry. It is based on the
following equation:
WD  w(e1 )  w(e2 )  w(e3 ) (3.7-28)
where w(e) is a function of the principal logarithmic strain (Hencky

strain) and e1, e2, and e3 are the principal logarithmic strains.
 The primary goal of the model is to fit given uniaxial

tension/compression data very well. This goal is accomplished by
using a spline to fit the derivative of w(e), as described in detail
below.
Of course, when uniaxial tension/compression data is known, a
curve fitting approach can, in theory, be used to determine the
constants for the other hyperelastic models, e.g. the Ogden material
model. But this curve fitting in practice does not provide good fits
to the data under many circumstances.

 This strain energy density expression assumes a totally

incompressible material (I3 = 1) and is modified as explained below
for plane strain, axisymmetric or 3-D analysis.
 The Sussman-Bathe model is given in the following reference
ref. T. Sussman and K.J. Bathe, “A model of incompressible

isotropic hyperelastic material behavior using spline
interpolations of tension-compression test data”,
Commun. Numer. Meth. Engng, Vol. 25, Issue 1, pp. 53-
63, January 2009.

 The following gives a quick summary of the Sussman-Bathe
model. In this summary, we assume that the material is totally
incompressible. Differences due to slight compressibility are small.
Theoretical background:
1) The Cauchy stress  i corresponding to the principal strain ei is
WD
i   p  w(ei )  p (3.7-29)
ei
where w(ei )  dw / de .
2) In uniaxial tension/compression (Fig. 3.7-1), e1  e,

e2  e3   12 e so
  w(e)  w( 12 e) (3.7-30)
(3.7-30) can be inverted to obtain
  

w(e)   2
1 k
e (3.7-31)
k 0
The series converges when  (e)  0 as e  0 .

Original
Deformed
L2
F1
L3 0
L1
A1 = L2 L3 = 0 A1
L1
L1
L1 1 F1
e = e1 = ln 0
e2 = e3 = - e1 t1 = t2 = t3 = 0
L1 2 A1
Fig. 3.7-1: Uniaxial tension/compression test
3) The asymptotic conditions for w are w(e)   as e   ;

w(e)   as e   . These asymptotic conditions correspond to
the asymptotic conditions of infinite stresses for infinite strains.
4) For a stable material, it is necessary (but not sufficient) that

w(e)  0 for all e. Not all materials for which  (e)  0 have
w(e)  0 . For example, the material with
 (e)  ET e, e  0
 EC e, e  0
where ET and EC are constants greater than zero, has w(e)  0

only if
1
2 ET  EC  2 ET
5) Given only simple tension data for  (e) , there are multiple
w(e) that exactly correspond to  (e) , for positive e only. Two
such w(e) are

w(e)  0, e  0
  (e), e  0
and
w(e)   (2e), e  0
 0, e  0
Hence the material is not uniquely described given only uniaxial

tension (or uniaxial compression) data. Both uniaxial tension and
uniaxial compression data must be provided to uniquely describe
the material.
6) The small-strain Young’s modulus E is found by differentiating

the uniaxial stress-strain curve, and evaluating at e = 0, and, since
the material is almost incompressible, the small-strain shear
1
modulus G is G  E . The results are
3
3 1
E w(0) , G  w(0) (3.7-32a, b)
2 2
8) The Ogden material model can be considered a special case of

(3.7-28), since the Ogden material model can be written in terms of
w(e) :
w(e)   n  exp( n e)  1 (3.7-33)

n
Spline representation of w(e) :
In the Sussman-Bathe model, we choose w(e) to fit given uniaxial

tension/compression data very well, as follows.
The uniaxial tension-compression data is in the form of user-

specified data points  ei , i  . From these data points, we build a
non-uniform cubic spline representation of the uniaxial

tension/compression stress-strain data    (e) , as shown in Fig.

3.7-2. For the non-uniform cubic spline representation of  (e) ,
t(e)
Segment of Slope > 0

cubic spline
emin,t
emax,t e
Measured data point

Slope > 0
Fig. 3.7-2: Uniaxial tension/compression stress-

strain spline
1) A spline segment is placed between two successive user-input

data points. The user-input data points need not be equally spaced.
2) The range of the cubic spline is between the first and last user-
input data points.
3) Outside the range of the cubic spline, the slope of  (e) is greater
than zero. This ensures that the asymptotic conditions of
 ()   ,  ()   are met.
Using the non-uniform cubic spline representation of  (e) and

(3.7-31), we build a uniform cubic spline for w(e) as shown in
Fig. 3.7-3. For the uniform cubic spline representation of w(e) ,

w’(e)
Segment of Slope > 0

cubic spline
emin,w’
emax,w’ e
Slope > 0
Fig. 3.7-3: w’(e) spline
1) The same number of spline segments is used in tension and in

compression.
2) The range of the cubic spline is the same in tension and in

compression. This range includes the range of the user-input data
points.
3) Outside the range of the cubic spline, the slope of w(e) is

greater than zero, whenever possible.
In order to measure the accuracy of the spline representation of

w(e) , we define the relative interpolation error
 (e)   (e)
r  max (3.7-34)
e  (e)
in which  (e) is the stress evaluated from the spline representation

of w(e) (using 3.7-30), and  (e) is the stress evaluated from the
spline representation of  (e) .
The number of spline segments is automatically chosen to make
the interpolation error r smaller than a user-specified value.
Typically only a few spline segments need be used for w(e) in

order to reduce the interpolation error to a value smaller than

experimental error.
Each cubic spline segment in w(e) can be written
w(e)  Ai 1 z  Bi 1 z 3  Ai (1  z )  Bi (1  z )3 (3.7-35)
for the segment ei  e  ei 1 where z  (e  ei ) /(ei 1  ei ) . With

this definition, w(e) can be written
 z2 z4 1  (1  z ) 2 1  (1  z ) 4 
w(e)  Ci  (ei 1  ei )  Ai 1  Bi 1  Ai  Bi 
 2 4 2 4 
(3.7-36)
The program determines the constants Ai, Bi, Ci from uniaxial

stress-strain data, as described above.
Plane stress analysis: The material is assumed to be totally

incompressible. Therefore WV is zero and W = WD. A
displacement-based finite element formulation is used, exactly as
for the Mooney-Rivlin material model described above.
Plane strain, axisymmetric and 3-D analysis: The material is

modeled as compressible (that is, the bulk modulus is not infinite),
but the bulk modulus can be set high so that the material is “almost
incompressible”.
Equation (3.7-28) is modified by 1) substituting the deviatoric
principal strains for the corresponding principal strains, 2)
removing the condition e1 + e2 + e3 = 0, and 3) adding the
volumetric strain energy density
WV    J ln J  ( J  1)  (3.7-37)
where κ is the bulk modulus. The relationship between the pressure

and the volume ratio is
p   ln J    e1  e2  e3  (3.7-38)
which is a generalization of the small-strain pressure-strain

relationship. The u/p formulation is always used. For comments

about the u/p formulation, see the corresponding comments in the
Mooney-Rivlin material description.
Data input considerations:
1) Data input is in the form of a set of stress-strain data points, with

positive stresses/strains corresponding to uniaxial tension and
negative stresses/strains corresponding to uniaxial compression.
Compression and tension data are entered together in the same data
set.
2) The data set should contain both tension and compression data
(compression data is possibly converted from equibiaxial tension
data, see below). If the data set contains only tension data, the
program will assume that the true stress / true strain curve in
compression is a straight line, which is most likely not a good
assumption.
3) The stresses and strains in the set of stress-strain data points can
be either
a) True stresses and logarithmic strains (SSTYPE=True in

MATHE)
b) Engineering stresses and engineering strains (SSTYPE=Eng in
MATHE)
c) Engineering stresses and stretches (SSTYPE=Stretch in
MATHE)
4) Data points from equibiaxial tension experiments can be

converted into equivalent uniaxial compression data. The
conversion formulas are:
eu  2eb , u  b2 , 0 eu  1  0 eb   1
2
(3.7-39)
 u   b , 0  u   0  b b3
where eu is the is the equivalent uniaxial logarithmic strain (< 0), eb

is the equibiaxial logarithmic strain (> 0), u is the equivalent
uniaxial stretch, b is the equibiaxial stretch, 0 eu is the equivalent
uniaxial engineering strain, 0 eb is the equibiaxial engineering

strain, u is the equivalent uniaxial true (Cauchy) stress,  b is the

equibiaxial true (Cauchy) stress, 0u is the equivalent uniaxial
engineering stress, 0b is the equibiaxial engineering stress. All of
these conversion formulas assume that the material is
incompressible.
5) The Sussman-Bathe model fits the data so closely that roughness

and waviness in the data causes roughness and waviness in the
w(e) splines. The program does not smooth the data in order to
eliminate roughness and waviness. If the original data set contains
roughness and waviness that should not be present in the analysis,
the data set should be smoothed before entering the data into the
program.
6) If the data set corresponds to a stable material, then the

Sussman-Bathe model is stable, otherwise the Sussman-Bathe
model may not be stable.
7) The strain range of the data set should contain the range of
strains anticipated during the analysis.
8) Do not confuse uniaxial compression with hydrostatic

compression. These two test cases are very different.
3.7.6 Thermal strain effect
 When the material is temperature-dependent, a coefficient of

thermal expansion can be included. The coefficient of thermal
expansion is constant. The thermal strain is calculated as
 th      0  (3.7-40)
where  0 is the initial temperature, and the thermal strain is

assumed to be isotropic. This is similar to the formula as is used for
the other thermo-elastic materials, see Section 3.1.6 assuming a
constant thermal expansion coefficient.
 When the thermal strain is non-zero, the deformation gradient

X is assumed to be decomposed into a thermal deformation

gradient Xth and a mechanical deformation gradient X m , using
X  X m Xth (3.7-41)
The thermal deformation gradient is
Xth  (1  eth )I (3.7-42)
therefore the mechanical deformation gradient is
X m  (1  eth ) 1 X (3.7-43)
the mechanical Cauchy-Green deformation tensor is
Cm  (1  eth ) 2 C (3.7-44)
and the mechanical Green-Lagrange strain tensor is
1
ε m  (1  eth ) 2 ε 
2
1  (1  eth ) 2  I (3.7-45)
For small thermal strains, the last equation reduces to

ε m  ε  eth I , so that the strains are nearly the sum of the
mechanical and thermal strains, as in small strain analysis.
However, we do not assume that the thermal strains are small.
 The strain energy densities are computed using the mechanical

deformations. This is done by computing all invariants and
stretches using the mechanical deformations, e.g. the mechanical
Cauchy-Green deformation tensor.
The 2nd Piola-Kirchhoff stresses are obtained by differentiating
the strain energy density with respect to the total strains. Since the
strain energy density is a function of the mechanical strains, we
obtain

1  W W 
Sij    
2   )ij  () ji 
1  W  m ) ab W  m )ba 

    (3.7-46)
2   m ) ab  )ij  ( m )ba  ) ji 
 

 1  eth 
2
 12  W)
m ij

W
 ( m ) ji


With this definition, the 2nd Piola-Kirchhoff stresses are conjugate

to the Green-Lagrange strains.
3.7.7 Viscoelastic effects (Solution 601 only)
 Viscoelastic effects can be included in the Mooney-Rivlin,

Ogden, Arruda-Boyce, hyperfoam and Sussman-Bathe material
models.
The viscoelastic model used is due to Holzapfel, see the
following references:
ref. G. A. Holzapfel, “On large strain viscoelasticity:

continuum formulation and finite element applications to
elastomeric structures”, Int. J. Num. Meth. Engng., Vol.
39, pp 3903-3926, 1996.
ref. G. A. Holzapfel, Nonlinear solid mechanics. A

continuum approach for engineering. John Wiley &
Sons, Chichester, pp 278-295, 2000.
ref. G. A. Holzapfel, “Biomechanics of soft tissue”, in

Lemaitre (ed.), Handbook of Materials Behavior
Models: Nonlinear Models and Properties, Academic
Press, 2001, pp 1057-1071.
In the following, we give a brief discussion of the Holzapfel

model for finite strain viscoelasticity.
Equivalent 1D model: The equivalent 1D model is shown in Fig

3.7-4. It is the same as a generalized Maxwell model with many

chains. A generic chain is denoted with superscript  , as shown in

the figure.
Strain e,
¥
stress s
E
E1 h1
..
E .
a ha
Strain Strain
ga Ga
Fig. 3.7-4: Equivalent 1D model

for viscoelastic effects
The spring E  is equivalent to the elastic stiffness of the

model. Each chain contains a spring with stiffness E  and dashpot
with viscosity   . (Note that the superscripts  and  do not
denote exponentiation.) The strain in each chain is the sum of the
strain in the spring g  and the strain in the dashpot  . The
observed stress is
      q (3.7-47)

  is
where    E  e is the elastic stress and q  E  g     

the stress in chain  . Using the definition    and the
E
assumption E     E  , the following expression is obtained:
1
q   
q      (3.7-48)


Assuming that 0 q  0 , (3.7-48) can be written in convolution

form as
 t  t 
t
q   exp       t   dt  (3.7-49)
0   
from which the total stress is
 t  t  t 
t

   E  1     exp    
e dt (3.7-50)
0     
Evidently the relaxation modulus is

  t 
E (t )  E  1     exp      which is a Prony series
    
expression.
The dissipation in dashpot  is
  1 q  
D  q   q e  g   q  e    
   

  
 (3.7-51)
  E 
and the total dissipation is D  


D . In the above, the
viscoelastic material constants for each chain are   and   .
Potential-based 1D model: The 1D model can be written in terms

of a potential as follows:
    (e)     ( g  ) (3.7-52)

1  2
where   (e)  E e is the strain energy of the elastic chain
2
1
and   ( g  )  E   g   is the strain energy in the spring of
2

   
chain  . In terms of   ,   g       g  . The 1D model
is recovered by defining
 
 , q   (3.7-53a,b)
e  fixed  e fixed
Notice that (3.7-47) and (3.7-53a) imply
 
q   .
e  fixed
g 
Finite strain model: The finite strain model is derived from the
potential-based 1D model as follows. The elastic potential is
defined as
 ε   W ε  (3.7-54)
 
where W  ij is the strain energy density from the elastic part of
the material model. The potential of each chain  is defined as
    ij , ij    W  Gij  , usage=combined

  WD  Gij  , usage=deviatoric (3.7-55a,b,c)
  WV  Gij  , usage=volumetric
in which the usage flag (which is a user-input flag) determines

whether the entire elastic strain energy density, deviatoric strain
energy density or volumetric strain energy density is taken for
chain  . Here Gij is analogous to the strain in the 1D spring g  ,
and we assume Gij   ij  ij . Note that with this definition of
Gij , we have

   
Sij  , Qij   
 ij ij fixed
 ij ij fixed
ij  ij fixed
Gij
(3.7-56a,b)
where Sij are the 2nd Piola-Kirchhoff stresses and Qij is analogous
to the stress q .
Following exactly the same arguments as in the 1D case, we
obtain
1
Q ij   Qij    Sij (3.7-57)

Assuming that 0Qij  0 , (3.7-57) can be written in convolution

form as
 t  t 
t
Qij   exp       t  Sij dt  (3.7-58)
0   
and (3.7-58) can be numerically approximated using
 t 
1  exp    
 t     t t S   t S 
Qij  exp     tQij   
t t 
    t
 ij ij 

(3.7-59)
(3.7-59) is exact when Sij does not change during the time step,
and is more accurate than the formula given by Holzapfel:
 t   t 
t t
Qij  exp     tQij    exp      t t Sij  t Sij 
    2 
(3.7-60)

t
especially when is large.

Dissipation calculations: If the dissipation is required, it is

calculated using

D  Qij  ij  Qij ij  G ij  (3.7-61)
where
 2     
Grs  Qij (3.7-62)
Gij Grs
is used to compute the unknown G rs from the known Q ij .

If usage=combined,
 2   2W
 
 
   Cijrs . (3.7-63)
Gij Grs  ij  rs
where the tensor Cijrs is evaluated at the strain state Gij . The
dissipation calculation requires the solution of a set of simultaneous
linear equations of at most order 6 (in the 3D case) at each
integration point.
If usage=deviatoric,
 2    WD
2
 
      CD ijrs (3.7-64)
Gij Grs  ij  rs
where the tensor  CD ijrs is evaluated at the strain state Gij . Here,
the dissipation calculation requires a singular value decomposition
of  CD ijrs , since  CD ijrs has a zero eigenvalue. A similar
situation applies when usage=volumetric, except that the
corresponding material tensor has only one nonzero eigenvalue.

The procedure given in (3.7-61) to (3.7-64) is only

approximate, since the fundamental assumption Gij   ij  ij
strictly speaking only holds for small strain analysis.
Restrictions and recommendations: The allowed values of the

usage flag depend upon the material model and finite element type,
as shown in Table 3.7-1.
In view of the restrictions, we recommend that
usage=“combined” be used in conjunction with the hyperfoam
material model, and that usage=“deviatoric” be used in conjunction
with the Mooney-Rivlin, Ogden, Arruda-Boyce and Sussman-
Bathe material models.
Table 3.7-1: Allowed values of the usage flag
Mooney-Rivlin, Ogden, hyperfoam

Arruda-Boyce,
Sussman-Bathe
Plane Plane strain, Plane Plane strain,
stress1 axisymmetric, stress3 axisymmetric,
3D2 3D
usage=combined yes no yes yes
(usage(i)=2)
usage=deviatoric yes yes no yes
(usage(i)=0)
usage=volumetric no no no yes
(usage(i)=1)
1. Usage cannot be equal to “volumetric”. This is because the material is

assumed to be fully incompressible, hence the volumetric strain energy
density is zero.
2. When the u/p formulation is used, the usage flag cannot be “combined” or
“volumetric”. This is because the modification to the volumetric stresses
caused when the usage flag is “combined” or “volumetric” is not taken into
account in the u/p formulation.
3. The only allowable value of the usage flag is “combined”. This is because
the out-of-plane stress component S xx must be zero, and in the Holzapfel
finite strain viscoelastic model, the only way that this can happen is if S xx
is zero.

Time-temperature superposition: The preceding derivation

assumes that the viscoelastic response is not temperature-
dependent. One method of including the effects of temperature is
the method of time-temperature superposition.
In time-temperature superposition, the actual time t is replaced
by the reduced time  . The relationship between the actual time
and reduced time is given by
d 1
 (3.7-65)
dt aT ( t )
where t is the temperature and aT ( t ) is the shift function.

Evidently
t
1
 
t
dt  (3.7-66)
a ( t  )
0 T
The shift function used here is either the WLF shift function,
C1 ( t   ref )
log10 aT ( t )   (3.7-67a)
C2  t   ref
or the Arrhenius shift function
1 1  t
log10 aT ( t )  C1  t   ,    ref
  
 ref 
(3.7-67b)
1 1  t
 C2  t 
   ,    ref
 ref 
where  ref is the reference temperature and C1 , C2 are material

d
constants. Notice that as t increases, aT ( t ) decreases and
dt
increases.

When using the Arrhenius shift function, the temperature unit

must be absolute (Kelvin or Rankine).
For the viscoelastic model used here, the differential equation of
the 1D model (3.7-48) becomes
dq 1  d 
  q   (3.7-68)
d  d
and using (3.7-65), (3.7-68) can be written as
1
q   q      (3.7-69)
aT ( ) 
It is seen that the effect of temperature is to modify the time

constants. As the temperature increases, the modified time
constants become smaller, that is, the material relaxes more
quickly.
The convolution equation of the finite strain model becomes
t
 t 
       d Sij
Qij   exp     

d  (3.7-70)
0    d
and (3.7-70) is numerically approximated by
  
1  exp    
      t t S   t S 
t t 
Qij  exp     t Qij   
   
 ij ij 

(3.7-71)
The only additional consideration is to calculate  , and this is

done using
t t
1
   t
aT ( t  )
dt  (3.7-72)

This integration is performed numerically assuming that ln aT ( )

varies linearly over the time step.
Heat generation: A user-specified fraction of the energy

dissipated by the viscoelastic model can be considered as heat
generation. This heat generation can cause heating in a TMC
(thermo-mechanical-coupling) analysis.
Specification of input: Viscoelastic effects are added to rubber-

like materials using the MATHEV bulk data entry. The MATHEV
bulk data entry includes:
 SHIFT: Indicates the shift function (none, WLF or

Arrhenius).
 C1, C2: The shift function material constants C1 , C2
 A table with one row for each chain. Each row in the table
contains beta(i)=   , tau(i)=   , hgen(i)= the heat
generation factor (fraction of dissipation considered as heat
generation, default value is 0.0), and usage(i)=usage flag
(default value is deviatoric). There is no restriction on the
number of chains permitted. The usage flag can be different
for each chain.
 It is seen that the dissipation calculation can be quite expensive.

Furthermore the dissipation is not required for the stress solution.
Therefore it is the default to not perform the dissipation calculation.
The dissipation is only calculated for the chain  when the heat
generation factor is non-zero.
3.7.8 Mullins effect (Solution 601 only)
When rubber is loaded to a given strain state, unloaded, then

reloaded to the same strain state, the stress required for the
reloading is less than the stress required for the initial loading. This
phenomenon is referred to as the Mullins effect.
The Mullins effect can be included in the rubber-like materials.
The material model used is the one described in the following
reference:

ref. R.W. Ogden and D. G. Roxburgh, “A pseudo-elastic

model for the Mullins effect in filled rubber”, Proc. R.
Soc. Lond. A (1999) 455, 2861-2877.
We briefly summarize the main concepts below.
Fig 3.7-5 shows the Mullins effect in simple tension. On initial

loading to force Fc , the specimen follows the force-deflection
curve a-b-c. When the load is removed, the specimen follows the
unloading curve c-d-a. On reloading to force Fc , the specimen
follows the reloading curve a-d-c, and on further loading to force
Ff , the specimen follows the loading curve c-e-f. When the load is
removed, the specimen follows the unloading curve f-g-a, and, on
reloading to force Fg , the specimen follows the reloading curve a-
g-f.
Force
h
f
e
c
b
d g
a
Deflection
Fig 3.7-5: Mullins effect loading-

unloading-reloading curves
Note that any permanent set associated with the Mullins effect
is not included in the Ogden-Roxburgh model used here.
The Ogden-Roxburgh model, as implemented in Advanced
Nonlinear Solution, uses the following strain energy density
expression:
W  WD ( ij )   ( ), all except hyper-foam

(3.7-73a,b)
=W ( ij )   ( ), hyper-foam

where W ( ij ) is the total elastic strain energy density, WD ( ij ) is

the deviatoric elastic strain energy density,  is an additional
solution variable describing the amount of unloading and  ( ) is
the damage function. W is referred to as the pseudo-energy
function. In our implementation, the deviatoric strain energy
density is used for the (almost) incompressible materials and the
total strain energy density is used for compressible materials. For
ease of writing, we discuss only the case of compressible materials;
for incompressible materials, replace W by WD in the equations
below.
 is computed as
1 1 
  1  erf  Wm  W   (3.7-74)
r m 
where erf( x) is the error function
x
2
erf( x )   exp  u  du
2
(3.7-75)
 0
Wm is the maximum value of W encountered during the

deformation history and m and r are material constants.
 ( ) is defined by
d ( )
 W (3.7-76)
d
and is computed by numerical integration of ( )  W . For a

given value of Wm , there is a minimum value of  computed as
1 W 
 m  1  erf  m  (3.7-77)
r m

The value of  ( ) at   m is written  ( m ) . (Note: the

subscript m in the term Wm means “maximum”, but the subscript m
in the term  m means “minimum”.) The time rate of change of
 ( m ) can be written as
1 W 
( m )  1   m  Wm  erf  m  Wm (3.7-78)
r m
Physically,  ( m ) is interpreted as the dissipation.

During loading, Wm  W , Wm  0 and   1 . Therefore
( )  0 and (m )  0 during loading.
During unloading or reloading, Wm  W , Wm  0 and
 m    1 . Therefore ( )  0 and ( m )  0 during unloading
or reloading.
Material constants m and r do not have any direct physical
significance. However Fig 3.7-6 shows the dependence of an
unloading-reloading curve in simple tension on these parameters. It
is seen that, for an unloading-reloading loop in which Wm  m ,
the initial slope of the reloading curve is reduced by the factor
1
1  . r must therefore be greater than 1.
r
It can also be shown that the dissipation of a loading-unloading
cycle, as shown in Fig 3.7-7, can be written as
m Wm W  1    W 2   
E
   0  de   erf  m   1  exp    m    
r m  m     
A    m    
(3.7-79)
where
C
Wm   0  de (3.7-80)
A

Force
Slope kb
Fb
2 1
Slope k b + Fb2
Slope ka p rm
Deflection
æ 1 éW ù ö
Slope ç 1 - erf ê m ú ÷ k a
è r ë m ûø
Wm = strain energy density at b
Fig. 3.7-6: Dependence of reloading curve on Mullins

effect material constants
s=Force/original area
0
B
Dissipation = shaded area
D
A
E e=displacement/
original length
Fig. 3.7-7: Dissipation in Mullins effect in a

loading-unloading cycle
0 is the engineering stress, e is the engineering strain. Therefore,

given  and Wm from two loading-unloading cycles of different
amplitude, m and r can be computed.
Heat generation: A user-specified fraction of the energy

dissipated by the Mullins effect model can be considered as heat
generation. This heat generation can cause heating in a TMC
(thermo-mechanical-coupling) analysis.

Specification of input: Mullins effects are added to the rubber-like

material model using the MATHEM bulk data entry.
The rubber-Mullins data set includes:
 R, M: The material constants r and m.

 HGEN: The heat generation factor (fraction of dissipation
considered as heat generation). The default value is 0.
3.8 Gasket material model (Solution 601 only)

 Gaskets are relatively thin components placed between two
bodies/surfaces to create a sealing effect and prevent fluid leakage
(see Fig. 3.8-1). While most gaskets are flat, any arbitrary gasket
geometry can be modeled in Solution 601.
Gasket thickness direction

(material X axis)
Gasket in-plane
directions
Only one element

through the thickness
Fig. 3.8-1: Schematic of gasket
 The sealing effect is created when the compressive load, applied

in the direction of the gasket thickness, exceeds the initial yield
stress of the gasket. The sealing effect is maintained as long as the
compressive stress does not drop beyond a specified threshold
value. The gasket ruptures if the compressive stress exceeds the
gasket’s ultimate stress. Unlike rupture, if a gasket leaks it still
maintains its load-deflection characteristics.
 The gasket model can be used with 3-D solid elements. It can
also be used with small displacement/small strain, large
displacement/small strain formulations.

 The gasket behaves as a nonlinear elasto-plastic material when

compressed in the thickness or gasket direction. Its load-
deformation characteristics are typically represented by pressure-
closure curves. Tensile stiffness can be assumed to be constant or
zero. Hardening is assumed to be isotropic.
 Since the gasket material has different properties in different

directions, it is considered to be an orthotropic material. The gasket
thickness direction is defined as the x direction of the material
coordinate system. The program attempts to determine this
direction automatically, see below.
 The closure strain is always defined as the change in gasket

thickness divided by the original gasket thickness. It is positive in
compression. The gasket pressure has units of stress, and it is also
positive in compression.
 The gasket’s uni-directional plasticity model speeds up

computations, and allows more flexibility in defining the shape of
the loading and unloading curves.
User input
The gasket material is obtained using the MATG material entry to

define the transverse and through-thickness gasket properties
together with an elastic isotropic MAT1 material entry to define the
in-plane gasket properties.
EPL is the Young's modulus for tensile loading in the thickness
direction ( Etensile ).It is allowed to set EPL to a very small number,
such as 1E-5, to model a gasket with very small tensile stiffness.
GPL is the transverse shear modulus.
YPRESS specifies the initial yield pressure. This value must
correspond to a point on the TABLD curve, see below.
IDMEM is the ID of a MAT1 entry that provides the material
constants for the membrane (in-plane) behavior.
TABLD and the TABLUi are used to enter the pressure -
closure strain relationship. These curves are defined using the
TABLES1 entry.
The leakage pressure is automatically set to 1% of the initial
yield pressure.

3.8: Gasket material model
Pressure - closure strain input
We now describe the pressure - closure strain input in more detail.

Fig. 3.8-2 shows a typical pressure - closure strain relationship. The
main loading curve is specified by parameter TABLD. The
corresponding TABLES1 entry can contain any number of points,
and successive points are connected by straight lines.
Last yield
point (rupture
point)
Gasket pressure
Main loading curve

Initial yield 2nd yield
point point
pyield Unloading curves
Leakage
pressure
Closure strain
Etensile
Fig. 3.8-2: Pressure-closure strain input for a gasket material
The point number of the first yield point on the yield curve is
given by the value of p yield (parameter YPRESS). p yield must
match the third, fourth or higher point on the yield curve. For future
reference, the point number corresponding to p yield is denoted
NPOINTS. NPOINTS must be greater than or equal to 3.
For each point on the main loading curve higher than the first
yield point, there is an associated unloading curve, as shown.
Notice that each unloading curve must have the same number of

points, and this number of points is equal to NPOINTS. There

must be at least three unloading curves, and therefore the total
number of points on the yield curve must be at least NPOINTS + 2.
There are two options for specification of each unloading curve:
1) The unloading curve can be specified using one of the

TABLUi. The corresponding TABLES1 entry must contain
NPOINTS points, the first point must have zero pressure,
and the last point must correspond to the (closure strain,
pressure) of the main loading curve yield point.
2) The program can automatically construct the unloading
curve by interpolation of the user-input unloading curves.
It is necessary to specify at least one of the unloading curves using

the TABLUi.
User input for nonlinear elastic gasket
We now describe how to input material data for a nonlinear elastic

gasket.
Fig. 3.8-3 shows an example for input of pressure - closure
strain data for a nonlinear elastic gasket. In this example, we
assume that points 2 to 4 correspond to experimental data. Points 5
to 7 are on a line extended from segment 3-4.
NPOINTS for this example is 5.
Points NPOINTS to NPOINTS+2 should correspond to very
high closure strain, so that the program never reaches the closure
strain corresponding to point NPOINTS.
The gasket material will be nonlinear elastic as long as the
closure strain is less than the closure strain at point NPOINTS, and
the relationship between closure strain and gasket pressure is given
by the segments connecting points 1 to NPOINTS.
Two unloading curves must be entered, corresponding to points
NPOINTS+1 and NPOINTS+2 on the loading curve, as shown.
Each unloading curve has NPOINTS equally spaced points, with
unloading point 1 for pressure =0 and unloading point NPOINTS
for the point coincident with the loading curve.
The tensile behavior of the gasket is governed by constant EPL
( Etensile ).

Original nonlinear elastic loading curve
Gasket pressure
7
Extended nonlinear elastic loading curve
5
Unloading curve 1
Unloading curve 2
4
NPOINTS=5 in this example 6
5
pyield 5 4
3
3
4
2
2
3
1 2 1 1
Closure strain
Etensile
Fig. 3.8-3: Example input for nonlinear elastic gasket
Pressure - closure strain response
Fig. 3.8-4 shows a typical pressure - closure strain response. The

gasket is loaded to point B, then unloaded to point C. When the
closure strain becomes negative, the response is governed by
Etensile . Then the gasket is loaded to point D. Up to this time, the
gasket behaves as a nonlinear elastic material. Then the gasket is
loaded to point E and unloaded to point F. The unloading to point F
is done on an unloading curve based at point E; this unloading
curve is interpolated from the input unloading curves. The gasket is
loaded to point G and unloaded to point H. The gasket follows an
unloading curve based at point G and interpolated from the input
unloading data. During the loading to point I, the gasket follows the
unloading curve to point G, then follows the loading curve to point
I. The gasket is then unloaded to points J and K. When the pressure
becomes negative, the response is again governed by Etensile .
During reloading to point L, the gasket follows the same unloading
curve as used for points J and K. The gasket is then loaded to point
M, at which time the gasket ruptures. The pressure drops to zero
and remains zero thereafter (point N).

Gasket pressure
Interpolated unloading
curve based on highest M
closure strain I
Initial yield
point
G
E L
pyield
D
H
J
B
Leakage F
pressure A
N
C K Closure strain
Fig. 3.8-4: Typical pressure - closure strain history
 Each gasket can have one of the following five states:
Open: The gasket pressure is less than the leakage pressure

and the gasket has never undergone plastic
deformation .
Closed: The gasket pressure is higher than the leakage
pressure and the gasket has never undergone plastic
deformation.
Sealed: There has been plastic gasket deformation, and
thereafter the pressure has always remained above
the gasket leakage pressure.
Leaked: There has been plastic gasket deformation, and the
gasket pressure has dropped below the gasket
leakage pressure for at least one solution step after
the plastic deformation occurred.
Crushed: Gasket closure strain has exceeded the rupture value
for at least one solution step.

Modeling issues
 The gasket must be modeled as a single layer of 3-D elements.

Only linear elements are possible (6-node wedge and 8-node brick
elements).
 If the user does not explicitly set the material axes coordinate
system in gasket elements (using CORDM of the PSOLID entry),
the program attempts to automatically determine the gasket
material axes directions. This automatic determination is made
based on the element layout. Fig 3.8-5 shows element layouts in
which the program can and cannot automatically determine the
gasket material directions. If the program can automatically
determine the gasket material directions, then the program
constructs a material coordinate system in which the x direction
corresponds to the gasket thickness direction.
 Warning, if the program cannot automatically determine the

gasket material directions, and if the material axes for the gasket
are not explicitly defined by the user, then the basic coordinate
system is used, and the x direction of the basic coordinate system
corresponds to the gasket thickness direction. If the gasket
thickness direction is in fact not the x direction of the basic
coordinate system, the results will be incorrect.
 The top and bottom surfaces of a gasket can be separate from

those of the mating surfaces. In this case, they should be connected
via contact. The gasket can also share a common surface with the
intended mating surface. In this case, contact is not needed,
however, the gasket cannot separate from its target. A gasket
surface can also be attached to its mating surface via tied contact,
mesh glueing, constraint conditions, or rigid elements.
Output variables: The following gasket output variables are

available: Gasket pressure, Gasket closure strain, Gasket yield
stress, Gasket plastic closure strain, Gasket status.
 Note that all these output variables are scalar quantities.

(a) Each element is surrounded by neighboring elements on adjacent sides.

Gasket direction can be automatically determined.
(b) At least one element is surrounded by neighboring elements on opposite sides only.
Gasket direction cannot be automatically determined.
Fig. 3.8-5: Automatic determination of gasket material directions
3.9. Shape memory alloy (Solution 601 only)

 The Shape Memory Alloy (SMA) material model is intended to
model the superelastic effect (SE) and the shape memory effect
(SME) of shape-memory alloys. It is defined using the MATSMA
material entry.
 The SMA material model can be used with rod, 2-D solid, 3-D
solid and shell elements. It is available only for implicit analysis
(Solution 601).
 Shape memory alloy materials can undergo solid-to-solid phase

transformations induced by stress or temperature. The high
temperature phase is called austenite (A) with a body-centered
cubic structure and the low-temperature phase is called martensite
(M) with a monoclinic crystal structure in several variants.

3.9: Shape memory alloy
Fig.3.9-1 shows a schematic SMA stress-temperature diagram.

The martensite phase in two generalized variants and the austenite
phase are shown, as well as stress- and temperature-dependent
transformation conditions. The martensite phase is favored at low
temperatures or high stresses. Upon heating from low temperature
the material begins transforming from martensite to austenite at
temperature As. The transformation is 100% complete at
temperature Af. If the material is then cooled again, the austenite
starts transforming back to martensite at temperature Ms. This
transformation is 100% complete at temperature Mf. These four
temperatures are also stress dependent with high stresses favoring
the martensite phase. This stress dependence is assumed linear with
slope CM and CA for the martensite and austenite temperatures,
respectively. A typical variation of volume fraction of martensite in
the SMA material with temperature is shown in Fig. 3.9-2.
 A typical uniaxial isothermal stress-strain curve is shown in Fig.

3.9-3.
 The superelastic effect is evident when the material is deformed

at temperature   Af and is displayed in Fig. 3.9-3(a). The stress
cycle application induces transformations from AM and then
from MA to exhibit the hysteresis loop. The shape memory
effect is evident when the material is deformed at temperature
  As and is displayed in Fig. 3.9-3(b). A residual transformation
strain remains after unloading; however heating the material to
temperature above A f leads to thermally induced MA
transformation and the recovery of transformation strain.

seff
Detwinned
martensite CM CM CA CA
Austenite
Twinned
martensite
Mf Ms As Af
Temperature, q
Fig. 3.9-1: SMA stress-temperature phase diagram
Transformation at seff = 0
Transformation at seff > 0
Mf Mfs As Ass
Volume fraction of martensite, m
1.0
0.8
0.6
0.4
0.2
0.0
Ms Mss Af Afs
Temperature, q
Fig. 3.9-2: Volume fraction of martensite vs. temperature

s s
q > Af q < As
(a) Superelasticity e (b) Shape memory effect e
Fig. 3.9-3: Schematic of stress-strain curves for shape-memory

alloys


 Both shape memory effects due to transformation from
martensite to austenite and due to re-orientation of the martensite
are captured by modeling the twinned and detwinned martensites as
different phases.
 The SMA material model is based on the following equations:
 The total strain,
   e  t 
where
 e= elastic strain
  = thermal strain
 t = transformation strain; to be evaluated
 The one-dimensional macro-scale model,
   s  t ; 0   
   = 1
 t =  tmax s

  ((1   ) E A   EM )(      t max s )
t = twinned martensite volume fraction

s = detwinned martensite volume fraction
A = austenite volume fraction
tmax = maximum recoverable residual strain; a material
property usually obtained from a simple tension test
when the material is fully detwinned martensite (s = 1)
The flow rule of three-dimensional constitutive model,
 ijt   s max
t
nijt
3  sij 
nijt    ; for the martensitic transformation
2 
3   ij 
t
nijt    ; for the reverse transformation

2   t 
where
sij = deviatoric stresses
3
 sij sij is the effective von Mises stress
2
3 t t
t   ij  ij is the effective transformation strain
2
This results in the following equation for deviatoric stress

calculation:
t t
E  
t t
sij 
1 t t
  
 t t
 ij   ijt 

where
t t
 ij  t t  ij  t  ijt
Four phase transformation conditions,
1. Starting condition for the martensitic transformation
f M S  3J 2  CM (  M S )
2. Ending condition for the martensitic transformation
f M f  3 J 2  CM (  M f )
3. Starting condition for the reverse transformation
f AS  3J 2  C A (  AS )
4. Ending condition for the reverse transformation
f Af  3J 2  C A (  Af )
The phase transformation rate using linear kinetic rule

(Auricchio and E. Sacco, 1999),
R
  f f
ff
t t
3 J2  t J2
f   c( t t  t )
2 
t t
where, for the austenite to martensite transformation,
f f  - f M f , c  CM and R = 1 - 
and for the reverse martensite to austenite transformation,

f f  f A f , c  C A and R = 
 Evolution of single-variant detwinned martensite:
Martensite re-orientation is based on the following condition
f R  3 J 2  CR   R
where
 R  material yield property at  = 0
CR = slope of yield function temperature variation
Austenite to martensite transformation leads to
s  
Martensite to austenite transformation leads to proportional

transformation of the twinned and detwinned phases:
s 
s  

t 
t  

 The computational steps for the stress-integration of the SMA

model are as follows (Kojić and Bathe, 2005):
1. Calculate the trial deviatoric stresses, assuming no additional

phase transformation or re-orientation,
E ( t ) t t
t t TR
s  (  ij )
1  ( t )
ij
2. Check for martensitic re-orientation,

f R  0 and t s  t and t  1
Check for austenite to martensite transformation,
f M f f M s  0 and t  1 and f  0
Check for martensite to austenite transformation,
f Af f As  0 and t  0 and f  0
3. In case of martensitic re-orientation solve the following

governing equation:
3 t t t t
g 2 ( s )  sij sij  CR t t   R  0
2
(3.9-1)
E ( ) t t "
t t
t t
sij  (  ij   s   max
t t t t
nij )
1   ( )
t t
The martensite reorientation calculation step is optional; it is

activated when σR > 0 is input.
4. In case of austenite to martensite transformation, solve the

following governing equation:
t t
R  3 t t
J 2 ( )  t J 2 
g ( )     c( t t  t )   0
t t
f f  2 
t t

(3.9-2)
where
E ( ) t t "

t t
t t
sij  (  ij   s ( )   max
t t t t
nij )
1   ( )
t t

R  1   , f f   f M f and c = CM
5. In case of martensite to austenite transformation, solve the

governing equation (3.9-2) with
R   , f f  f Af and c = CA
6. Update history-dependent variables for this time step/iteration

step.
7. Calculate the consistent tangent constitutive matrix.
ref. M. Kojić and K.J. Bathe, Inelastic Analysis of Solids and

Structures, Springer, 2005
ref. F. Auricchio and E. Sacco, “A Temperature-Dependent

Beam for Shape-Memory Alloys: Constitutive
Modelling, Finite-Element Implementation and
Numerical Simulation”, Computer Methods in Applied
Mechanics and Engineering, Vol. 174, pp. 171-190
(1999)
3.10 Viscoelastic material model (Solution 601 only)

 The viscoelastic model can be used with the rod, 2-D solid, 3-D
solid and shell elements.
 The viscoelastic model can be used with the small

displacement/small strain, large displacement/small strain and
large displacement/large strain kinematics (2-D solid and 3-D
solid elements only).
When used with the small displacement/small strain kinematics,
a materially-nonlinear-only formulation is employed, when used
with the large displacement/small strain kinematics, a TL
formulation is employed and when used with the large
displacement/large strain kinematics, the ULH formulation is
employed.

3.10: Viscoelastic material model
 The mechanical behavior for an isotropic and linear viscoelastic

material may be expressed in tensor notation as
dG ( )
t
sij (t )  2G (0)eij (t )  2  eij (t   ) d (3.10-1)
0
d
dK ( )
t
 kk (t )  3K (0) kk (t )  3   kk (t   ) d (3.10-2)
0  d
1
where t is the time, sij   ij   ij kk is the deviatoric stress,  ij
3
1
is the Kronecker delta,  ij is the stress, eij   ij   ij  kk is the
3
deviatoric strain,  ij is the strain, G(t) is the shear modulus and
K(t) is the bulk modulus.
In the presence of a temperature variation  (t ) the stresses for
an isotropic and thermorheologically linear viscoelastic material
may be written as

dG ( )
sij (t )  2G (0)eij (t )  2  eij (   ) d (3.10-3)
0
d
 
dK ( ) dK ( )
 kk (t )  3K (0)   kk (t )   (t )   3   kk (   )
th
d  3 kkth (t )  d
d d
kk
0 0
(3.10-4)
where
 t
        d ,         d (3.10-5)
 
0 0
and the thermal strain is given by
 kkth  t   3  (t )    t   TALPHA   3  0   0  TALPHA  (3.10-6)

  (t )  is the temperature-dependent coefficient of thermal

expansion and  (t ) is the shift function, which obeys
d
 (TALPHA )  1,  ( )  0, 0 (3.10-7)
d
Note that TALPHA is the reference temperature used for thermal
strain calculation.
In equations (3.10-3) and (3.10-4) it is assumed that the
mechanical and thermal responses are uncoupled. Furthermore if
the temperature is constant, equations (3.10-3) and (3.10-4) reduce
to equations (3.10-1) and (3.10-2).
 We assume the following thermo-material properties:

G
G (t )  G   Gi e  it (3.10-8)
i 1
K
K (t )  K    K i e  it (3.10-9)
i 1
 (t )  0 (3.10-10)
    (t )  (3.10-11)
where G and K  are the long-time shear modulus and bulk

modulus respectively,  i and  i are the decay constants for the
shear modulus and bulk modulus respectively and G and  K are
the number of time-dependent terms for the shear modulus and
bulk modulus respectively. Equations (3.10-8) and (3.10-9) are
referred to in the literature as Prony or Dirichlet series. G and  K
are limited to a maximum value of 15.
The shift function used is either the Williams-Landell-Ferry
(WLF) equation, written as follows
C1    0 
log10  ( )   (3.10-12)
C2     0 

3.10: Viscoelastic material model
or the Arrhenius shift function
1 1 
log10  ( )  C1    ,    0
  0 
(3.10-13)
1 1 
 C2    ,    0
  0 
in which C1 and C2 are material constants, and  0 is defined as the

initial temperature of the model, which must be the same as the
reference temperature of the viscoelastic material model.
 The viscoelastic material is specified using the MATVE bulk

data entry. The MATVE bulk data entry uses TABVE bulk data
entries for the input of the shear and bulk modulus relaxation
functions.
 The nodal point temperatures are input as discussed in Section

5.6.
 For more information, see the following references:
ref. W.N. Findley, J.S. Lai and K. Onaran, Creep and

relaxation of nonlinear viscoelastic materials, Dover
Publications, 1976.
ref. R.L. Frutiger and T.C. Woo, “A thermoviscoelastic
analysis for circular plates of thermorheologically simple
material”, Journal of Thermal Stresses, 2:45-60, 1979.
3.11 Heat transfer materials (Solution 601 only)

 Heat transfer materials are available for heat transfer analyses
and coupled structural heat transfer analyses (SOL 601,153 and
SOL 601,159)
 The isotropic materials in this section (MAT4 and MATT4) are

available for rod, beam, 2-D solid, 3-D solid, and shell elements.
The orthotropic materials (MAT5 and MATT5) are only available
for 3-D solid and shell elements.

 The convection heat transfer coefficient and heat generation

capacity are input via the MAT4/MAT5 entries. However, in this
manual they are considered as loads and boundary conditions and
are therefore addressed in Chapter 5.
3.11.1 Constant isotropic material properties
 This material model is obtained with a MAT4 material entry.

The thermal conductivity and heat capacity are independent of
temperature and time and do not exhibit any directional
dependence due to the material.
3.11.2 Constant orthotropic conductivity
 This material model is obtained with a MAT5 material entry.

The thermal conductivity is orthotropic, that is, the model exhibits
a directional dependency. Three constants k1 , k2 , k3 give the
thermal conductivity along material axes (1,2,3), respectively.
 The heat capacity is isotropic for this model.
3.11.3 Temperature dependent thermal properties
 Both the constant isotropic and the constant orthotropic material

models can be made temperature dependent by adding MATT4 or
MATT5 material entries.
 Both thermal conductivity and heat capacity can be made

temperature dependent. In this case they are defined using
piecewise linear input curves.

4. Contact conditions
 Contact conditions can be specified in Advanced Nonlinear
Solution to model contact involving 3-D solid elements, shell
elements, 2-D solid elements, and rigid surfaces.
 Very general contact conditions are assumed:
 The points of contact are assumed not known a priori.

 Friction can be modeled according to various friction laws
(only standard Coulomb friction for Solution 701).
 Both sticking and sliding can be modeled.
 Repeated contact and separation between multiple bodies is
permitted in any sequence.
 Self-contact and double-sided contact are permitted.
 Tied contact can be modeled (Solution 601 only).
 A small displacement contact feature is available.
Some of the contact algorithms used in Advanced Nonlinear
Solution are described in the following references:
ref. KJB ref. Bathe, K.J. and Chaudhary, A., "A Solution Method for
Section 6.7 Planar and Axisymmetric Contact Problems," Int. J.
Num. Meth. in Eng., Vol. 21, pp. 65-88, 1985.
ref. Eterovic, A. and Bathe, K.J., "On the Treatment of
Inequality Constraints Arising From Contact Conditions
in Finite Element Analysis," J. Computers & Structures,
Vol. 40, No. 2, pp. 203-209, July 1991.
ref. Pantuso, D., Bathe, K.J. and Bouzinov, P.A."A Finite
Element Procedure for the Analysis of Thermo-
mechanical Solids in Contact," J. Computers &
Structures, Vol. 75, No. 6, pp. 551-573, May 2000.
 Contact in Advanced Nonlinear Solution is modeled using

contact sets, contact surfaces (regions), contact segments and
contact pairs, as explained in much greater detail below.

Chapter 4: Contact conditions
 Table 4-1 lists the case control commands related to contact,

Table 4-2 lists the bulk data entries related to contact surface
definition, and Table 4-3 lists the bulk data entries related to
contact set definition.
 Most of the features and tolerances needed for contact sets are
provided in the BCTPARA entry. An explanation of this entry is
provided in the NX Nastran Quick Reference Guide. Some contact
settings however apply to all contact sets (such as contact
convergence tolerances, suppression of contact oscillations). These
settings are provided in the NXSTRAT entry.
Table 4-1: Case Control commands related to contact
Contact Case Control Description

Command
BCSET Selects which contact set to use
BCRESULTS Selects which contact results to output
Table 4-2: Bulk Data entries related to contact surface definition
Contact Surface Bulk Description

Data Entry
Define contact surface on 3-D solid elements (by element and
BSURFS
nodes)
BSURF Define contact surface on shell elements (by element number)
BCPROP Define contact surface on shell elements (by property ID)
Define contact surface on free faces of 3-D solid elements (by
BCPROPS
property ID)
Define contact surface on 2-D axisymmetric, plane strain and
BEDGE
plane stress solid elements (by element edges)
Define contact surface on 2-D axisymmetric, plane strain and
BLSEG
plane stress solid elements (by node numbers)
BCRPARA Set parameters for contact surface

Table 4-3: Bulk Data entries related to contact set definition
Contact Set Bulk Data Description

Entry
BCTSET Define contact sets
BCTADD Define union of contact sets
BCTPARA Set parameters for contact sets
4.1 Overview
 Contact sets (and their contact surfaces) in Advanced Nonlinear
Solution can be either 2-D or 3-D. The contact surfaces should be
defined as regions that are initially in contact or that are anticipated
to come into contact during the solution.
 2-D contact surfaces are either axisymmetric or planar and

must lie in the global XZ plane, with all Y coordinates equal
to zero. Typical two-dimensional contact surfaces are
shown in Fig. 4.1-1.
 A 3-D contact surface is made up of a group of 3-D contact
segments (faces) either on solid elements, shell elements or
attached to rigid nodes. See Fig. 4.1-2 for an illustration.
 A contact pair consists of the two contact surfaces that may

come into contact during the solution. One of the contact surfaces
in the pair is selected to be the contactor surface and the other
contact surface to be the target surface. In the case of self-contact,
the same surface is selected to be both contactor and target.Self-
contact is when a contact surface is expected to come into contact
with itself during the solution.)
 Within a contact pair, the nodes of the contactor surface are

prevented from penetrating the segments of the target surface, and
not vice versa.

Contact surface 3
Body 2 l
l l
l l
l
D
l Body 3 l
l l
l
Contact surface 2 l l l
l
l
l l l l l l l l l
A
l l
l l
l l l
l Contact surface 1
C l
Bl
Body 1
Fig. 4.1-1: Typical 2-D contact surfaces

3-D contact surface pair
Contactor surface
(surface of cylinder)
Body 1
Target surface
(top surface of
body 2)
Body 2
Fig. 4.1-2: Typical contact surfaces and contact pair

4.1: Overview
 Fig. 4.1-3 shows the effect of contactor and target selection on

the different contact configurations.
No penetration Target
surface
Contactor
surface
Contactor
No penetration surface
Target
surface
Fig. 4.1-3: Contactor and target selection


 In Solution 601 at least one of the two contact surfaces in a
contact pair must not be rigid. If one surface is rigid, this surface
should, in most cases, be the target surface.
 In Solution 701 both contactor and target surfaces can be rigid if

the penalty algorithm is used. Otherwise, the same restriction
mentioned above for Solution 601 applies.
 Rigid surfaces have no underlying elements and therefore no

flexibility apart from rigid body motions. All their nodal degrees of
freedom must be either fixed, have enforced displacement, or be
rigidly linked to a master node which is defined on the BCRPARA
entry.
 Symmetric contact pairs can be defined, where in one contact

pair surface A can be the contactor and surface B the target, and in
another contact pair surface B is the contactor and surface A is the
target.

In Solution 601, a non-zero contact surface compliance should

always be used with symmetric contact pairs. (Contact surface
compliance is discussed in Section 4.4.) Without compliance, the
system of contact equations can become overconstrained, that is,
the same contact equation can be present more than once in the
system of contact equations. When this happens, the global system
of equations becomes singular and the solution diverges.
 Basic concepts
ref. KJB  The normal contact conditions can ideally be expressed as

Section 6.7.2
g  0;   0; g   0 (4.1-1)
where g is a gap, and λ is the normal contact force. Different

algorithms may vary in the way they impose this condition.
 For friction, a nondimensional friction variable  can be defined

as
FT
 (4.1-2)

where FT is the tangential force and  is the normal contact force.
 The standard Coulomb friction condition can be expressed as
 1
and   1 implies u  0 (4.1-3)
while   1 implies sign  u   sign  
where u is the sliding velocity.
 In static analysis, the sliding velocity is calculated by dividing

the incremental sliding displacement by the time increment. Hence,
time is not a dummy variable in static frictional contact problems.
 When (Coulomb) friction is used, the friction coefficient can be

constant or calculated from one of several predefined friction laws.

4.1: Overview
 The possible states of the contactor nodes and/or segments are
No contact: the gap between the contactor node and target

segment is open.
Sliding: the gap between the contactor node and the target
segment is closed; a compression force is acting onto the
contactor node and the node kinematically slides along the
target segments (either due to frictionless contact or to a
frictional restrictive force less than the limit Coulomb force.
Sticking: as long as the tangential force on the contactor node

that initiates sliding is less than the frictional capacity (equal to
the normal force times the Coulomb friction coefficient), the
contactor node sticks to the target segment.
 Old and new contact surface representations

Two types of contact surface representation are supported in
Advanced Nonlinear Simulation, an old and a new contact surface
representation (set via the CSTYPE parameter in the NXSTRAT
entry). The new contact surface representation is the default. The
main differences between the two representations are:
 In the old representation, contact segments are linear (2

nodes for 2D contact; 3- or 4 nodes for 3D contact). In the
new representation, contact segments can be linear or
quadratic (up to 3 nodes for 2D contact, up to 9 nodes for
3D contact).
 In the new representation, contact is based on the actual
faces of the contact segments which results in more accurate
contact constraints.
 The new representation uses a more accurate contact
traction calculation algorithm.
 The new representation generates more accurate contact
constraints for 3-D contact segments resulting from 10, 11
node tet elements and 20, 21 node brick elements. These
elements generate zero (10, 11 node tets) or negative (20, 21
node bricks) contact forces at their corner nodes when
subjected to a uniform contact pressure.
 Tractions are reported as nodal quantities in the new surface
representation.

The new contact surfaces cannot be used with the following

features:
 Segment method algorithm

 Rigid target algorithms
 Single-sided contact
For single-sided contact, which is defined using NSIDE=1
parameter on the BCTPARA card (see Fig. 4.1-4), one side of the
contact surface is assumed to be internal and the other side to be
external. Any contactor node within the internal side of a target
surface is assumed to be penetrating and will be moved back to the
surface. This single-sided option is ideal for contact surfaces on the
faces of solid elements since in that case it is clear that one side is
internal to the solid while the other is external. In this case, the
external side can usually be predicted from the geometry. This
option is also useful for shells when it is known that contact will
definitely occur from one direction. In this case, however, the
program cannot intuitively predict the internal side of the contact
surface.
Internal side
(no contactor nodes allowed)
Target
surface
External side
Fig. 4.1-4: Single-sided contact surface
 Double-sided contact
In double-sided contact, which is defined using NSIDE=2
parameter on the BCTPARA card (see Fig. 4.1-5), there are no
internal or external sides. The contactor surface nodes in this case
are prevented from crossing from one side of the target contact
surface to the other during solution. This option is more common

4.1: Overview
for shell-based contact surfaces. If a contactor node is on one side

of the target surface at time t, it will remain on the same side at
time t  t . Note that double sided contact is only supported in
3-D.
Contactor node cannot
penetrate lower side
Target
surface
Contactor node cannot

penetrate upper side
Fig. 4.1-5: Double-sided contact
 Tied contact
When the tied contact feature is selected for a contact set (TIED
parameter in the BCTPARA card), Solution 601 performs an initial
contact check at the start of the analysis. All contactor nodes that
are found to be in contact or overlapping are permanently attached
to their respective target segments. Contactor nodes that are not in
contact are also set to be tied if the contact gap is less than a user-
specified contact tolerance (TIEDTOL parameter in the BCTPARA
card). This tolerance is useful when the contact gap is due to non-
matching finite element discretizations of the contacting surfaces.
The tied contact feature is conceptually similar to using Rigid
elements or multipoint constraints to attach the node to the target
surface. The main difference is that the coefficients for the rigid
elements are automatically determined by the program and they are
only applied for the nodes that are initially in contact. The basic
idea is illustrated in Fig. 4.1-6.
Tied contact is not "real" contact because there can be tension
between tied contact surfaces. Also no sliding can occur between
tied contact surfaces.
The tied contact option can be used to connect two incompatible
meshes. However, the mesh glueing feature described in Section
5.9 produces more accurate results.

Contactor surface
Permanent rigid
connections
Gap between
contact surfaces
(exaggerated)
Target surface
Fig. 4.1-6: Tied contact option
If the contact surfaces initially overlap, they are not pushed back
to eliminate the overlap. Similarly, if there is an initial gap it is not
eliminated.
The tied contact constraint equations are computed based on the
initial nodal positions only. The constraints generated in tied
contact are not updated during the analysis. Hence, the constraints
will be inaccurate if the bodies experience large rotations.
 Small displacement contact

If the small displacement contact feature is used (CTDISP = 1 in
the NXSTRAT entry or DISP =1 in the BCTPARA entry), the
contact constraints are generated once in the beginning of the
analysis and are kept constant, as shown in Fig. 4.1-7. A target
location is identified for each contactor node if possible, and its gap
and normal direction are determined. The local coordinates of the
target point and the normal direction are then kept constant for the
remainder of the analysis. This is in contrast to the standard large
displacement contact, where the contact constraints are updated
every iteration, and the contactor nodes can undergo any amount of
sliding.

4.1: Overview
Target surface
x*1
N x*1
x1 x1
Contactor surface
Original geometry: Determine target Small displacement contact acceptable

point x*1 and normal vector N
x*1
x1 x*1
x1
Small displacement contact not Small displacement contact not

acceptable due to excessive displacement acceptable due to excessive rotation
Fig. 4.1-7: Small displacement contact feature
This feature is useful when there is very little relative

deformation around the contact region. For such problems, it is
much more computationally efficient to perform only one detailed
contact search at the beginning of the analysis, rather than
repeating the search every iteration. Also, in some cases,
convergence can also be slow or unachievable with the general
algorithm, for example as nodes oscillate between one target
segment and another equally valid neighboring target segment.
 Contact result output is controlled by the BCRESULTS Case

Control command. The user can request output of nodal contact
forces, nodal contact tractions and/or contact gap distances.
Tractions and contact gap distances are only generated on contactor
surfaces.

4.2 Contact algorithms for Solution 601

 Solution 601 offers three contact solution algorithms (set via the
TYPE flag in the BCTPARA entry):
Constraint-function method,
Segment (Lagrange multiplier) method, or
 Rigid target method
 Each contact set must belong to one of these three contact

algorithms. However, different contact sets can use different
algorithms.
 All 3 contact algorithms can be used with or without friction.
4.2.1 Constraint-function method
 In this algorithm (selected using TYPE=0 on BCTPARA card),

constraint functions are used to enforce the no-penetration and the
frictional contact conditions.
The inequality constraints of Eq. (4.1-1) are replaced by the
following normal constraint function:
2
g   g  
w  g,        N
2  2 
where εN is a small user-defined parameter. The function is shown

in Fig. 4.2-1. It involves no inequalities, and is smooth and
differentiable. The parameter εN is set via the EPSN variable in the
BCTPARA entry. The default value of 1.0x10-12 is suitable for
most applications and should rarely be modified.
It is possible to set EPSN=0.0. In this case Solution 601
automatically determines EPSN. However, this determination may
not result in correct results for some problems. Hence EPSN=0.0
should not be used in general.

4.2: Contact algorithms for Solution 601
w(g,l)
Fig. 4.2-1: Constraint function for

normal contact
 The constraint function method is also used to approximate the

rigid non-differentiable stick-slip transition of Eq. (4.1-3). This
results in a smooth transition from stick to slip and vice versa, and
it also results in a differentiable friction law that is less likely to
cause convergence difficulties.
Two friction regularization algorithms are available in
Advanced Nonlinear Solution. Both constraint functions take the
form v  u ,   0 .
The newer default algorithm involves a more accurate
linearization of the frictional constraints and, in general, converges
much faster than its predecessor. The v function is defined
implicitly as a multilinear function as shown in Fig. 4.2-2. Here εT
is a small parameter (EPST parameter in the BCTPARA entry)
which has the physical meaning of the "sticking velocity", that is,
the maximum velocity corresponding to sticking conditions. The
default value of EPST is 0.001.
In the old friction algorithm, the v function is defined implicitly
via
  u  v 
  v  arctan  0
  T 
Here εT is a small parameter (EPST parameter in BCTPARA entry)

which provides some elastic slip to the Coulomb friction law as
shown in Fig. 4.2-3. Guidelines for selecting εT are provided in
section 4.7.3.

.
-10 -2 -1 1 2 _
10 u
eT
-1
Fig. 4.2-2: Frictional contact constraint function for new

friction algorithm
t
_
2
p
1
.
-10 -2 -1 1 2 _
10 u
eT
-1
Fig. 4.2-3: Frictional contact constraint function for old

friction algorithm
The old friction algorithm can still be accessed via the FRICALG
parameter in the NXSTRAT entry.

4.2.2 Segment (Lagrange multiplier) method
 In this method (selected using TYPE=1 on BCTPARA card),

Lagrange multipliers are used to enforce the contact conditions of
Eq. (4.1-1). The kinematic conditions are enforced at the contactor
nodes, and the frictional conditions are enforced over the contact
segments.
 This method involves distinct sticking and sliding states. It also

calculates this state for each contactor node based on the contact
forces on the target segment.
4.2.3 Rigid target method
 This is a simplified contact algorithm (selected using TYPE=2

on BCTPARA card). The algorithm is fully described in Section
4.8.
4.2.4 Selection of contact algorithm
 Our experience is that in most frictionless contact problems the

constraint function method is more effective than the segment
method. The constraint function method is the default.
 For problems involving rigid targets, either the constraint

function or the rigid target algorithm can be employed.
 Note that the target surface can be rigid in all three contact
algorithms. The presence of a rigid target does not mean that the
rigid target algorithm must be used.
4.3 Contact algorithms for Solution 701

 Solution 701 offers three contact solution algorithms (set via the
XTYPE flag in the BCTPARA entry):
Kinematic constraint method,

Penalty method, or
 Rigid target method

 Each contact set must belong to one of these three contact

algorithms. However, different contact sets can use different
algorithms.
 All Solution 701 contact algorithms can be used with or without

friction.
4.3.1 Kinematic constraint method
 This algorithm is selected by setting XTYPE=0 on BCTPARA

card). It is the default explicit contact algorithm for Solution 701.
 A predictor step is first done without applying contact

constraints or forces. Then displacements are evaluated and
penetration is detected and corrected. The exact correction of
displacements requires the solution of a non-diagonal system of
equations. Instead, a good approximation is done. In this case, for
each penetrating contactor node, a penetration force
N
FCN  M C aCN *  M C N
t 2
is calculated. This is the force required to remove the penetration at

the contactor node. However, not all the penetration will be
removed by moving the contactor. The target will get some motion
depending on its mass relative to the contactor and how many
contactor nodes are touching it. So, the FCN force above is
projected to the target segment nodes:
FTNi  N i FCN
where Ni is the shape function relating the contactor displacement

to that of each target node. Similarly, the mass of the contactor
node is projected to the target in the same way:
M Ti  N i M C
and this mass is added to that of the target node itself. Then the
acceleration of the target node is determined as

aTN ( M T   M Ti )   FTNi
This correction is then used to update the target displacements. The

contactor acceleration is
aCN  aCN *   aTN N i
 For friction, a similar approach is used. A correction force is

calculated
vT
FCT *  M C
t
where vT is the tangential sliding velocity. However, this force

cannot exceed the limit force based on the normal force and the
coefficient of friction
FCT  min   FCN , FCT * 
The rest of the procedure is very similar to the case of normal

contact. The form of the equations is different if there is damping,
and is also different if the previous and current time steps are not
the same.
 A modification is also required for rigid targets, which are

common in contact. The form of the equations in this case depends
on whether the rigid target has natural or essential boundary
conditions.
 The kinematic constraint algorithm should not be used when the

target surface degrees of freedom are fixed.
4.3.2 Penalty method
 In this algorithm (selected using XTYPE=1 on BCTPARA

card), contact conditions are imposed by penalizing the inter-
penetration between contacting surfaces. When a penetration is
detected, a normal force of

F  A P  A N ( K N  N  K DN )
is applied to the contactor node, where KN is the normal stiffness,

KD is a normal rate stiffness, δN is the penetration, N is the
penetration rate, N is the normal vector pointing towards the
contactor, A is the contact area and P is the normal contact traction.
An opposing force is distributed to the target nodes.
 Similarly, in the presence of friction, the relative sliding

velocity between the two bodies is penalized as follows:
 x 
FT  min  A KT xT ,  FN T 
 xT 
where xT is the relative tangential sliding displacement.
 The normal and tangential penalty stiffnesses KN and KT can be

selected by the user, or determined automatically by the program
based on the following BCTPARA parameters: XKN, XKNCRIT,
XKT, XKTCRIT. The penalty rate stiffness KD can be explicitly
selected by the user, or determined by the program as a ratio of
critical damping for the contact node (using the XDAMP and
XNDAMP parameters).
 When penalty stiffnesses are automatically determined they are

chosen based on the masses of the contactor nodes and the time
step. They are selected such that they have a minimal effect on the
existing time step.
Note that unduly small penalty stiffnesses will lead to excessive
penetrations, and unduly large penalty stiffnesses will lead to
excessive oscillations or unstable explicit time integration.
4.3.3 Rigid target method
 This algorithm is similar to the rigid target method used in

Solution 601. It is selected using XTYPE=3 on the BCTPARA
card. The algorithm is fully described in Section 4.8.

4.3.4 Selection of contact algorithm
 The kinematic constraint method is the default in Solution 701.
 The penalty method is the simplest and fastest of the explicit

contact algorithms. It can also handle rigid contactor and target
surfaces. It also allows a contactor node to be in contact with
multiple targets simultaneously.
 The main disadvantage of the penalty method is that contact

conditions are not exactly satisfied and it usually shows oscillations
in contact forces. These oscillations can usually be removed by
using the penalty rate stiffness factor K D . The penalty method is
also sensitive to the choice of the penalty stiffness. If that stiffness
is too large it leads to instability and oscillations, and if it is too
small it leads to excessive penetrations.
 The default penalty stiffness selected by Solution 701 is, in

most cases, a suitable compromise.
4.4 Contact set properties

This section describes the main options available for contact sets.
 Contact surface offsets

Penetration of a contact surface occurs when the plane or line
defined by the contact segment nodes is penetrated. However, an
offset distance can be specified which causes the actual contact
surface to be offset from the plane defined by the contact surface
nodes. In the case of double-sided contact, the offset creates two
separate surfaces above and below the reference surface. Fig. 4.4-1
shows the possibilities for single and double-sided contact. Note
that the offset distance should be small compared to the contact
surface length.
Offsets for a whole contact set are specified via the OFFSET
parameter in the BCTPARA entry, while offsets for a specific
contact surface are set via the OFFSET parameter in the
BCRPARA entry. If one of the contact surfaces has a defined
offset, it will overwrite the contact set offset.

Offset
t Actual contact
surface
Internal side Defined contact
surface
(a) single-sided contact
(using NSIDE=1, OFFSET=t on BCTPARA card)
Offset
Actual contact
t surfaces
t
Offset
Defined contact surface
(b) double-sided contact
(using NSIDE=2, OFFTYPE=1, OFFSET=t on BCTPARA card)
Fig. 4.4-1: Contact surface offsets
If the contact surface is on a shell then half the true shell

thickness can automatically be used as the offset (OFFTYPE=2 in
the BCTPARA entry). In this case the shell thickness at a node is
obtained as the average of all of the shell thicknesses at that node.
The use of contact surface offsets in double-sided contact is not
recommended.
 Continuous normals (Solution 601 only)

The normal direction to a contact segment will in general not be
continuous between segments as illustrated in Fig. 4.4-2. This
sometimes causes convergence difficulties due to the non-unique
normals at nodes and segment edges. The continuous normals
feature first calculates nodal normals as averages of all the normals
from the attached segments, and then interpolates these nodal
normals across the segment. This leads to a uniformly varying
normal direction.

4.4: Contact set properties
Nodal normals
(a) Discontinuous normals (b) Continuous normals
Fig. 4.4-2: Contact surface normals
The SEGNORM parameter in the BCTPARA card determines

the setting for continuous normals. Continuous normals
(SEGNORM=1) is the default.
In modeling target surfaces with sharp corners, either use
discontinuous normal vectors, or use small segments near the
corners, in order that the normal vectors for segments near the
corners be computed correctly. See Section 4.7.2 for modeling tips
related to this feature.
Continuous normals give poor results with explicit time
integration. Therefore, they are blocked from Solution 701.

 Contact surface depth
By default, the contact region extends for an infinite distance below
the contact surface (for single-sided contact). However, a contact
surface depth can be defined (by setting the PDEPTH parameter in
the BCTPARA card), below which the contact surface is no longer
active. The default PDEPTH=0.0 results in an infinite contact depth
extension. Fig. 4.4-3 shows some of the possibilities.
 Initial penetration
The treatment of initial penetrations in Solution 601 is governed by
the INIPENE parameter in the BCTPARA entry. By default, if
there is initial overlap (penetration) between a contact node and a
target segment in the first solution step, the program attempts to
eliminate the overlap. Advanced Nonlinear Solution can eliminate
the overlap at the first step or over a user-specified time using the
TZPENE parameter in BCTPARA. This feature is useful if the
initial penetrations are too large to be eliminated in a single step.

Contact surface is used in self-contact.

A B A B
C PDEPTH C
G
Target depth G
D D
F F
E E
Continuum created
by segment A-B Nodes F and G have not
penetrated segment A-B.
Nodes F and G have
penetrated segment A-B.
a) Target depth option not used b) Target depth option used
Fig. 4.4-3: Contact surface depth.
The program can also calculate initial penetrations at the start of

solution and ignore them in future steps. In this case the program
does not detect penetration for a contactor node if the amount of
penetration is less than or equal to the recorded amount. Fig. 4.4-4
shows some of the possibilities. See Section 4.7.2 for modeling tips
related to this feature.
This feature can be interpreted as a redefinition of the contact
gap, as follows:
 Eliminate penetration option (INIPENE=0): The contact

gap is equal to the geometric gap.
 Ignore penetration option (INIPENE=2):
— Initial geometric gap is positive or zero: The contact gap
is equal to the geometric gap.
— Initial geometric gap is negative: The contact gap is equal
to the geometric gap minus the initial geometric gap.
Initial penetrations can also be set to gap override (see below).

Eliminate penetration
Contactor node
INIPENE = 0, TZPENE = 0.0
Target surface
Continuum
Solution start, contactor node First solution time,

initially penetrates target surface overlap is eliminated
Eliminate penetration
over time period Ignore penetration
INIPENE = 0, TZPENE > 0.0 INIPENE = 2
Overlap
distance
Overlap is gradually eliminated First solution time,

overlap is recorded
Fig. 4.4-4: Initial penetration options
 Gap override
In the gap override feature (set via the initial penetration flag
INIPENE=3 in BCTPARA), the gaps and penetrations calculated
from the finite element mesh are replaced by a fixed user-specified
value (GAPVAL parameter in the BCTPARA entry). A positive
value represents an initial gap, zero means that the contact is
touching the target, and a negative value represents an initial
penetration (which can be removed either immediately or over a
user-specified time as explained in the Section on initial
penetration above).

This feature can also be interpreted as a redefinition of the

contact gap, as follows:
 Gap-override option (INIPENE=3): The contact gap is

equal to the geometric gap minus the initial geometric gap
plus the gap override GAPVAL. This formula is applied
regardless of whether the initial geometric gap is positive,
zero or negative.
This feature is useful for problems involving curved meshes in

close proximity, such as the shrink fit example shown in Fig. 4.4-5.
The gaps and penetrations measured from the discretized finite
element mesh are sometimes inaccurate for such problems (unless
matching meshes are used). In some problems, such as that shown
in the figure, a constant geometry based overlap should be applied
to all nodes, which corresponds to a gap override value of –.
Note that mesh refinement and quadratic elements reduce the
error in the measured overlaps but frequently a very high mesh
density would have to be used if gap override is not used.
Two rings with a geometric overlap d (shrink fit)
Directly measured overlaps

are incorrect at most nodes
Proper initial overlap at
all nodes should be d
Fig. 4.4-5: Significance of gap override for curved non-matched geometries
Note also that the error in mesh based gaps and penetrations for
curved surfaces can be more significant when low precision
numbers are used for the node coordinates (such as when short

input file format is used). Gap override is also useful for such
cases.
 Contact surface extension

The target surface can be enlarged beyond its geometric bounds, so
that contactor nodes that slip outside the target can still be
considered in contact (via the EXTFAC parameter on the
BCTPARA card). This feature is useful where the edge of the
contactor and target surfaces coincide, as shown in Fig. 4.4-6. Each
target segment is enlarged by an amount equal to the contact
surface extension factor multiplied by the length of the segment.
Target
surface e = L ´ contact surface extension
L
e
Extended target
surface
Corner contactor node may slip

outside bounds of target due to
numerical round-off, or lateral
displacements
Fig. 4.4-6: Contact requiring contact surface extension
 Contact surface compliance (Solution 601 only)

Contact surface compliance is set via the CFACTOR1 parameter
on the BCTPARA card and is only available with the constraint
function algorithm in Solution 601. Contact surfaces are commonly
assumed to be rigid meaning that no interpenetration is allowed.
This situation corresponds to a contact surface compliance of 0.0.
However, the contact surface compliance feature can be used to
simulate soft or compliant surfaces. The amount of allowed
interpenetration between the contacting surfaces in this case is
penetration   p  normal contact pressure (4.4-1)

where
normal contact pressure = normal contact force / contact area
The constraint function in the presence of a compliance factor is

modified as shown in Fig. 4.4-7. A is the contact area of the
contactor node.
eP /A
Fig. 4.4-7: Constraint function for

compliant contact
The benefits of contact compliance are that it:
Improves the convergence rate. By adding compliance to the

contact surface, more contactor nodes come into contact with
the target surface. It is better for convergence to have many
nodes in contact with small forces at each contactor node, than
to have few nodes in contact with large forces at each contactor
node.
Improves the contact tractions. Without contact compliance, the

discretization error of the mesh often cause some contactor
nodes on the contact area to not be in contact which results in
spotty contact (or patchy) contact tractions. An example of
spotty contact is shown in Fig. 4.4-8.

Contact force
Closed
Open
Fig. 4.4-8: Spotty contact on a plate
Reduces numerical oscillations that sometimes result from

dynamic time integration. Numerical oscillations, such as
chatter or ringing, can be triggered by contact reversal. A node
that is out of contact at time t  t , in contact at time t and out
of contact at time t  t is said to have had contact reversal.
Contact surface compliance can be used to prevent these
oscillations.
However, the drawbacks of contact compliance are:
Nodal contact overlap. The contactor nodes penetrate the target

segment when contact compliance is used. The compliance
factor must be selected such that the contact pressures do not
cause excessive nodal overlap.
Additional (spurious) energy is stored in the contact pair.

Elastic energy is stored in the compliant contact surface. This
energy is released when the contact pair opens which can have
an effect in certain dynamic problems.
Can reduce the resultant contact force and stresses. If a

prescribed displacement is applied, the contact surface
compliance feature will reduce the stress in the body for a given
displacement. The stresses are not reduced if the body is
subjected to an applied load.
By default, the compliance factor is set to zero (  p  0 ). However,

in general, some amount of contact compliance should always be
used. It is not the default because the appropriate compliance factor

is model dependent.
The value of the contact compliance factor is usually set by trial
and error. A good starting value is 105 . The objective is to find the
smallest compliance factor that results in convergence with good
tractions (i.e., not spotty contact), but does not cause excessive
penetration (nodal contact overlap).
After contact compliance is used, the amount of penetration
should always be checked to make sure the maximum overlap is
acceptable.
When deciding on what penetration is acceptably small we must
consider the mesh used. For example, if the target surface is
curved, there will be a geometric error associated with using a
coarse contractor surface. There is no advantage if the maximum
penetration is less than the geometric error. So, if the target surface
is coarse, a large maximum penetration can be used.
 Consistent contact stiffness (Solution 601 only)

The consistent contact stiffness feature is set via the parameter
CSTIFF on the BCTPARA card. Changes in the direction of the
contact normal provide an additional contribution to the stiffness
matrix that is proportional to the value of the contact force and the
change in the normal direction. Therefore, higher convergence rates
(closer to quadratic) can sometimes be obtained by selecting the
consistent contact stiffness option which accounts for these
additional stiffness contributions. This results, however, in an
increase in the size of the stiffness matrix which is detrimental for
large problems. This option is more beneficial when discontinuous
contact normals are selected, because the derivation assumes that
the contact normals are discontinuous.
The consistent contact stiffness feature is not used when the
target surface is rigid.
Consistent contact stiffness is not used in dynamic analysis.
The default is CSTIFF=0 (consistent contact stiffness is not
used).
 Contact birth/death
The contact birth feature activates a contact set at a specific time,
while the contact death feature disables a contact set at a specific
time. They are set via the TBIRTH and TDEATH parameters on
the BCTPARA card. A 0.0 birth time means that the contact set
starts active at the beginning of the analysis, and a death time less

than or equal to the birth time means that the contact set does not
die.
 Friction delay (Solution 601 only)

When the friction delay feature is activated (FRICDLY parameter
in the BCTPARA entry), frictional conditions are applied to a
contactor node one time step after contact is established. This
feature can be useful in many problems, since it delays the non-
linearity associated with friction until contact is established.
Note that the relative sliding velocity cannot be uniquely
determined when a node was not in contact at time t, and is in
contact at time t  t . That velocity depends on the exact time at
which contact started, which is somewhere between times t and
t  t (see Fig. 4.4-9). Delaying friction is equivalent to assuming
that contact was established close to time t  t , and hence the
sliding velocity is zero and so is the frictional force.
Frictoin delay is off by default (FRICDLY=0).
Time = t
t+2Dt t+Dt
tx
1
t+2Dt t+Dt .
FT FT x1 R= Relative sliding velocity
t+2Dt t+Dt t+Dt
FN FN Friction delay FT = 0
t+2Dt . t+Dt .
x1 Runique x1 R non-unique
Fig. 4.4-9: Friction delay feature

4.5 Friction
Advanced Nonlinear Solution has a general Coulomb type friction
model, where the coefficient of friction  can be a constant or
calculated based on several pre-defined friction laws. Solution 701
however, only supports standard Coulomb friction.
 In friction, the slip velocity is used to determine whether

sticking or slipping conditions are occuring.
The slip velocity at a contactor node is defined as the velocity of
the contactor node relative to the contacting point of the target
surface, projected onto the contact plane. This definition is
opposite to the definition used in ref KJB equation (6.310) (in
which the slip velocity is defined as the velocity of the contacting
point of the target surface relative to the contactor node, projected
onto the contact plane).
The slip velocity is evaluated as the change in slip displacement
divided by the change in solution time. In this way the same
evaluation is used in static and dynamic analysis.
4.5.1 Basic friction model
By default, a constant coefficient friction is used. It is specified for

each contact pair via the BCTSET entry.
4.5.2 Pre-defined friction models (Solution 601 only)
One of the following predefined friction laws can be used instead

of constant Coulomb friction. The friction law and its input
parameters are set via the BCTPARA entry. The following
variables are used in the friction laws: the magnitude of the relative
sliding velocity u , the contact traction Tn , the consistent contact
force Fn , the current nodal coordinates x, the direction of sliding v,
and the time t. The setting for the FRICMOD parameter required
for each friction law is given in parentheses. The A1 through A5
constants used in the predefined friction laws are set up via
FPARA1 through FPARA5 parameters in BCTPARA.

4.5: Friction
 Constant coefficient of friction (FRICMOD = 1)
  A1
 Different static and dynamic friction coefficients

(FRICMOD = 4)
 A1 if u  A3
 
 A2 if u  A3
 Friction coefficient varying with sliding velocity

(FRICMOD = 5)
 u
 A1  ( A3  A1 ) if u  A2
 A2
 A3 if u  A2

 Anisotropic friction model (FRICMOD = 6)
 ( A v ) 2  ( A v ) 2  ( A v ) 2 if u  A5
 1 (1) 2 (2) 3 (3)
 A4 if u  A5
where v(1), v(2) and v(3) are the x, y and z components of the sliding
direction.
 Friction coefficient varying with consistent contact force

(FRICMOD = 7)
  A1  A2 Fn , 0    1
 Time varying friction model (FRICMOD = 8)

 t
 A1  ( A3  A1 ) if t  A2
 A2
 A3 if t  A2


 Coordinate-dependent friction model (FRICMOD = 9)
 A1  A3 x (1)  A4 x (2) in 2D
 , 0    A2
 A1  A3 x (1)  A4 x (2)  A5 x (3) in 3D
 Friction model 1a (FRICMOD = 2)
1  exp( A2Tn )

Tn A1
 Friction model 1b (FRICMOD = 12)
1  exp( A2 Fn )

Fn A1
 Friction model 2a (FRICMOD =3)
  A2  ( A2  A1 ) exp( A3Tn )
 Friction model 2b (FRICMOD = 13)
  A2  ( A2  A1 ) exp( A3 Fn )
4.5.3 Frictional heat generation
The heat generation resulting from frictional contact can be

accounted for in a coupled TMC analysis. The user selects the
fractions of the generated heat going into the contactor and target
surfaces via the TMCFC and TMCFT parameters in the BCTPARA
entry. If these two fractions do not add up to 1.0 the remaining
portion is assumed to be lost.

4.6: Contact analysis features
4.6 Contact analysis features

4.6.1 Dynamic contact/impact
For Solution 601

 Oscillations in velocities and accelerations can sometimes be
present in implicit dynamic contact analysis especially for high
speed impact problems. These oscillations can be reduced by
Applying post-impact corrections,

Settingthe Newmark parameter  = 0.5,
Adding compliance to the contact surfaces,
Using the Bathe time integration method.
 In post-impact corrections, the velocities and accelerations of

the contactor and target can be forced to be compatible during
contact (only in the normal contact direction). This feature is
activated by setting IMPACT = 1 in the NXSTRAT entry. This is
achieved by modifying the velocities and accelerations of the
contact nodes once convergence is reached such that they satisfy
conservation of linear and angular momentum.
The post-impact correction option requires additional memory
and computations.
The post-impact correction feature should not be used together
with compliant contact surfaces, since the velocities and
accelerations of the contactor and target surfaces are no longer
expected to be identical.
If post-impact correction is activated, all target nodes, except
those with all degrees of freedom fixed or enforced displacements,
must have a positive non-zero mass. The contactor nodes can have
zero mass.
 Setting the Newmark  = 0.5 instead of the default  = 0.25

(trapezoidal rule — see Section 6.3) results in an accurate solution
of rigid body impact problems, and frequently has a positive effect
on reducing numerical oscillations in flexible body contact. This
feature can be activated by setting IMPACT = 2 in the NXSTRAT
entry, or by changing ALPHA to 0.5 also in the NXSTRAT entry.
It is, however, recommended that the Bathe method be used
instead, whenever possible.

 Adding compliance to the contact surface can also significantly

reduce the numerical oscillations that result from dynamic time
integration. This is done by setting a non-zero CFACTOR1 in the
BCTPARA entry. In this case, the compliance factor must be
selected based on Eq. (4.4-1) such that the contact pressures do not
cause excessive penetration. Allowing penetration of the order of
1% of the element size usually eliminates numerical oscillations.
 The Bathe time integration method provides some numerical

damping to the high frequency content of the solution, which
includes the contact oscillations.
For Solution 701

 Oscillations in velocities and accelerations can sometimes be
present in explicit dynamic contact analysis especially for high
speed impact problems. These oscillations are more common with
the penalty contact algorithm. In that case, they can be reduced by
 Reducing the normal penalty stiffness,

Adding penalty contact damping.
See Section 4.3.2 for details on the explicit penalty contact

algorithm.
In addition, other sources of damping such as Rayleigh damping
can reduce contact oscillations by damping the high frequency
modes that generate them.
 Oscillations in results can also occur when using the kinematic

constraint algorithm. These oscillations can be due to a mismatch
in the masses of the two contacting surfaces. See Section 4.7.3 for
more details.
4.6.2 Contact detection
 As explained earlier in this chapter, the contact conditions

prevent the contactor nodes from penetrating the target segments.
During each equilibrium iteration, the most current geometry of the
contactor and target surfaces is used to determine and eliminate the
overlap at the contactor nodes.

 For single-sided contact, the calculation of overlap at a

contactor node k consists of a contact search, followed by a
penetration calculation. The contact search starts by identifying all
possible target surfaces where node k can come into contact. For
each of these target surfaces:
 Find the closest target node n to node k.

Find all the target segments attached to node n.
Determine if node k is in contact with any of these
segments.
 If node k is in contact, update the information.
 If no appropriate target segment is detected, the contact
search is expanded beyond the target segments attached to
node n.
 For double-sided contact, the contact search algorithm uses time

tracking and checks whether the contactor node penetrated a target
segment between times t and t + Δt.
4.6.3 Suppression of contact oscillations (Solution 601)
 In some problems contactor nodes may oscillate during

equilibrium iterations between several (usually two) neighboring
target segments. Frequently, both solutions are acceptable. A
special procedure can be used to prevent such oscillations. This is
done by selecting a non-zero NSUPP parameter in the NXSTRAT
entry. In this case, the program records the pairing target segment
for each contactor node in the previous NSUPP iterations. Once
this array is full, and the contactor node is still in contact, and the
pairing target segment is one of those recorded in previous
iterations, the suppression feature is activated. The contactor node
from this iteration onwards is associated with only that target
segment. It may remain in contact with the segment, or in contact
with an infinite plane passing through the segment, or it can
separate from contact completely. The node is released from its
restrictions once iteration ceases, either because convergence is
reached, or due to non-convergence.
The default is not to use oscillation suppression (NSUPP=0).
 If this oscillation suppression feature is used, it is recommended

that NSUPP be set greater or equal to 5 and at least 5 less than the

maximum number of iterations.
 Note that there is memory overhead associated with this feature,

where an integer array of size NSUPP is defined for all contactor
nodes.
4.6.4 Restart with contact
 Changes in contact parameters are allowed between restarts,

with some exceptions. Some restrictions exist, such as no restart
from friction to frictionless and vice versa.
 The contact algorithm itself for a certain contact set can also
change in a restart. For this purpose, the contact algorithms can be
divided into two categories. The first category includes the
constraint function (implicit), Lagrange multiplier segment
(implicit), kinematic constraint (explicit), and penalty (explicit).
Restarts are possible between different algorithms in this category.
The second category includes the implicit and explicit Rigid Target
algorithms. Restarts are possible between these algorithms.
However, restarts are not allowed between the two categories.
4.6.5 Contact damping
 The contact damping feature allows the user to add normal and
tangential grounded viscous dampers to all contactor nodes in the
model. The damping is activated via the CTDAMP parameter in
NXSTRAT, and the normal and tangential damping coefficients are
CTDAMPN and CTDAMPT. Using the same value for both
normal and tangential direction results in isotropic viscous
damping. The damping force on each node is
FDamp  CN u N  CT u T
 This damping can be useful in static problems for stabilizing the

model especially when there are insufficient boundary conditions to
remove rigid body modes. It can also be useful in dynamic analysis
to dampen out high frequency numerical oscillations. The damping
can be set to act only at the initial time step (CTDAMP=1), or to be
constant throughout the analysis (CTDAMP=2).

 Using the initial damping option, the damping will be active at

the beginning of the first time step, and will be reduced gradually
(between iterations) until it fully dies out by the end of the first
time step. Thus the final solution at the first time step will be free
of any damping. Note that if contact is not established and nothing
else stabilizes the model, the program will not converge and will
give an appropriate warning message.
 Constant damping remains active throughout the analysis. In

this case, the program outputs the sum of all damping forces in the
output file, and the user must check that these forces are
significantly smaller than the sum of the reaction forces (also
written to the output file).
See Section 4.7.6 for modeling hints on using contact damping
to handle improperly supported structures and how to choose the
damping constants.
4.7 Modeling considerations

4.7.1 Contactor and target selection
 For some contact problems, the contactor and target surfaces in

a contact pair can be interchanged without much effect on the
solution. However, for many cases, one of the two alternatives is
better.
 If it is more important for the nodes of one surface not to

penetrate the other, then that surface should be the contactor. This
factor is usually important when one surface has a much coarser
mesh than the other as shown in Fig. 4.7-1. The coarse surface
should preferably be the target in this case. A related condition
occurs around corners or edges as shown in Fig. 4.7-2. The upper
surface should preferably be the contactor in this case.

Body 1 (contactor)
Body 2 (target)
Contactor nodes
Target segments
Fig. 4.7-1: Effect of incorrect contactor-target selection due to

mesh density
Contactor
surface
Target
surface
Fig. 4.7-2: Target selection for surfaces of

different sizes
 If one of the surfaces has mostly dependent degrees of freedom,

it should be the target. This dependency can be due to boundary
conditions, constraints or rigid elements. The surface can also be
rigid if its nodes are not attached to any elements. In that case too it
has to be the target (except in the explicit penalty algorithm where
this is permitted).

4.7: Modeling considerations
 If one surface is significantly stiffer than the other, it should

preferably be the target, unless one of the two conditions above
also exist.
4.7.2 General modeling hints
 Advanced Nonlinear Solution automatically defines the

direction of the contact surfaces on the faces of solid elements
(defined using the BSURFS or BCPROPS entries in 3-D or the
BEDGE or BLSEG entries in 2-D). For target contact surfaces
defined on shells (using the BSURF or BCPROP entries) the user
has to ensure that the correct direction is defined using the
BCRPARA entry (except when double-sided contact is used).
 In some cases, even though the contact surface is on the faces of

3D solid elements, it is more convenient to define the surface using
shell elements. In this case, fictitious shell elements should be
defined and referenced in the BSURF or BCPROP entries, and
TYPE should be set to COATING in BCRPARA. The program
will automatically transfer the contact surface to the underlying
solid elements and delete the fictitious shell elements.
 Rigid target surfaces can be modeled using nodes with no

degrees of freedom or nodes with enforced displacements for all
active degrees of freedom. As a result, a fine discretization of a
complex rigid surface geometry is possible with only a small
increase in the solution cost.
 The commands for 3-D contact surface definition all require the
contact surface nodes to be connected with 3-D solid or shell
elements. Therefore, to model a rigid target, dummy shell elements
should be used to define the surface. These shell elements are
removed from the model if they are found to be attached to a rigid
contact surface. A contact surface is deemed rigid if:
 It is the target of a contact pair in a contact set using the

rigid target algorithm, or
 The TYPE flag in the BCRPARA entry is set to RIGID.
 If the contact surface is rigid the MGP parameter in the

BCRPARA command can also be used to define a master node that

will control the motion of the rigid surface. Internally, rigid links
are created between the master node and all the nodes on the rigid
target.
 It is acceptable for the nodes on the contactor and target

surfaces to be coincident (have identical coordinates). In this case,
it is important to ensure that the two surfaces do not share the same
nodes.
 In general it is recommended that the lengths of segments on the

contactor and target surfaces be approximately equal. This is
particularly important if multiple contact surface pairs are
considered in the analysis or if the contact surface geometries are
complex.
 If required, a contactor surface can be modeled as almost rigid

by choosing a reasonably high Young's modulus for the finite
elements modeling the contactor surface. However, the stiffness of
the surface elements should not be excessively high and make the
model ill-conditioned.
 If the degrees of freedom of a node on a contactor surface are

used in constraint equations or attached to a rigid element (see
Section 5.8), the contactor node degrees of freedom should
preferably be independent.
 If the contact surfaces are smooth (i.e., the coefficient of friction

is small), the frictionless model is recommended as it is less costly
to use. It is also recommended that prior to any contact analysis
involving friction, the frictionless solution is first obtained,
whenever possible.
 It is not recommended that contact pairs with friction coexist

with contact pairs without friction in the same contact set.
 A contactor node should preferably not belong to more than one

contact surface in a contact set, otherwise the contactor node may
be over-constrained. If it is necessary for a contactor node to
belong to more than one contact surface, then contact compliance
should be used.

 For problems in which the contactor and target surfaces are

initially relatively close to each other and no significant sliding
between these surfaces is expected throughout the analysis, the
small displacement contact feature may be used. The analysis will
be faster in this case, since the relatively time consuming contact
search is only performed once, and convergence difficulties
associated with a contactor node oscillating between one target to
another are eliminated. It is the user’s responsibility however, to
make sure that the problem is suitable for small displacement
contact.
 The friction delay feature can sometimes lead to better

convergence since friction will only act once a converged contact
solution is established. This feature is also very useful for many
problems involving initial penetrations. In this case, the first time
step during which these initial penetrations are removed will be
frictionless.
 Restarting from frictionless contact to contact with friction and

vice versa is not possible. However, it can be done if the
frictionless analysis is replaced by a frictional analysis with a very
small friction coefficient.
 Ignoring initial penetrations is a useful option when these
penetrations are just a product of the finite element discretization,
meaning that they do not exist in the physical model. Fig. 4.7-3
illustrates one such case involving contact between concentric
cylinders. In this situation, if initial penetrations are eliminated, the
contact algorithm will try to push the penetrating contactor nodes to
the target surface segments in the first step, creating initial
prestressing. These initial penetrations and any prestressing that
they might cause are unrealistic. Ignoring them is useful in this
case. Note however, that if either cylinder is significantly rotated
the initial penetrations calculated at each contactor node (in the
initial configuration) will no longer be valid. In this case, the best
alternative would be to use a much finer mesh.

Outer cylinder
Inner cylinder
Contactor surface,
marked with
Target surface,
marked with
Geometry before
discretization
Overlap to ignore
Fig. 4.7-3: Analysis of contact between concentric cylinders, initial

penetration is ignored
 When higher order elements are used in contact, specifically the

10-node tetrahedral and the 20-node brick elements, tensile
consistent nodal contact forces can develop even when the
underlying contact tractions are compressive. The program can
accept such tensile forces as if they are compressive. This is done
via the TNSLCF flag in NXSTRAT. Accepting these tensile forces
gives more uniform results for problems involving the above
mentioned elements. However, it may slow down or even prevent
convergence in other problems. It is off by default.
 The option NXSTRAT DIAGSOL=2 outputs much useful

information. In Solution 601, this information is output every
equilibrium iteration. This information includes the number of
nodes going into and out of contact, the number of nodes
transisioning between stick and slip (in frictional contact), the
number of nodes exceeding the boundaries of target surfaces, etc.
Because this information is so useful, it is recommended that

NXSTRAT DIAGSOL=2 be used when there are convergence

difficulties (not the default).
4.7.3 Modeling hints specific to Solution 601
 It is recommended that the ATS method be used in contact

analysis (see Section 6.2.4). It can also be effective to use the low-
speed dynamics option of the ATS method (ATSLOWS parameter
of the NXSTRAT entry).
 Line search can sometimes be beneficial for contact problems.
 Frictional contact problems using the constraint function

algorithm can be sensitive to the choice of frictional regularization
constant (EPST parameter in BCTPARA entry). For most
problems, this parameter should be one or two orders of magnitude
smaller than the expected sliding velocity. Using an excessively
large value leads to a smoother friction law, which generally
converges faster but results in smaller frictional forces or more
sliding. Using an excessively small value enforces the Coulomb
law more accurately but is more likely to experience convergence
difficulties.
 Friction is not regularized or smoothed in the Lagrange

multiplier segment algorithm. This results in accurate enforcement
of stick and slip, but is more likely to experience convergence
difficulties.
 Geometric and material nonlinearities can highly depend on the

sequence of load application. Thus, for problems involving such
features, the load steps should be small. The time step Δt should
also be small in dynamic analysis and when time dependent
material constitutive relations (e.g., creep) are used.
 If rigid elements are connected to contact surface nodes, the

flexible option can be used. In this case, the rigid element does not
create any dependent degrees of freedom. This feature is activated
via the EQRBAR and EQRBE2 flags in the NXSTRAT entry.

 If a contact surface with corners or edges is modeled with
continuous contact normals, the normal vectors may be inaccurate

as shown in Fig. 4.7-4(a). In this case, switch to discontinuous

normals, use different contact surfaces for each smooth part
(Fig. 4.7-4(b)) or use a fine mesh close to the corners or edges
(Fig. 4.7-4(c)).
Contact surface
with sharp corners
Contact surface 1
Arrows correspond to normal a) Single contact surface

vectors pointing to exterior side
Contact
Contact surface 3
surface 1 Contact surface 1
(with fine
corner mesh)
Contact surface 2
b) Three separate contact surfaces c) Single contact surfaces with

fine mesh at corners
Fig. 4.7-4: Defining contact surfaces (with continuous normal vectors)

in the presence of corners
4.7.4 Modeling hints specific to Solution 701
 The penalty algorithm is preferred when both surfaces are rigid

or have many fixed or prescribed nodes.

 Large oscillations in the contact forces may occur when using

the penalty method even though the model is stable. These can be
reduced by adding a penalty damping term and/or reducing the
penalty stiffness.
 When using the penalty contact algorithm it is important to

check that the contact stiffnesses are properly selected. Unduly
small penalty stiffnesses will lead to excessive penetrations, while
unduly large penalty stiffnesses will lead to excessive oscillations
or unstable time integration.
 Large mismatches between the masses of contacting surfaces

should be avoided when using the kinematic constraint method.
This mismatch is common when contact involves a rigid surface
with a small mass and an applied force, as shown in Fig. 4.7-5. The
best solution in such cases is to minimize the mismatch by
increasing the mass of the rigid surface.
The inaccuracy in this case results from the way the contact is
enforced. The kinematic constraint method first predicts
displacements without contact then applies a contact correction.
The contact conditions are satisfied more accurately when the
penetrations in the predicted configuration are small which is
usually the case due to the small time step size of explicit analysis.
However, some cases such as that mentioned above lead to large
projected penetrations which results in incorrect contact conditions
and tensile contact forces.

F Small mass assigned to rigid surface
One node
in contact
Original configuration Correct solution
Tensile forces
Wrong contact
region
Projected configuration
(before contact correction) Final configuration
Fig. 4.7-5: Performance of kinematic contact algorithm when contact

surfaces have a large mass mismatch
 Large mismatches between the masses of contacting surfaces

can also lead to problems when using the penalty method. In this
case, the normal penalty stiffness required to avoid instability
(without reducing the time step) can be unduly small leading to
excessive penetrations. The best solution in such cases is to
minimize the mismatch by increasing the mass of the rigid surface,
or increase the penalty stiffness by setting it manually or by
reducing the time step.
4.7.5 Convergence considerations (Solution 601 only)
 When Solution 601 fails to converge during the incremental

analysis, the intermediate printout given by Solution 601 in the
output listing can provide some useful information (see Fig. 4.7-6).

OUT-OF- NORM OF ...
BALANCE OUT-OF-BALANCE NORM OF INCREMENTAL ...
ENERGY FORCE MOMENT DISP. ROTN. CFORCE ...
NODE-DOF NODE-DOF NODE-DOF NODE-DOF CFNORM ...
MAX VALUE MAX VALUE MAX VALUE MAX VALUE
box b box c box d box e box f

ITE= 0 1.14E+00 1.41E+02 9.99E-17 5.35E-02 5.12E-02 1.27E-15 ...
36-Z 35-X 31-Z 31-X 0.00E+00 ...
-1.00E+02 4.71E-17 -5.68E-03 -3.30E-02
ITE= 1 -1.29E-03 2.56E+01 1.92E-04 1.56E-02 2.45E-01 2.65E+03 ...

121-Z 31-X 64-Z 34-X 1.27E-15 ...
-9.85E+00 -1.06E-04 5.07E-03 1.26E-01
ITE= 2 3.32E-04 2.51E+01 1.88E-04 1.80E-02 1.77E-01 1.95E+03 ...

117-Z 31-X 64-Z 32-X 5.08E+01 ...
-9.66E+00 -1.04E-04 4.97E-03 -9.02E-02
ITE= 3 7.69E-02 4.46E+02 8.18E-04 1.04E-03 1.17E-02 1.95E+03 ...

64-Z 34-X 120-Z 33-X 2.00E+03 ...
3.21E+02 5.15E-04 -1.33E-04 -7.92E-03
... CONVERGENCE RATIOS CONVERGENCE RATIOS OUT-OF-BALANCE LOAD

... FOR OUT-OF-BALANCE FOR INCREMENTAL VECTOR CALCULATION
... ENERGY FORCE DISP. CFORCE BETA RATIO
... MOMENT ROTN. (ITERNS)
COMPARE WITH COMPARE WITH

ETOL RTOL DTOL RCTOL
1.00E-03 1.00E-02 (NOT USED) 5.00E-02
... 1.00E+00 1.41E+01 0.00E+00 1.27E-05

... 9.99E-17 0.00E+00
... -9.69E-03 2.56E+00 0.00E+00 2.65E+05 1.00E+00 -5.54E-02

... 1.92E-04 0.00E+00 ( 1)
... 2.49E-03 2.51E+00 0.00E+00 3.85E+01 1.92E-02 5.08E-03

... 1.88E-04 0.00E+00 ( 9)
... 5.76E-01 4.46E+01 0.00E+00 9.77E-01 1.00E+00 3.94E+03

... 8.18E-04 0.00E+00 ( 2)
Fig. 4.7-6: Solution 601 output listing of convergence criteria

during equilibrium iterations
 Three non-contact related norms are given: first, the energy

convergence criterion, the displacement and rotation convergence
criterion (boxes d and e), and the force and moment convergence
criterion (boxes b and c). Each box has 3 lines of output with the
top one giving the norm of the quantity, the second one giving the
node number corresponding to the maximum value, and the third
line giving the maximum value. See Chapter 6 for definitions and
more details on these norms.

 Box f of Fig. 4.7-6 shows the contact related norms. Parameter

CFORCE indicates the norm of the change in the contact forces
(between two iterations), and parameter CFNORM gives the norm
of the contact force vector.
 The following additional convergence criterion is used when

contact is present:
CFORCE
 RCTOL
max(CFNORM,RCONSM)
where RCTOL is the contact force convergence tolerance and

RCONSM is the reference contact force to prevent possible
division by zero in the contact convergence criterion above.
RCONSM and RCTOL are set in the NXSTRAT entry.
 When the maximum number of iterations is reached without

convergence, and all norms are decreasing, the maximum number
of iterations should be increased.

 When the norms are rapidly changing before convergence fails,
it is commonly caused by applying the load too quickly or using a
large time step.
 When CFNORM is stable but CFORCE changes rapidly during

equilibrium iterations, the contact can be oscillating between 2 or
more close solutions. In this case, try to change the time stepping,
or turn on the suppression of contact oscillations feature. When
CFNORM varies rapidly, usually the other three norms also vary.
 The additional output from NXSTRAT DIAGSOL=2 can also

assist in determining the reasons for convergence difficulties.
4.7.6 Handling improperly supported bodies
Many static problems depend on contact to provide the boundary

conditions necessary for a stable problem (one in which there are
no rigid body modes). Some examples are shown in Fig. 4.7-7.

Contact pair 1
Contact Contact
pair 2 pair 3
Applied Blankholder
load pressure
Symmetry
Fixed die
Fig. 4.7-7: Examples of improperly supported bodies
In such cases, the stiffness matrix is singular if the contact

constraints are inactive. Even if the constraints are active the
stiffness matrix is still not positive definite. The problem is more
serious if natural boundary conditions are applied (forces,
moments, or pressure). Weak springs can be added by the user to
make the model stable. However, the selection of appropriate
locations and stiffnesses for such springs may not be feasible.
Some models may be better suited for a dynamic analysis or the
low-speed dynamics feature. However, in many cases, this too is
not a feasible option. Therefore, several other modeling techniques
are available in Solution 601 to handle such problems. These are
stiffness stabilization, contact damping and limiting incremental
displacements. These techniques can be used separately or
combined in the same model.
Stiffness stabilization (see Section 10.6 for details).

This feature provides a stabilizing effect by scaling all diagonal
stiffness terms without affecting the right-hand-side load vector.
The outcome of each iteration will be affected, but the final

converged solution will not be (within the bounds of the

convergence tolerances). Since the stabilization constant in non-
dimensional, it should always be a small number. Typical values
are between 10-12 and 10-9.
Contact damping (see Section 4.6.5 for details).

Contact damping adds grounded viscous dampers to all contactor
nodes. Setting the damping to be only at the initial time step is
sufficient for some problems such as those in Fig. 4.7-7. When the
first time step converges contact must be established and damping
will have been removed. This way, the converged solution will be
free of any contact damping. Other problems however, require the
damping to be constantly present. In this case, the program outputs
the damping forces at every time step. These forces should be
compared with the reactions in order to ensure that damping is not
excessive.
The damping constants have units of force per unit velocity.
Hence, their proper value is problem dependent. If initial contact
damping is used to stabilize a problem involving two contact
bodies at least one of which is unsupported, and with a gap
between them, then a good estimate of the damping constants C N
and CT is one in which the gap is nearly closed in the first
iteration. Starting with the dynamic equations of motion (see
Equation 6.3-1) and canceling out the inertial term (static analysis),
and the stiffness term (since one or both bodies are initially
unsupported), we obtain
 R
C U
where C is the total damping matrix, which in this case is diagonal,

and R is the applied load vector. We can assume the normal and
tangential damping constants to be equal, the total damping
contribution to be the damping constant times the number of
contact nodes on the unsupported contact surface N (the top
circular body in the example in Fig. 4.7-7), and the velocity to be
approximately equal to the minimum initial gap between the two
bodies, g , divided by the time step size t . This leads to the
following value of the damping constants

R t
CN  CT 
Ng
where R is the sum of the applied loads at the first time step.
Note that this is only an estimate, but is frequently an acceptable
one.
Limiting maximum incremental displacement (see Section 6.2.1
for details).
Limiting the maximum incremental displacement per iteration is

useful when a load is applied to a body that is not initially in
contact. The model at that stage is unstable and even when stiffness
stabilization or viscous damping is used, the initial displacement
can be excessive leading the program away from the converged
solution, and thus making the return to the proper solution difficult.
Setting the limiting displacement to about the element length size
in this case would scale down the potentially huge displacement in
the first iteration so that the results remain close to the converged
solution.
Note that this feature does not stabilize the stiffness matrix, so
in many cases it may be necessary to use it together with stiffness
stabilization or viscous contact damping or both.
4.8 Rigid target contact algorithm

4.8.1 Introduction
 The rigid target contact algorithm is intended for use in

applications in which the target surfaces are considered to be rigid.
It is only available for 3-D contact.
Fig. 4.8-1 shows a typical application in metal forming.

Prescribed punch
displacement
Blank holder force
Punch
Blank holder
Blank
Die
a) Physical problem
Contact
surface Contact
4 (target) surface 2
(target)
Contact
Contact surface 3
surface 1 (target)
(contactor),
offsets used
to model blank
thickness
b) Modeling with contact surfaces
Fig. 4.8-1: Sample metal forming analysis using the

rigid-target contact algorithm

4.8: Rigid target contact algorithm
 A target surface can either be stationary, can translate as a rigid

body or can rotate as a rigid body.
 Contact can be frictionless or can include Coulomb friction.
 The rigid target contact algorithm is completely revised in

Advanced Nonlinear Solution of NX 5. However the rigid target
contact algorithm in Advanced Nonlinear Solution of NX 4 is
retained in Advanced Nonlinear Solution for backwards
compatibility. The revised rigid target contact algorithm of NX 5 is
the default.
Throughout this section, the rigid target contact algorithm in
Advanced Nonlinear Solution of NX 4 is referred to as the “NX4”
rigid target contact algorithm. This section does not describe the
NX4 rigid target contact algorithm; for information on the NX4
rigid target contact algorithm, see the NX Nastran 4 Advanced
Nonlinear Theory and Modeling Guide.
Models that were set up using the NX4 rigid target contact
algorithm may need to be revised when using the current rigid
target contact algorithm. We suggest that new models not be set up
using the NX4 rigid target contact algorithm.
 It is also possible to solve many problems involving rigid

targets using the constraint function and segment contact
algorithms described earlier in this chapter. However, the rigid
target contact algorithm described here is frequently more
effective, because the rigid target contact algorithm uses the
assumption of rigid targets to simplify the contact searching.
4.8.2 Basic concepts
4.8.2.1 Contactor surfaces
Similar to the other contact algorithms in Advanced Nonlinear

Solution, the contact surfaces are organized into contact sets. Each
contact surface consists of 3- or 4-node contact segments. A
contact pair consists of a contactor surface and a target surface. In
the rigid target algorithm, it is allowed for a contactor surface to be
in contact with more than one target surface simultaneously.

Contactor surface: The contactor surface definition includes the

possibility of offsets.
When there are no offsets specified, the contactor surface is
described entirely by the contactor nodes. (Fig. 4.8-2(a)).
Contact Contact
segment 1 segment 2
Segment normals
a) Contactor segments without offsets. Segment normals are not used.
Radius of sphere
Upper surface = contactor offset
Lower surface
b) Contactor segments with spherical offsets. Segment normals are not used.
Fig. 4.8-2: Contactor segments
When there are offsets specified, the offsets can either be

described using spheres centered around the contactor nodes (Fig.
4.8-2(b)), or using the contactor normals (Fig. 4.8-2(c)). In either
case, the offset magnitude is either constant or taken from the
current thickness of attached shell elements, as described below.

Upper contactor point
Offset
vector
Contactor
node normal
Lower contactor point
Contactor node normal = average of all segment normals

Offset vector = contactor node normal ´ offset magnitude
c) Contactor segments with offsets, normals used to describe offsets
Fig. 4.8-2: Contactor segments (continued)
When the offsets are described using the contactor normals,

offset vectors are constructed using the averaged contactor normals
and the offset magnitude. The upper and lower contactor points are
constructed from the contactor nodes and the offset vectors.
When the target surface is concave, it is possible for the contact
situation to be similar to the one shown in Fig. 4.8-3. In this case,
when the offsets are described using spheres, the center contactor
node cannot be in contact with both target segments at once, hence
the center contactor node will oscillate between them. The center
contactor node cannot be in contact with target edge 1 since edge 1
is farther away than either of the target segments. Equilibrium
iterations in static and implicit dynamic analysis will not converge,
because of the oscillation. However contact is correctly modeled
when the offsets are described using normals, because the center
contactor node can be in contact with target edge 1.

Contactor and target surfaces viewed from the side

for ease of visualization.
This contactor node cannot be in contact In contact with target segment 2

with both target segments at the same time.
In contact with target segment 1 Spherical offset
Contactor surface
Targ ent 2
et seg gm
men et se
t1 Targ
a) Spherical offsets, contact incorrectly modeled
Fig. 4.8-3: Concave target surface, contactor surface with offsets
4.8.2.2 Target surfaces
Each target surface is either stationary, or can rigidly move

(translate, rotate or a combination of translations and rotations).
4.8.2.3 Determination of contact between contactor and target
No contactor offsets: It is allowed for a contactor node to be in

contact with a target segment, target edge or target node. The
program searches for the target segment, edge or node for which
the absolute value of the distance d between the contactor node and
the target segment, edge or node is minimized, where the distance
is measured in the direction opposite to the target normal (Fig.
4.8-4). A positive distance corresponds to a geometric gap; a
negative distance corresponds to geometric overlap.

In contact with target segment 2

In contact with target segment 1
In contact with target edge 1
Normals offset
Targ ent 2
et seg gm
men
t1 a rg et se
T
Target edge 1
b) Normals offsets, contact correctly modeled
Fig. 4.8-3: (continued)
Notice that, for interaction between a contactor node and target

edge, or between a contactor node and target node, the normal
direction is taken from the line segment connecting the contactor
node and the target, as shown in Fig. 4.8-4.
Fig 4.8-5 shows two target segments with a common target
edge. The shaded volumes indicate which of the target entities any
contactor node is closest to. Notice that the shaded volume in
which the contactor node is closest to the target edge depends upon
the angles between the target segments attached to the edge.
Once the target segment, edge or node is determined, then the
contact gap is computed using
g  d  GAPBIAS
where GAPBIAS can be chosen to, for example, not model contact
even if there is geometric overlap. (The default for GAPBIAS is 0.)

Contactor node
Target segment
Closest point on
target segment
Target segment normal
n
a) Interaction between contactor node and target segment
Contactor node
Contactor node
d
d
Target edge
Closest point on Target node

n
target edge
n
b) Interaction between contactor node c) Interaction between contactor
and target edge node and target node
Fig. 4.8-4: Interaction between contactor node and target surface
If the corresponding gap is negative, and less than DEPTH, then

contact occurs, in other words  DEPTH  g  0 is the contact
condition. DEPTH can be chosen to limit the depth of the target
surface, exactly as in the other contact algorithms.

Contactor node in this shaded volume

Contactor node in this
is closest to edge 1.
shaded volume is
closest to segment 1.
Contactor node in this
shaded volume is
closest to segment 2.
Edge 1
Segment 1
Segment 2
Fig. 4.8-5: Interaction of contactor node with target segments and edges
Contactor offsets described using spheres: In this case, the

distance d is determined exactly as if there are no offsets. Then the
contact gap is computed using
g  d  OFFSET  GAPBIAS
where OFFSET is the offset magnitude. The process is illustrated

for interaction between a contactor node and target segment,
assuming that GAPBIAS = 0 (Fig. 4.8-6). The same idea is used for
interaction between a contactor node and target edge or node.

Contactor node
OFFSET
d
d - OFFSET
Target segment
Closest point on
target segment
Fig. 4.8-6: Interaction between contactor node and target segment,

spherical offsets
Contactor offsets described using normals: In this case, contact is

detected using the upper and lower contactor points instead of the
contactor nodes.
Oscillation checking: The search for the nearest target segment,

edge or node is performed every equilibrium iteration in Solution
601. During the equilibrium iterations, it is possible for the
contactor node to move in such a way as to be alternately in contact
with two neighboring segments. This is especially true if the target
surface is concave. When the contactor node oscillates between two
neighboring segments, the solution cannot converge unless
oscillation checking is turned on. When oscillation checking is
turned on, then, when oscillation is detected between two
neighboring segments, the contactor node is placed into contact
with the shared target edge. In many cases, this procedure allows
the iterations to converge, if in fact the contactor node “should”
have been in contact with the shared target edge.
Oscillation checking only forces the contact between the
contactor node and shared target edge for the current equilibrium
iteration. For the successive equilibrium iterations, the contactor
node is always in contact with the nearest target segment, edge or
node. So oscillation checking cannot force contact to the “wrong”
target segment, edge or node in a converged solution.

Contact normal force: The normal force corresponding to contact

is computed as Fn   kn g where the normal force acts in the
direction opposite to the target normal direction (Fig. 4.8-7). kn is
the contact normal stiffness, entered as a parameter (see Section
4.8.3 for hints about choosing kn ). kn can be considered to be a
penalty parameter.
Slope -kn
Fn
-DEPTH g
Tensile contact curve
Fig. 4.8-7: Normal contact force vs. gap
Tensile contact: During equilibrium iterations in Solution 601, a

node can temporarily be in “tensile contact”. The basic ideas for
tensile contact are illustrated in Figs. 4.8-8 to 4.8-10.
Fig. 4.8-8 shows the iteration history when tensile contact is not
used. For iteration ite-2, the contactor node and target segment
overlap. Hence contact is assumed between the contactor node and
target segment. For iteration ite-1, because of the relative motion of
the contactor node and target segment, the contactor node and
target segment do not overlap. For this iteration, no contact is
assumed between the contactor node and target segment. For
iteration ite, there is a large overlap because the contactor spring
unloads, since there are no forces acting on the contactor spring,
and the target does not provide any stiffness to the contactor node.
This large overlap causes large contact forces, which can cause
trouble in convergence in the successive iterations.

Spring is compressed
Contactor node Spring is uncompressed
Target segment
a) Iteration ite-2, b) Iteration ite-1, c) Iteration ite,

contact force Fite-2 no contact force contact force Fite
is compressive is large and compressive
Fig. 4.8-8: Iterations when tensile contact is not used
Fig. 4.8-9 shows the iteration history when tensile contact is

used. Now, in iteration ite-1, tensile contact is assumed between the
contactor node and target segment. In tensile contact; the target
surface still provides stiffness to the contactor node. Hence the
overlap in iteration ite is small.
Contactor node
Target segment

contact force Fite-2 contact force Fite-1 contact force Fite
is compressive is tensile is compressive
Fig. 4.8-9: Iterations when tensile contact is used
Fig.4.8-10 shows the iteration history when tensile contact is

used, and the gap is large. In iteration ite-1, tensile contact is
assumed, and in iteration ite, no contact is assumed.
It is seen that tensile contact speeds up the convergence when
the converged solution is in contact, and slows down the
convergence when the converged solution is not in contact.

Contactor node
Target segment

contact force Fite-2 contact force Fite-1 no contact force
is compressive is tensile
Fig. 4.8-10: Iterations when tensile contact is used, converged solution

not in contact
It is not permitted for a solution in which tensile contact is

present to converge, unless the tensile forces are all less than the
value of a user-input parameter (see Section 4.8.3). Hence the
tensile contact feature does not affect the converged solution.
Tensile contact is always used in Solution 601.
4.8.2.4 Frictional contact
The friction force is calculated using the relative sliding velocity

between the target and contactor. The relative sliding velocity u f
is calculated from the velocities of the contactor and target using
u f   u c  u t     u c  u t   n  n
where u c is the velocity of the contactor node, u t is the velocity

of the target and n is the target normal.
In static analysis, the contactor and target velocities are
calculated using the nodal incremental displacements divided by
the time step. In dynamic analysis, the contactor and target
velocities are taken from the nodal velocities.
The friction force magnitude is computed using

u f
Ff   Fn , u f  u f min
u f min
  F , u f  u f min
where Fn is the normal contact force and  is the Coulomb

friction constant (Fig. 4.8-11). u f min is the minimum sliding
velocity, entered as a parameter (see Section 4.8.3 for hints about
choosing u f min ). The direction of the friction force is always
opposite to u f .
Ff
mFn
Sliding
. . .
-uf uf u
-mFn
Sliding Sticking
Fig. 4.8-11: Friction force vs. velocity
When u f  u f min , the friction is sticking, otherwise the friction is

sliding.
Oscillation checking with friction: During equilibrium iterations in

Solution 601, it is possible for the contactor node to undergo
“sliding reversals”. Namely, the contactor node slides in one
direction for an equilibrium iteration, then reverses sliding
direction for the next equilibrium iteration. When sliding reversals
occur, the solution cannot converge unless oscillation checking is
turned on. When oscillation checking is turned on, then, when
sliding reversals are detected, the contactor node is placed into

sticking contact, even if the sticking force is larger than the sliding
force, and convergence is prevented for the current equilibrium
iteration.
Oscillation checking only forces sticking friction for the current
equilibrium iteration. For successive equilibrium iterations, the
frictional state (sliding or sticking) is determined as usual from the
sliding and sticking forces. So oscillation checking cannot
converge to a solution in which the frictional state is wrong.
4.8.3 Modeling considerations
Selection of rigid target contact: For Solution 601, set TYPE=2 on

the BCTPARA card. For Solution 701, set XTYPE=3 on the
BCTPARA card.
Algorithm used: The current rigid target contact algorithm is

selected by default. To select the NX4 rigid target contact
algorithm, set RTALG=1 on the NXSTRAT card.
Modeling of target surfaces: If the target surface translates or

rotates, all of the nodes on the target surface must be connected to a
“master node”, either using constraint equations or using rigid
links. For example, in Fig. 4.8-12, all of the nodes on the lower
target surface are connected to a master node using rigid links.
It is not allowed for the nodes on a target surface to have
independent degrees of freedom. All degrees of freedom for the
nodes on a target surface must be fixed or constrained.
Modeling of contactor surfaces: The amount and description of

offset is determined by the BCTPARA parameters OFFTYPE,
OFFSET and OFFDET. If OFFTYPE=0 (the default), there is no
offset. If OFFTYPE=1, a constant offset of value OFFSET is used.
If OFFTYPE=2, an offset equal to half of the current shell element
thickness is used.

Fixed target surface

Fixed nodes
Moving target surface

Slave nodes
Rigid links
Master node, independent degrees of freedom
Fig. 4.8-12: Modeling of fixed and moving target

surfaces
When there is an offset, then BCTPARA parameter OFFDET

determines the description of the offset. If OFFDET=0, then
Advanced Nonlinear Solution determines the offset description
(either spheres or normals, see Section 4.8.2.1). The criterion used
by Advanced Nonlinear Solution is that an offset description of
spheres is used for each target surface that is convex or flat, and an
offset description of normals is used for each target surface that is
concave. If OFFDET=1, then the offset description is spheres, and
if OFFDET=2, then the offset description is normals.
When normals are used for the offset description, small steps
should be used in Solution 601. This is because the offset vectors
are assumed to remain constant during the equilibrium iterations. In

particular, at convergence, the offset vectors corresponding to the

previous converged solution are used.
Determination of contact, modeling issues: Contact is affected by

variables GAPBIAS and DEPTH, as described in Section 4.8.2.3.
GAPBIAS is set using BCTPARA parameter GAPBIAS (default
=0) and DEPTH is set using BCTPARA parameter PDEPTH
(default=0).
It is possible for the closest target segment, edge or node to not
be the expected one. An example is shown in Fig. 4.8-13. In this
example, the rim of the wheel is modeled with target segments.
Because the distance between a contactor node and a target
segment is measured in the direction of the target segment normal,
a contactor node interacts with the lower target surface, and the
contact algorithm detects a large overlap between this contactor
node and the lower target surface.
Another example is shown in Fig. 4.8-14. In Fig. 4.8-14(a),
there is a gap between the contactor node and the closest target
segment, as expected. In Fig. 4.8-14(b), the punch has moved
upward relative to the contactor node. Now there is a large overlap
between the contactor node and the closest target segment. This
segment is the only segment with a normal that points in the
direction of the contactor node.
In both Fig. 4.8-13 and Fig. 4.8-14, the large overlap is
unintended. In Solution 601, the equilibrium iterations would most
likely not converge. In Solution 701, the large forces between
contactor and target would overdistort the elements attached to the
contactor node.
One way to avoid the large overlaps is to use the DEPTH
feature so that contact is not detected between the contactor node
and the incorrect target segment. Another way to avoid this issue is
to create additional target segments as shown. Then the contactor
node is closest to one of the additional target segments.

Contactor node is closest to lower target. Contactor node is not closest to upper target,
since there is no upper target segment with normal that points in the direction of the
contactor node.
Contactor node
is closest to
upper target
Upper target
Lower target
a) Orthographic view of wheel b) Side view of wheel
Contactor node cannot be Contactor node is closest

in contact with lower target to target edge
Upper target
Upper target
Lower target
Lower target
DEPTH
c) DEPTH feature used d) Wheel modeled with additional

target surfaces
Fig. 4.8-13: Modeling of a wheel

Closest point Closest point

on target surface on target surface
that interacts that interacts
with contactor node with contactor node
Target surface
Contactor
node
Contactor
node
a) Intended interaction of contactor b) Unintended interaction of contactor

node and target surface; contactor node and target surface; contactor is
node is not in contact with target in contact with target with large overlap
Closest point
on target surface
that interacts
with contactor node
DEPTH
c) DEPTH feature used, contactor node

is not in contact with target d) Additional target segments used, contactor
is not in contact with target
Fig. 4.8-14: Modeling of a punch
Choice of kn : kn is set using BCTPARA parameter NCMOD. The

default value of the normal contact stiffness kn is 1E11. However,
kn can be chosen for optimal convergence. Note that increasing kn
causes the maximum overlap between the contactor and the target
to become smaller. Also, increasing kn can lead to convergence
difficulties.
We recommend that the smallest value of kn be used such that
the maximum overlap is still acceptably small. For example, if the
target surface is curved, there will be a geometric error associated
with using a coarse contractor surface (Fig. 4.8-15). There is no
advantage if the maximum overlap is less than the geometric error.
So, if the mesh is coarse, a large maximum overlap can be used.

Contactor surface, coarse mesh used
Geometric error
Target surface, fine mesh used
Fig. 4.8-15: Modeling of a curved target

surface
Another consideration for the choice of kn is the following.

Because the rigid target algorithm is node-based, and because the
contact stiffness is the same for each node in contact, the stresses
computed in higher-order elements on the contactor surface will be
inaccurate, if kn is too small. For example, in a problem involving
pressing an element onto a contact surface, kn should be greater
EA
than 100 where E is the Young’s modulus, A is the contact
nL
area, L is the element thickness (in the contact direction) and n is
the number of nodes on the contact area. The basic concept is
illustrated in Fig. 4.8-16.
This issue also arises when lower-order elements are used, but
when lower-order elements are used, the variation in the consistent
nodal point forces is much less, so kn can be smaller for the same
accuracy in the element stresses.

Uniform pressure load
Young’s
modulus E Contact
L
area A
Contact forces acting on contactor nodes

are all nearly equal
a) Soft target surface, kn small, element stresses are

inaccurate
Uniform pressure load
Contact forces acting on contactor nodes

are nearly equal to consistent forces
corresponding to pressure load
b) Hard target surface, kn large, element stresses are

accurate
Fig. 4.8-16: Higher-order elements and rigid-target contact

Time step for Solution 701: For Solution 701, the time step should
be smaller than
m
t  2
kn
This formula is derived from the following considerations.

Consider a single contactor node with mass m and no additional
stiffness or damping, with a velocity normal to the target. If this
node just touches the target at time t  t , and penetrates the target
at time t , the node should remain in contact at time t  t . The
choice of t in the above equation satisfies this condition.
Clearly, decreasing kn will increase the time step t .
A node that is out of contact at time t  t , in contact at time t
and out of contact at time t  t is said to have had a contact
reversal (Fig. 4.8-17).
t+Dtu.
Contactor node at t+Dt
Contactor node at t-Dt

t-Dtu.
Contactor node at t
Fig. 4.8-17: Contact reversal due to too

large time step in Solution 701
Time step selection in frictional contact: The time step size will
affect the frictional velocities and hence the results. This is
because, in static analysis, the nodal velocities used in the friction
calculations are calculated as the incremental displacements
divided by the time step.
In those parts of the analysis in which friction is important, a
“realistic” time step should be used.

In those steps of the analysis in which friction is not important,

a large time step can be used, which causes the velocities to be
small. For example, in metal forming analysis, a large time step
size can be used when establishing the blank holder force, and
during springback calculations.
Choice of u f min for frictional contact: u f min is set using

BCTPARA parameter SLIDVEL. The default value of the
minimum sliding velocity u f min is 1E-10. However u f min can be
chosen for optimal convergence. Decreasing u f min can lead to
convergence difficulties.
We recommend that u f min be chosen from experimental data, or
that the largest acceptable value of u f min be used.
Time step for Solution 701 for frictional contact: For Solution 701,
the time step should be smaller than
2mu f min
t 
 Fn
to prevent reverse sliding. This formula is derived from the finite

difference equation corresponding to explicit time integration,
when applied to a single contactor node with mass m and no
additional stiffness or damping, sticking to the target, but with a
nonzero sticking velocity. If t is larger than the value in the
above equation, the velocity will increase, and eventually the node
will slide. The sliding will then tend to “reverse”, that is, for a
given time step, the sliding direction will be opposite to the sliding
direction in the previous step (Fig. 4.8-18).
2mu f min
Note that when the time step is greater than t  , the
 Fn
solution is still stable.

Contactor node .at t Contactor node at t+Dt

tu t+Dtu.
f f
Fig. 4.8-18: Reverse sliding due to too large

time step in Solution 701
Automatic time step selection in Solution 701: When using the

automatic time step selection options in Solution 701, the time step
returned from the rigid target contact algorithm is
mi
t  min
i kn
where the minimum is taken over all contactor nodes. Notice that
friction is not considered in the automatic time step selection; this
is because the model remains stable even if the time step is larger
than the friction time step discussed above.
Birth/death: Rigid target contact surfaces can have a birth and

death time, similar to other contact surfaces. The birth and death
times are set by BCTPARA parameters TBIRTH and TDEATH
(default =0.0, corresponding to no birth and death).
Other user-input parameters:
BCTPARA TFORCE
The maximum tensile force for a node in tensile contact for which
convergence is allowed (default value 0.001). All nodes in tensile
contact must have a tensile force less than this value for the
solution to converge. Tensile contact is not used in Solution 701.
BCTPARA OCHECK
If OCHECK = 0, then oscillation checking (described in Section

4.8.2.3) is turned off. If OCHECK = ITE>0, then oscillation

checking is activated after equilibrium iteration ITE. The default is

5. Oscillation checking is not used in Solution 701.
4.8.4 Rigid target contact reports for Solution 601
The following messages are output at the end of each converged

solution.
Maximum overlap at convergence:
Meaning: self-explanatory
Recommend: If the maximum overlap is too large, increase kn ;

if the maximum overlap is too small, decrease kn .
Maximum tensile contact gap during iterations for nodes in contact

at convergence:
Meaning: A node that is in contact at the start of the time step

may temporarily move out of contact during the iterations, then
go back into contact before convergence. This report item
reports the maximum contact gap of all such nodes. When the
tensile contact gap is large, then convergence may be difficult.
Recommend: Either reduce the time step or decrease kn to

reduce the tensile contact gap.
Maximum friction velocity at convergence:
Meaning: For nodes in frictional contact, this is the maximum

friction velocity of a node (either sticking or sliding).
Recommend: If the maximum velocity is less than u f min , and

the corresponding node should be sliding, decrease u f min .

Number of nodes in contact, number of nodes in sticking contact,

number of nodes in sliding contact:
Meaning: Self-explanatory. Each node is counted once for each

target surface that the node is in contact with. So a node that is
in contact with two target surfaces simultaneously is counted
twice.
Change of contact status during iterations:
Meaning: The number of nodes that switch contact status (not in

contact to in contact, or vice versa), is reported. If there are
many nodes that switch contact status, this may cause
convergence difficulties.
Recommend: Either reduce the time step or decrease kn .
In contact at convergence, in tensile contact during iterations.
Meaning: The number of nodes which were in tensile contact

during the iterations (meaning that the nodes were almost out of
contact) and in contact in the converged solution. When there
are many such nodes then convergence may be difficult.
Recommend: Either reduce the time step or decrease kn to

reduce the likelihood that nodes go into tensile contact.
Change in frictional contact status during iterations:
Meaning: The number of nodes that change frictional contact

status (from sticking to sliding or vice versa) is reported.
Recommend: If there are many nodes that switch frictional

contact status, reduce the time step or increase u f min .
The following messages are output at the end of each solution that
did not converge.

Maximum change of contact force at end of iterations:
Meaning: The contactor node for which the contact force had
the largest change is output.
Recommend: Examine the model near that contactor node for

hints about why the solution did not converge.
Change of contact status at end of iterations:
Meaning: The number of nodes that are changing contact status

at the end of the iterations.
Recommend: Reduce the time step or decrease kn .
Sliding reversal at end of iterations:
Meaning: The number of nodes that are undergoing sliding

reversals at the end of the iterations.
Recommend: Reduce the time step or increase u f min
Change of target entity at end of iterations.
Meaning: The number of nodes that are oscillating between

different target entities at the end of the iterations.
Recommend: If oscillation checking is not turned on, turn it on.

Otherwise refine the target surfaces, or reduce the time step.
4.8.5 Rigid target contact report for Solution 701
The following items are output for each time step in which results
are printed or saved:
Number of nodes in contact, number of nodes in sticking contact,

number of nodes in sliding contact:
Meaning: See the corresponding message in Section 4.8.4.

Maximum overlap since solution start; maximum overlap since last

report:
Recommend: See the corresponding recommendations in

Section 4.8.4.
Maximum friction velocity since solution start, maximum friction

velocity since last report:
Recommend: See the corresponding recommendations in

Section 4.8.4.
Contact reversals since solution start, since last report:
Meaning: This is a count of the total number of contact

reversals. Also the number of contact reversals for the node with
the most contact reversals is given, along with the mass of the
node.
Recommend: To reduce the number of contact reversals, either

reduce the time step or decrease kn .
Sliding reversals since solution start, since last report:
Meaning: This is a count of the total number of sliding

reversals. Also the number of sliding reversals for the node with
the most sliding reversals is given, along with the mass of the
node.
Recommend: To reduce the number of sliding reversals, either

reduce the time step or increase u f min .
4.8.6 Modeling hints and recommendations
 For a time step in which contact is established over a large area,

many equilibrium iterations may be required in Solution 601. This

is because the solution cannot converge until the nodes in and out
of contact are determined, and it may take many equilibrium
iterations to determine which nodes are in and out of contact. An
example is shown in Fig. 4.8-19. The ATS cutback method will not
be effective for this time step. Rather, the maximum number of
iterations should be set very large, so that the program can find the
converged solution.
Prescribed displacement
Target surface 1 Contactor surfaces
Target surface 2
a) Solution step before contact with target surface 2
Possible contact in this area, many

equilibrium iterations may be required
Target surface 1
to accurately determine contact here.
Target surface 2
b) Solution step after contact with target surface 2
Fig. 4.8-19: Establishment of contact over a large area during a

solution step
 When forming a part that is relatively thin, setting the

PLASALG flag of the NXSTRAT card to 2 can allow the use of
larger time steps.

 The contact search algorithm may take a relatively long time for
the first iteration of the first time step. Similarly, the contact search
algorithm may take a relatively long time for the first iteration of
any time step in which a contact set is born.
 As the contactor surface is refined, keeping k n constant, the

overlap and contact force will decrease at each contactor node.
Hence kn may need to be adjusted as the mesh is refined. In
general, as the mesh is refined, kn can be decreased in order to
keep the overlap reasonable.
 Convergence in Solution 601 may become difficult when
contactor nodes that were not in contact with the target suddenly
interact with the target. An example is illustrated in Fig. 4.8-20.
Eventually the contactor nodes on the right will come into contact
with the target, and convergence may be difficult. Alternate ways
to model this situation are shown in Fig. 4.8-20.
Contactor node in contact

Contactor node not in contact
Drawing
direction
a) Poor modeling
b) Good modeling
c) Good modeling
Fig. 4.8-20: Contactor nodes suddenly coming into contact

Another example is shown in Fig. 4.8-21. In Fig. 4.8-21(a), the

top target surface is flat, and the indicated node suddenly comes
into contact with the top target surface. Convergence is very
difficult, because a very small change in the position of the
indicated node can cause the contact status of that node (and hence
the contact force) to change abruptly. In Fig. 4.8-21(b), the top
target surface has a round corner, and the indicated node gradually
comes into contact with the top target surface. Convergence is
easier, because a very small change of the position of the indicated
node results in only a very small change in the contact force.
In metal forming analysis, this situation in frequently
encountered in the modeling of the blank holder. The modeling is
easiest if the blank holder is modeled as a flat target surface. But
convergence is easier if round corners are added to the blank holder
wherever nodes on the blank are anticipated to contact the blank
holder during drawing.
This node This node

suddenly comes gradually comes
into contact into contact
a) Poor modeling b) Good modeling
Fig. 4.8-21: Additional example of nodes suddenly coming into contact

Chapter 5: Loads, boundary conditions and constraint equations
5. Loads, boundary conditions and constraint

equations
5.1 Introduction
The objective of this chapter is to present the various options
available in Advanced Nonlinear Solution for the description of
applied loads, boundary conditions, and constraint equations.
 Table 5-1 lists the Case Control commands used for loading,
boundary conditions, and initial conditions.
 Note that the selected DLOAD set can be used for any time
varying loads in both static and dynamic analysis. Similarly, the
selected LOAD set can be used for defining constant loads in both
static and dynamic analyses.
 Table 5-2 lists the load, boundary condition and initial condition
Bulk Data entries supported in Advanced Nonlinear Solution.
 Table 5-3 lists the Bulk Data entries used for combining applied
loads and/or enforced displacements in Advanced Nonlinear
Solution.
Table 5-1: Case Control commands in Advanced Nonlinear Solution
Case Control Command Comments

DLOAD Select load set (time varying)
LOAD Select load set (non-time varying)
Select single-point constraint set (including
SPC
enforced displacement)
MPC Select multipoint constraint set
Select initial conditions set (displacements
IC
and velocities)
TEMPERATURE Select initial and applied thermal load sets
BOLTLD Select bolt preload set
DTEMP Select time-dependent temperature set

5.1: Introduction
Table 5-2: Bulk data entries for defining loads, boundary conditions and
constraints
Bulk Data Entry Comments

FORCE, FORCE1, FORCE2 Concentrated force on nodes
MOMENT, MOMENT1, Concentrated moment on nodes
MOMENT2
SPC1,2,4, SPC14, SPCADD4 Fixed or enforced degrees of freedom
on nodes
SPCD4 Enforced displacement on nodes
PLOAD Uniform pressure on shell element or
3-D solid face
PLOAD1 Distributed load on beam element
Concentrated force on beam nodes
PLOAD2 Uniform pressure on shell element
PLOAD4 Pressure or distributed load on shell or
3-D solid face
PLOADE1 Varying pressure on plane stress or
plane strain 2-D solid element
PLOADX1 Varying pressure on axisymmetric 2-D
solid element
TEMP Applied temperature on nodes
TEMPD Applied default temperature
DTEMP Time-dependent temperature loading
GRAV Mass proportional inertial load
RFORCE Centrifugal load
RFORCE1 Centrifugal load with more than one
load applied to the model
MPC, MPCADD Define multipoint constraints
3
TIC Initial displacement and velocity on
nodes
BOLTFOR Preload force on bolts

Table 5-2: (continued)

TEMPBC Applied temperature on nodes in heat
transfer analysis
QHBDY, QBDY1 Uniform heat flux on boundary
element
QBDY2 Varying heat flux on boundary
element
QVOL Uniform volumetric heat addition
CONV Free convection on boundary element
RADBC Space radiation on boundary element
Notes:
1. SPC can also enforce displacement.

2. If enforced displacements are always 0.0 they become a boundary
condition.
3. Initial conditions are discussed in Section 10.1.
4. Can also be used to fix or enforce temperature in a heat transfer analysis.
 The LOAD entry is used for combining loads that are constant
throughout the analysis while DLOAD is used for combining
time-varying loads. The DLOAD entry references a load defined
through a TLOAD1 entry. The TLOAD1 entry references the type
of load (applied load or enforced displacement), as well as the table
entry (TABLED1 or TABLED2) defining the time variation of the
load.
 Both LOAD and DLOAD can be used in static and dynamic

analyses in Advanced Nonlinear Solution.

5.1: Introduction
Table 5-3: Bulk data entries for applying loads and enforced
displacements
Defines a linear combination of

LOAD
constant loads
Defines a linear combination of time
DLOAD varying loads (by combining different
TLOAD1 entries)
Defines time varying loads and
TLOAD1
enforced motion
Defines the time functions used by the
TABLED1, TABLED2
loads
 
 A time function is defined as a series of points t , f i  t  in
which t is time and fi  t  is the value of time function i at that time.
Between two successive times, the program uses linear
interpolation to determine the value of the time function.
 Advanced Nonlinear Solution does not support subcases. If

subcases are only used to change the applied load in a static
analysis, then they can be equivalently defined in Advanced
Nonlinear Solution as time-varying loads in a single case.

 A typical time-varying load such as the enforced displacement
shown in Fig 5.1-1 (on the y direction of node 100) will be applied
as follows:
DLOAD, 1, 1.0, 10.0, 5

TLOAD1, 5, 3,, DISP, 7
SPCD, 3, 100, 2, 1.0
TABLED2, 7, 0.0,
,0.0, 2.0, 2.0, 4.0, 4.0, 4.0, 5.0, 0.0,
,7.0, -2.0, 8.0, 0.0, ENDT
TSTEP, 1, 8, 1.0, 4

Time functions:
f(t)
4
3
2
1
t
2 4 6 8 10
-1
-2
Resulting load values for tR = 10f(t):
Time 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
tR 20 30 40 40 40 0 -10 -20 0
Fig. 5.1-1: Typical time-varying load


Note that the TSTEP entry is used for both linear and nonlinear
analyses. In this case, 8 steps of size 1.0 are selected with output
every 4 steps.
Note that in Solution 701 with automatic time step selection, the
above input will not result in 8 steps. Instead, the critical time step
for the model will be used and output of results will be done as
soon as the solution time exceeds 4.0 and 8.0. See Section 7.1 for
details.
 The LOAD case control command can point to a LOAD entry

or to individual loads, and similarly the DLOAD case control
command can point to a DLOAD entry or directly to a TLOAD1
entry. The initial and applied temperature load sets must be
selected by the TEMPERATURE case control command if needed.
The active initial conditions must be selected by the IC case control
command.
 Boundary conditions can be grouped into two classes: essential

and natural boundary conditions (see ref. KJB, Section 3.3.2).

5.1: Introduction
Essential boundary conditions can be enforced displacements or

rotations. Natural boundary conditions include all applied forces
and moments.
 Displacement boundary conditions include fixed nodal degrees

of freedom, enforced displacements and constraint equations.
 Force and moment boundary conditions include numerous types

of applied loading available in Advanced Nonlinear Solution.
 All displacement and force boundary conditions are referred to

the displacement coordinate system at the node at which they act.
 The externally applied load vector used in the governing

equilibrium equations is established using contributions from the
various applied loads.
For concentrated loads, the contributions of these nodal loads
are directly assembled into the externally applied load vector.
For pressure loading, distributed loading, centrifugal loading
and mass proportional loading, Advanced Nonlinear Solution first
calculates the corresponding consistent nodal load vectors
(consistent in the sense that the principle of virtual work is used)
and then assembles these load vectors into the externally applied
load vector. The evaluations of the consistent nodal load vectors for
the various types of loading are described in the following sections.
 Temperatures in Advanced Nonlinear Solution are used in

conjunction with material models which include temperature
effects.
 The definition of the time variation of the externally applied

load vector for the various time steps in the solution period depends
on whether automatic step incrementation is used or not.
Time variation of externally applied loads when automatic step

incrementation is not used: Each applied load or enforced
displacement is associated with a time function which defines the
time variation of the load throughout the solution period.
 In a static analysis in which time-dependent effects (such as

creep or friction) are not included in the material models, time is

a "dummy" variable which is used, via the associated time

function of each applied load, to define the load intensity at a
step. Thus, the time step increment directly establishes the load
increments. So, in the example shown in Fig. 5.1-2, the same
solution is obtained regardless of the size of the time step
increment.
tR
tR tR
200 200
150 l 150 l
100 l 2 steps 100 l 2 steps
1 2 t 2 4 t
Run 1: Dt = 1.0 Run 2: Dt = 2.0
Note: identical results are obtained in Run 1 and Run 2

for a linear static analysis.
Fig. 5.1-2: Example of time varying loads
 In a dynamic analysis or if time-dependent effects are

included in the material models in a static analysis, time is used
in a similar way to define the load intensity of an applied load at
a step. However, in these cases, time is a "real" variable because
the time step increment is employed in the actual integration of
the equations of motion in a dynamic analysis, and in the
integration of the element stresses in a creep analysis. Hence, in
these cases the choice of the time step increment is no longer
arbitrary.

5.1: Introduction
Time variation of externally applied loads when automatic step

incrementation is used: Two options are available:
 Using the automatic-time-stepping (ATS) procedure (see

Section 6.2.4), the loads are defined for all times Δt, 2Δt,...
as for no automatic step incrementation. In addition, when
the algorithm subdivides a time (load) step, the load vector
is established by linear interpolation of the load vectors at
times t and t + Δt.
 Using the load-displacement-control (LDC) procedure (see
Section 6.2.6), the applied loads are not associated with any
time function and the time variation of the loads cannot be
specified by the user. The contributions from all the loads
are assembled into a constant load vector denoted as the
reference load vector. During the response calculation, this
reference load vector is scaled proportionally using a load
multiplier (in general different from one step to the next)
automatically computed by the program.
 The activation of the various applied loads can be delayed using

the X1 field in the TABLED2 entry. The arrival/delay time option
does not apply, however, to centrifugal and mass-proportional
loading, see Section 5.4.
The specification of a nonzero arrival time corresponds to a
shifting of the associated time function forward in time. If the time
function is used by a force boundary condition, this corresponds to
using a time function multiplier of zero for all times t smaller than
the arrival time; see illustration given in Fig. 5.1-3. However, if the
time function corresponds to a enforced displacement/rotation the
associated degrees of freedom are assumed to be free prior to the
arrival time (not having a zero prescribed value).
The arrival time feature is used only if NXSTRAT TFSHIFT=1
(the default). If NXSTRAT TFSHIFT=0, any arrival time data
specified by TABLED2 X1 is ignored.

Time functions:
f(t) Input time function Time function shifted
by arrival time
4
2
Time at solution
1 start=0.0
t
2 4 6 8 10
-1 Arrival
time Load active
-2
Solution period
-3
Fig. 5.1-3: Example of the use of the arrival time option
5.2 Concentrated loads

 Concentrated loads are nodal point forces applied at the
specified nodes using the FORCE, FORCE1, or FORCE2 entries.
Concentrated moments are also applied to specific nodes using the
MOMENT, MOMENT1, or MOMENT2 entries. Concentrated
forces on beam nodes can also be applied using the PLOAD1 entry.
 The direction in which a concentrated load acts depends on the

displacement coordinate system assigned to the node.
 Note that concentrated moments applied to shell nodes convert

the shell nodes automatically to 6 degree of freedom nodes. This is
done since the local V1 and V2 directions at shell nodes are
unknown to the user, and hence cannot be used in defining
moments.
 When the FORCE1 or MOMENT1 entries are used in a large

displacement analysis, they can be follower loads, meaning that the
direction of the applied force or moment can be updated during the
simulation based on the current coordinates of the G1 and G2

5.2: Concentrated loads
nodes. This however is only possible if either G1 or G2 is set to be

the node of load application.
 The direction of a follower load can be controlled using RBAR

or RBE2 rigid elements (see Section 2.7). An example is given in
Fig. 5.2-1.
G1
Rigid element
Node of load
application
G2
Thin cantilever
Follower force
at t = 0
a) Configuration at time t=0
Note: t qk is the rotation at node k at time t.
G1
tq
k
Rigid element
G2
Follower force
at time t
b) Configuration at time t
Fig. 5.2-1: Example of the use of a rigid element to

establish the follower load direction

5.3 Pressure and distributed loading

 Some examples of pressure and distributed loading are shown in
Fig. 5.3-1.
a) Axisymmetric 2-D solid element b) 3-D solid element

PLOADX1 PLOAD, PLOAD4
l
l
l
l
l
l
l
c) Beam element d) Shell element

PLOAD1 PLOAD, PLOAD2, PLOAD4
Fig. 5.3-1: Examples of distributed and pressure loading
 Distributed loads can be applied to beam elements using the

PLOAD1 entry. This entry can also be used to apply concentrated
forces on beam nodes.
 Pressure loads can be applied to shell elements using the

PLOAD or PLOAD4 entries, and to shell 3-node and 4-node
elements only (CTRIA3 and CQUAD4) using the PLOAD2 entry.
 Pressure loads can be applied to axisymmetric 2-D solid

elements using the PLOADX1 entry. Pressure loads are input as
force per unit area.

5.3: Pressure and distributed loading
 Pressure loads can be applied to plane stress or plane strain 2-D

solid elements using the PLOADE1 entry. Pressure loads are input
as force per unit area.
Deformation-dependent pressure loads that act onto the edges of
plane stress elements do not take into account the change in
element thickness due to in-plane deformations. For example, a
plane stress element that undergoes uniaxial tension due to a
deformation-dependent pressure load has internal stresses larger
than the pressure load by the ratio (original thickness)/(current
thickness).
 Pressure loads can be applied to the faces of 3-D elements

(HEXA, CPENTA, CTETRA, CPYRAM) using the PLOAD or
PLOAD4 entries.
 When applied through the PLOAD4 entry, the pressure can be

normal to the face of the element, or along a specified direction.
 For each pressure/distributed load surface specified, a consistent

nodal load vector is calculated to represent the pressure/distributed
loading.
 In a large displacement analysis, the pressure/distributed

loading can be specified as deformation dependent for all element
types via the LOADOPT parameter in the NXSTRAT entry. In this
case, the calculations of the consistent load vectors are based on the
latest geometry and configuration of the loading surface.
 In Solution 601, deformation dependent loading should only be

used in a large displacement analysis. Equilibrium iterations (see
Chapter 6) should in general be performed if deformation
dependent loading is present.
 The loading direction for distributed loads can be along the

basic coordinate system or the element coordinate system. Loads
along element coordinate systems can be deformation dependent in
large displacement analysis.
 For pressure loading on 2-D and 3-D solid elements, the

consistent load vector consists of nodal forces acting on the
translational degrees of freedom only. The calculation of this load

vector is given in ref. KJB, Section 4.2.1. The same effect occurs in
shells since the nodal translations and rotations are interpolated
independently.
 The distributed loading on a beam element results in equivalent

concentrated forces and moments acting at the beam nodes as
shown in Fig. 5.3-2. The calculation of these consistent forces and
moments also follows the equations in ref. KJB Section 4.2.1.
q2
q1
1l l2
L
(a) Beam distributed loading
L L
F1 = (7q 1 + 3q 2) F2 = (7q 2 + 3q 1)
20 20
l l
L
L2 (3q + 2q ) L2 (3q
M1 = 1 2 M2 = 2 + 2q 1)
60 60
(b) Fixed-end forces/moments representation
Fig. 5.3-2: Representation of beam distributed loading
 Displacements and stresses in the model are calculated by

representing the actual distributed loading using the consistent load
vector defined above. Hence, the calculated solution corresponds
only to these equivalent concentrated nodal forces and moments,
and may not correspond entirely to beam theory results taking

5.3: Pressure and distributed loading
account of the distributed loading more accurately, see Fig. 5.3-3.

In order to capture the applied moment more accurately, more
beam elements are required between sections A and B.
A B
Bending
moment
MA
MB
a) Actual beam structure bending moment diagram
A B
l l l l
Bending
moment
MA
MB
b) Finite element bending moment diagram
using 3 beam elements
Fig. 5.3-3: Beam element bending moments

when subjected to distributed loading
5.4 Inertia loads ─ centrifugal and mass proportional

loading
 Centrifugal and mass proportional loading can be used to model
the effect of body forces which arise from accelerations to which
the structure is subjected.
 Centrifugal loading is generated using the RFORCE and

RFORCE1 entries, and mass proportional loading is generated

using the GRAV entry.
 The mass matrix used in the calculation of centrifugal and mass

proportional loading can be lumped or consistent depending on the
mass setting for the whole model. Note that the computational
effort involved in the evaluation of a lumped mass matrix is, in
general, much less than the effort for a consistent mass matrix.
 Centrifugal and mass proportional loading can both be present

in a static or dynamic analysis. In a dynamic analysis, the type of
mass matrix employed in the load calculation and in the dynamic
response calculation are the same.
 When more than one centrifugal load is applied to part of the

model, each load acts independently. This means that certain inertia
forces due to the coupling of the velocities (e.g. gyroscopic forces)
are not accounted for in the analysis. In certain problems, the
coupling terms are significant. Therefore, in general, it is
recommended to only apply one centrifugal load to any part of the
model.
 The angular velocities in centrifugal loading are not coupled

with the nodal velocities in the dynamic response. This means that
certain inertia forces due to the coupling of the velocities (e.g.
gyroscopic forces) are not accounted for in dynamic analysis. It is
recommended to only use centrifugal loading in static analysis.
 When elements die (due to rupture), their contribution to the

load vector is removed. Hence, the consistent load vector consists
(at all times) only of the contributions from the elements currently
alive.
 Centrifugal and mass proportional loading cannot be applied

with a delay/arrival time. The time function has to be shifted
manually to create this effect.
 Centrifugal or mass proportional forces at fixed nodes are taken

into account in the calculation of reaction forces.
 Centrifugal l loading is not supported for potential-based fluid

elements.

5.4: Inertia loads — centrifugal and mass proportional loading
 Centrifugal loads are only approximately computed for the large

displacement beam element. This is because the centrifugal loads
are calculated assuming that the beam is straight. The error in the
solution is reduced as the beam element mesh is refined.
Centrifugal loading
 The consistent load vector for centrifugal loading is computed

as follows (see Fig. 5.4-1):
t
R    [ α
t
 r

t
 t ω   t ω  t r  ] dV
  
(5.4-1)
angular acceleration centripetal acceleration
where t α    RACC f (t )  R is the angular acceleration vector,

t
 
ω   A f  t  R is the angular velocity vector, t r is the
radial vector from the axis of rotation to the node, A is the scale
factor of the angular velocity (in revolutions/unit time), RACC is
the scale factor of the angular acceleration (in revolutions / unit
time squared),  is the density, f(t) is the time function, and R (no
left superscript) is the rotation vector.
Note that the centrifugal force is directly proportional to the
time function f(t).
t
a
w = angular velocity vector

t t
w
a = angular acceleration vector
t
z
G
Centrifugal loads
y
x
Structure
Fig. 5.4-1: Convention used for centrifugal loading

Note that the consistent load vector consists of nodal point

forces applied to all the nodes of the finite element model. But
loads applied to fixed degrees of freedom do not have any effects in
the solution.
 When centrifugal loading is used in nonlinear analysis,

additional nonlinear terms are added to the stiffness matrix and the
load vector. These additional nonlinear contributions are described
in the following. Let the equilibrium of the finite element system be
calculated at time t + Δt, iteration i. Then
t t  (i )  t t K 
M t t U
i 1
U (i )  t t R ( i )  t t F 
i 1
(5.4-2)

where M = mass matrix, K = stiffness matrix = K L  K NL0 , U  
= acceleration vector, U = incremental displacement vector, R =
external load vector, F = nodal point forces corresponding to
stresses, (i) = iteration i, and t + Δt = time t + Δt.
The complete expression of the load contribution due to the
angular velocity part of the centrifugal loading, including all
nonlinear effects, is
t t
 
R ( i )     ω  ω   0r  t t U ( i 1)  U ( i )  dV 
(5.4-3)
0
where r = initial radial vector from the axis of rotation to the
node and ω = angular velocity vector. From the expression
(5.4-3), it can be seen that a deformation-dependent load is present,
given by
t t

R NL     ω   ω  t t U (i 1)  dV
i

and that an additional nonlinear contribution K NL1 to the stiffness
matrix is present, given by
t t

K NL1  U (i )    ω   ω  U (i )  dV
i 1

Both of these effects are included in the program.

The complete expression of the angular acceleration load

contribution, including all nonlinear effects, is
t t i
 
R       α  0 r  t t U 
i 1
 dV
In this case, a deformation-dependent load is present, given by
t t i

R NL     α  t t U 
i 1
 dV
However, Solution 601 does not include the nonlinear stiffness
matrix contribution from the angular acceleration load. The
stiffness matrix associated with this load is skew-symmetric due to
the single cross-product term.
 Nonlinear centrifugal loading can be used in static analysis and

implicit dynamic analysis. Nonlinear centrifugal loading cannot be
used in explicit dynamics (Solution 701).
 The correction to the stiffness matrix and the correction to the

loading are made when deformation dependent loading is requested
(LOADOPT parameter in the NXSTRAT entry).
 More than one centrifugal load can be specified using the

RFORCE1 entry.
 In the RFORCE1 entry, the GROUPID field can be used to

choose the elements onto which the centrifugal loads will be
applied. The GROUPID field selects a group of elements defined
using the GROUP bulk entry.
Mass-proportional loading
 The consistent load vector for mass proportional loading in

direction i is computed using the mass matrix of the entire finite
element system and the specified accelerations (only in the
translational degrees of freedom), as follows:
t
R i  t M di t ai (5.4-4)

where d i is a direction vector with "1" in the portions of the

translational degrees of freedom acting into the direction i and "0"
in the other portions, and ai is the acceleration magnitude in the
direction i.
In the calculation, the mass coupling term between active and
deleted degrees of freedom is included. This mass coupling term
can be clearly seen in the discussion of ground motion loads later in
this section.
 Mass proportional loading is commonly used to model gravity

loading. For gravity loading, t ai is the acceleration vector due to
gravity. For example, for the z coordinate in the vertical direction
(increasing z corresponds to movement away from the ground),
enter t az   g , where g is the (positive) acceleration due to
gravity.
 Mass-proportional loads are frequently used to model ground

motions. The basis for using mass-proportional loads in modeling
ground motions is given briefly now. The equations of motion for
linear dynamics, not including damping but including ground
motions, are
 M11 M12   
x1   K11 K12   x1   R1 
MT    T  (5.4-5)
 12 x 2  K12
M 22   K 22   x 2   R 2 
where x1 is the vector of nodal point displacements for nodes not

attached to the ground and x 2 is the vector of nodal point
displacements for nodes attached to the ground. R1 , R 2 are
externally applied forces (for example, concentrated forces).
When the ground motions are the same at all nodes attached to
the ground, x1  u1  d1i xgi , 
x1  u
1  d1i 
xgi , x 2  u 2  d 2i xgi ,
x2  u
  2  d 2i 
xgi , where u1 is the vector of nodal point
displacements relative to the ground for nodes not attached to the
ground and u 2 is the vector of nodal point displacements relative
to the ground for nodes attached to the ground. Clearly, u 2  0 .
Also d1i is the direction vector for the nodes not attached to the

ground and d 2i is the direction vector for the nodes attached to the
ground, with “1” in the portions of the translational degrees of
freedom acting into the direction i and “0” in the other portions.
The matrix equation of motion becomes
 M11 M12   u xgi    K11

1   d1i  K12   u1   d1i xgi    R1 
      T    
MT
 12

M 22    0  d 2i xgi    K12 K 22    0  d 2i xgi    R 2 
(5.4-6)
 K11 K12   d1i   0  d 
Now  T       since the vector di   1i 
 K12 K 22  d 2i   0  d 2 i 
corresponds to a rigid body motion. The matrix equation of motion
becomes
 M11 M12  u 1   K11 K12  u1   R1   M11 M12   d1i xgi 
MT  T    x 
 12 M 22   0  K12 K 22   0   R 2  M12
T
M 22  d 2i gi 
(5.4-7)
and therefore the system of equations solved is
1  K11u1  R1  M11d1i 

M11u xgi  M12d 2i 
xgi (5.4-8)
The mass coupling term between active and deleted degrees of

freedom ( M12d 2i 
xgi ) is included.
It is seen that the ground acceleration can be applied to the
model as a mass-proportional load, provided that the resulting
nodal point motions are interpreted as motions relative to the
ground.
Please note:
 To enter a positive ground acceleration xgi , specify a

negative mass-proportional load ai .
 All single-point fixities are relative to the ground. In other
words, fixing a node attaches it to the ground.
 All enforced displacements are relative to the ground.
 All single DOF scalar elements are attached to the ground.

Damping can be used. However, scalar dampers and single DOF

damping scalar elements are attached to the ground. Mass-
proportional Rayleigh damping acts relative to the ground motion.
Although we have illustrated the procedure only for linear
dynamics, the procedure is also valid in nonlinear dynamics.
5.5 Enforced motion

 Enforced displacements at specified degrees of freedom can be
applied in Advanced Nonlinear Solution using the single point
constraint entries (SPC, SPC1, or SPCADD) or the enforced
motion (SPCD) entry. The applied displacement can be constant or
described by a time function.
 Enforced velocities and accelerations are not supported.
 Nodal point translations and rotations can be enforced. The

degree of freedom is in the direction of the displacement coordinate
system assigned to the node.
 A nodal point can be "fixed" by prescribing a zero displacement

component for all degrees of freedom at this node. This is,
however, different from imposing a permanent single-point
constraint on the GRID entry because the enforced degrees of
freedom are retained in the system matrices (i.e., equation numbers
are assigned) whereas the degrees of freedom at which permanent
GRID constraints are imposed are deleted from the system
matrices.
 Note that enforced displacements are not recommended on

contactor surfaces (see Chapter 4).
 Arrival time option: Delay or arrival times can be used for

enforced displacements. In this case, the displacements are free
before the arrival time. Once the arrival time is reached, the
displacements are set to their enforced values. However, the
enforced value can be interpreted as an absolute or total
displacement or as a relative displacement based on the
configuration at the arrival time.

5.5: Enforced motion
 Relative enforced displacement option: Normally the specified

enforced displacement is directly applied to the model, so that the
enforced displacement at the node is equal to the specified enforced
displacement. However, there is a “relative enforced displacement
option” that works as follows. When the relative enforced
displacement option is active, then the enforced displacement at the
node is calculated using the specified enforced displacement plus a
preexisting displacement. The preexisting displacement is the
displacement at the arrival time (if the arrival time is nonzero), or
at the beginning of the current analysis (which might be a restart
analysis).
The time function for the specified enforced displacement can
be constant or time-dependent.
As a special case, the specified enforced displacement can be
zero. Then the enforced displacement is equal to the preexisting
displacement. This option is useful for “freezing” the motion of a
node, so that before the arrival time, the node has no enforced
displacement, and after the arrival time, the node has a enforced
displacement equal to the displacement at the arrival time.
The relative enforced displacement option is controlled by the
DISPOPT flag in the NXSTRAT entry. DISPOPT=0 selects
absolute enforced displacements (the default) and DISPOPT=1
selects relative enforced displacements.
 Enforced rotations can be applied to the rotational degrees of

freedom of large displacement formulation structural elements
(such as Hermitian beam elements), and also to the master
rotational degrees of freedom of large displacement rigid links. The
formula used is now discussed.
In large displacement analysis, the increment in enforced
rotation is computed as θ  t t θ  t θ , where t θ , t t θ are the
enforced rotations at times t , t  t . The components of the
increment in enforced rotation are θ  ( x ,  y ,  z ) .
Consider an incremental rotation with components
θ  ( x ,  y ,  z ) applied to the master node of a large
displacement rigid link with end coordinates t x m , t x s . ("m" =
master, "s" = slave). The vector connecting these coordinates is
t
x sm  t x s  t x m . When this incremental rotation is applied to the

master node of the rigid link, the vector t x sm changes to

  
 sin 
t t sin  1 2 S2
x sm  Q(θ) t x sm , where Q(θ)  I  S  
 2  
 2 
 0  z  y 
 
and S    z 0  x  ,    x2   y2   z2 . The
  y  x 0 

above formula can be found in ref KJB, p 580, exercise 6.56.
This formula gives the exact answer for the case when the
incremental rotations are finite. Physically, the incremental
rotations are interpreted as a vector, with the direction of the vector
interpreted as the axis of rotation, and the magnitude of the vector
interpreted as the amount of rotation about that axis.
The same formula is applied to the end-node triad of a beam
element local node, and to the director vector of a shell element
local node.
5.6 Applied temperatures

 In Advanced Nonlinear Solution, temperature can be prescribed
in any structural analysis. In addition, in heat transfer and TMC
analyses, temperature can be prescribed at certain parts of the
model, and the program will solve for the complete temperature
field.

 Temperature can be applied directly to a node using the TEMP
entry or to the whole model using the TEMPD entry. Direct nodal
values applied with TEMP override the default TEMPD value. This
is applicable to structural analysis, and to initial conditions for a
heat transfer analysis.
 The TEMPERATURE case control command selects the initial

or reference temperature field.
 In heat transfer analysis, temperature boundary conditions are

applied using the TEMPBC entry or the SPC entry.

5.6: Applied temperatures
 A time-dependent temperature load can be applied by using the

DTEMP bulk entry and referencing it in a DTEMP case control. In
previous versions, a time-dependent temperature load can be
defined with the TLOAD1 entry, and referencing the TLOAD1 in
the DLOAD case control. This is still supported for backwards
compatibility but it is recommended to use DTEMP instead. Note
that DTEMP cannot be used with TEMP(LOAD) or DLOAD for
temperature loading.
 It should also be noted that when using higher-order elements,

the temperatures can be significantly different within the element
than at the nodal points. For example, the temperature can be
negative at points within an element, although the nodal point
temperatures are all non-negative. This is illustrated in Fig. 5.6-1
(see element 1). This observation can be important when
performing an analysis with temperature-dependent material
properties.
 Applied temperature gradients cannot be specified on shell

element nodes.
 A node cannot be specified more than once in a temperature set.
 Temperature loads cannot be specified with both TEMP(LOAD)

and DLOAD case control commands. Exception: if TEMPOPT=1
in the NXSTRAT entry, then both TEMP(LOAD) and DLOAD are
allowed. This option is provided for backwards compatibility with
previous versions of NXN. However, it is recommended to specify
only TEMP(LOAD) for a constant temperature load or DTEMP for
a time-dependent temperature load in a model.

Linear elements:
q
q1
q0 l l l l l l l
x
El. 1 El. 2 El. 3 El. 4 El. 5 El. 6
Quadratic elements:
q1
q<q 0
q0 l l l l l l l
x
El. 1 El. 2 El. 3
Fig. 5.6-1: Interpolation of temperature boundary

conditions
5.7 Bolt preload

 Advanced Nonlinear Solution supports the preloading of bolt
elements. The bolt preloads or forces are applied during extra
solution steps performed at the very beginning of the analysis prior
to the rest of the step-by-step solution.
 The bolt preloads are applied via the BOLTFOR entry which
should be used together with a BOLTLD case control command.
 Bolt preload (and bolts in general) are only available in Solution

601.
 The BOLTSTP feature of the NXSTRAT entry allows the bolt

preload to be applied over BOLTSTP solution steps (default is

5.6: Applied temperatures
BOLTSTP=1.
 See Section 10.7 for more details on the bolt feature.
5.8 Constraint equations

 Advanced Nonlinear Solution supports single-point and
multipoint constraints. The single-point constraints are defined
using the SPC, SPC1, SPCADD entries. Permanent single-point
constraints can also be defined using the GRID entry. This case,
however, totally removes the associated degree of freedom from
the solution.
 Multipoint constraints are defined directly using the MPC and
MPCADD entries. They can also be defined through R-type
elements (see Section 2.7).
 The following relationship holds for multipoint constraints:
R u
j
j j 0

 Constraints can be enforced in two ways. The first approach
(called regular constraints approach) is for the first degree of
freedom in each constraint (u1) to be a dependent degree of
freedom. The second approach (called general constraints
approach) is to add a Lagrange multiplier to enforce each
constraint, and hence keep all constraint degrees of freedom
independent.
Parameter GENMPC in the NXSTRAT entry sets the approach
used for all constraint equations. GENMPC=0 selects regular
constrants (the default) and GENMPC=1 selects general
constraints.
 Note that in the regular constraints approach, each constraint

reduces the number of independent equations by one, while in the
general constraints approach, each constraint adds one extra degree
of freedom (the Lagrange multiplier). Hence, the regular
constraints approach should be used whenever possible. In some
cases, however, one cannot easily express a constraint in a way
such that dependent degrees of freedom are not constrained to other
dependent degrees of freedom.

 As a simple example, the constraints given by the equations

below cannot be expressed as regular constraints:
u z2  u1z  3 x1  0  valid regular constraint

u  u  u  4  0
3
z
2
z
1
z
2
x  invalid regular constraint

This DOF is already dependent
in which the subscript gives the direction and the superscript gives
the node number. These constraints can be expressed as general
constraints.
However, the second constraint can be written as a regular
constraint by manually applying the first constraint to obtain
u z3  2u1z  3 x1  4 x2  0
 Mesh glueing (Section 5.9) internally creates general constraint

equations that are enforced using Lagrange multipliers. All
independent degrees of freedom associated with the glued mesh
remain independent.
 Regular constraints can be applied both in Solution 601 (in
static and implicit dynamic analysis) and in Solution 701.
 General constraints cannot be used in Solution 701.


 Multipoint constraints are only approximately satisfied in an
explicit analysis (Solution 701), since imposing the constraint
exactly requires a non-diagonal mass matrix.

 For R-type elements in large deformation the multipoint
constraint can have variable coefficients that are updated based on
the deformation of the structure.
 Note that enforced displacements detailed in Section 5.5 are
internally enforced using single-point constraints.
 For an R-type element to produce multipoint constraints with
changing coefficients that capture large deformations, the
constraints must be between only 2 nodes. In addition, one of the
nodes should possess all the independent degrees of freedom and

5.8: Constraint equations
the other node should only possess dependent degrees of freedom.

These large displacement multipoint constraints are internally
called rigid links.
 The following comments apply to regular constraints of in

which the dependent degree of freedom is constrained to more than
one independent degree of freedom.
 If a consistent mass matrix is used and concentrated masses

are applied to constrained nodes, then the off-diagonal
terms are included in the global mass matrix.
 Similarly, if concentrated dampers are applied to
constrained nodes, then the off-diagonal terms are included
in the global damping matrix.
5.9 Mesh glueing

 The mesh glueing feature is used to attach two surfaces (or lines
in 2-D) together. These two surfaces usually involve different finite
element meshes (see Fig. 5.9-1). The glueing procedure results in a
smooth transition of displacements and tractions between the glued
surfaces. Mesh glueing sets are defined in the BGSET entry, and
the glued surfaces are defined via the BEDGE, BSURFS or
BCPROPS entries.
 This feature is useful for several applications:
 When a fine mesh is desired in a certain region and coarser

meshes are desired in other regions.
 When different regions are meshed independently with
unstructured free meshes.
 When different regions are meshed with different element
types (such as a tetrahedral mesh attached to a brick mesh).

Mesh glueing along dashed lines
Fig. 5.9-1: Examples requiring mesh glueing
 The proper glueing constraint between the two surfaces can be

expressed as
 
  (u1  u 2 )d   0 (5.9-1)
where u1 is the displacement of the first glued surface, u2 is the

displacement of the second surface and λ is the Lagrange multiplier
field imposing the constraint.
One of the glued surfaces is designated as the master and the
other as the slave. The Lagrange multiplier field involves nodal
degrees of freedom at the nodes of the slave surface, and the
integration is also performed over the slave surface. Hence Eq.
(5.9-1) becomes
 S
 S  (u M  u S ) d  S  0 (5.9-2)
The accurate integration of Eq. (5.9-2) is not trivial since the

displacements uM and uS are generally interpolated over different
domains. This integration is automatically performed by Advanced
Nonlinear Solution.

5.9: Mesh glueing
 Mesh glueing is not available in Solution 701.
 Glueing is superior to tied contact and should be used in its

place whenever applicable.
 Only 2-D solid elements can be used in the gluing of 2-D lines.
The glued element side can have linear or quadratic displacement
interpolation. Rod and beam elements are not supported.
 Only 3-D solid elements can be used in the gluing of 3-D

surfaces. The glued element faces can be triangles or quads, and
they can have linear or quadratic displacement interpolation. Shell
elements are not supported.
 Nodes on glued surfaces (both master or slave) cannot have

dependent translation degrees of freedom. Therefore, they cannot
be slaves in multi-point constraints involving translations.
 If one glue surface is smaller than the other, the smaller surface
should preferably be the slave. However, the glueing will also
work if the smaller surface is the master. The two glued surfaces
can also be partially overlapping. These cases are shown in Fig.
5.9-2.
 The two glued surfaces should ideally be smooth surfaces (no

sharp corners). If corners exist it is better to create multiple glued
meshes, as shown in Fig. 5.9-3.

Master surface
Slave surface
Slave surface
Master surface
Glue surface 2
Glue surface 1
Fig. 5.9-2: Examples of master and slave glue

surfaces
One glue-mesh Cross-over points
One glue-mesh
Four glue-meshes
(one for each side)
Fig. 5.9-3: Glueing non-smooth surfaces

5.9: Mesh glueing
 The Lagrange Multiplier field is modified at nodes where

multiple glue meshes intersect. These nodes are called cross-over
points.
 If the two surfaces have different mesh densities, either one can
be used as slave. However, it is recommended that the finer mesh
surface be the slave surface. Using the finer meshed surface as a
slave will produce more equations, since the Lagrange multipliers
degrees of freedom are on the nodes of the slave surface.
 The master glue surface can be enlarged beyond its geometric

bounds, so that the slave points that project slightly outside the
master can still be considered glued. This is done via the EXTi
parameter in the BGSET card.
 If the independent parts of the model have rigid body motions

(not considering the effect of glueing), the equation solver might
stop due to zero pivots. In this case, stiffness stabilization can be
employed (see Section 10.6).
 The mesh glue feature admits rigid body translations. Therefore,

when the model is rigidly translated, no spurious stresses and
strains are generated, and force equilibrium is satisfied between the
glued surfaces.
 The mesh glue feature only admits rigid body rotations when
the two glued surfaces exactly coincide, that is, when the two glued
surfaces are flat. Curved surfaces do not exactly coincide due to the
discretization error, as illustrated in Fig. 5.9-4.
Fig. 5.9-4: Glued surfaces do not exactly coincide when the

surfaces are curved.

As a result, if the glued surfaces are curved, moment

equilibrium is not satisfied between the glued surfaces, spurious
stresses and strains are generated when the model rotates, and
rotational motions are resisted, just as if the glued surfaces are
attached to a grounded rotational spring. The stiffer the glued
surfaces, the stiffer the rotational spring. This effect is less severe
when the coarser meshed surface is the slave surface of the glue
mesh pair because in this case the surfaces are not as “tightly”
glued (fewer glueing constraints are created), and hence the
distortional motion imposed by the mesh glue is less severe.
Fig. 5.9-5 shows the stress results for a large displacement static
analysis where a steel cylinder of radius 1mm is rotated by 10
degrees by prescribing the rotation on the outside surface of the
cylinder. Other than the prescribed rotation, the cylinder is
completely free. We observe that the exact solution is obtained
when there is no mesh glue and when the glued surfaces are flat,
but spurious stresses are generated when the glued surfaces are
curved, especially when the finer meshed surface is the slave
surface in the mesh glue pair.

5.9: Mesh glueing
Fig. 5.9-5: Stress results when cylinder is rotated 10 degrees.
5.10 Convection boundary condition

 Convection boundary conditions take the following form
q S  h  e   S 
where h is the convection coefficient,  e is the external ambient

temperature, and  S is the unknown body surface temperature.

 Convection boundary conditions are applied using surface

elements generated using the CHBDYE or CHBDYG entities
which point to a CONV entry. This entry provides some of the
necessary inputs and in turn points to a PCONV entry providing
more input. The heat convection coefficient h is provided in the
material definition entries.
 The following types of convection boundary conditions are

available:
Line convection boundary conditions, used in conjunction with

2-D planar elements and 2-D axisymmetric elements.
Surface convection boundary conditions, used in conjunction

with 3-D solid or shell elements.
 The convection coefficient h can be either temperature-

dependent (through a MATT4 or MATT5 entry), or time-
dependent. This is achieved via the Control Node setting in the
CONV entry. It cannot, however, be both temperature and time
dependent.
 The ambient temperature  e is obtained from node using

parameter TA1 in the CONV entry. The temperature at node TA1
must be prescribed, and can be time-varying.
 The heat flux, qS, is converted to nodal heat fluxes by consistent

integration over the convection boundary. See ref. KJB, Section
7.2.3 for details. In this integration, the temperatures are
interpolated from their nodal values, and if the heat transfer
coefficient, h, is temperature dependent, it is calculated for each
integration point based on its interpolated temperatures.
5.11 Radiation boundary condition

 Radiation boundary conditions take the following form

q S   f e  r4   S 
4


5.11: Radiation boundary condition
where  is the Stefan-Boltzmann constant, f is a view factor or

shape factor, e is the material emissivity,  r is the temperature of
the radiative source (or sink) and  S is the unknown body surface
temperature. Both temperatures are in the absolute scale. Note that
in the above equation the absorptivity is assumed to be equal to the
emissivity.
 Radiation boundary conditions are applied using surface

elements generated using the CHBYDE or CHBYDG entries which
point to a RADBC entry. This entry provides some of the necessary
inputs and in turn points to a RADM or RADMT entry providing
the rest of the inputs.
 The Stefan-Boltzmann constant () and the absolute

temperature offset are set in PARAM entries (SIGMA and TABS
parameters). Note that although  is a constant it must be input in
the proper units.
 The following types of radiation boundary conditions are

available:
Line radiation boundary conditions, used in conjunction with

2-D planar elements and 2-D axisymmetric elements.
Surface radiation boundary conditions, used in conjunction

with 3-D solid or shell elements.
 The emissivity coefficient e can also be temperature-dependent

by using the RADMT entry.
 The radiative source/sink temperature  r is specified in the

NODAMB parameter in the RADBC entry. The temperature at this
node NODAMB must be prescribed, and can be time-varying.
 The view or shape factor f is input via the FAMB parameter in

the RADBC entry.
 Default values of some radiation settings are defined using the

BDYOR entry.


integration over the radiation boundary. See ref. KJB, Section 7.2.3
for details. In this integration, the temperatures are interpolated
from their nodal values, and if the emissivity, e, is temperature
dependent, it is calculated for each integration point based on its
interpolated temperatures.
5.12 Boundary heat flux load

 Applied boundary heat flux is specified by equation (8.1-3):
ref. KJB
Section 7.2.1

kn  qS
n S2
where q S is the surface heat flux input to the body across some
part S2 of the body surface, kn is the body thermal conductivity in
direction n, the outward normal to the surface, and θ is the
temperature.
 Boundary heat flux loads are applied either directly to the nodes
defining a face of an element using the QHBDY entry, or by
pointing to existing surface elements (CHBDYi type) using the
QBDY1 or QBDY2 entries.

integration over the boundary. See ref. KJB, Section 7.2.3 for
details.
 Note that any boundary of the domain which does not have
either the heat flux or temperature specified will be assumed by
virtue of the formulation to have
qS  0
i.e., this part of the boundary is insulated, allowing no heat transfer

across it.

5.13: Internal heat generation
5.13 Internal heat generation

ref. KJB  This form of thermal loading results from the generation of heat
Section 7.2.3 within the domain, which is introduced into the governing equation
system by the term qB of equation (8.1-1).
 Internal heat generation is applied via the QVOL entry which

provides a load multiplier to the heat generation parameter HGEN
set in the MAT4 or MATT4 entries.
 A negative heat generation term qB indicates a loss of heat

within the body.
 The heat generation term can be temperature-dependent by

making the HGEN parameter temperature dependent using the
MATT4 entry.
 The heat flux generated per unit volume, qB, is converted to

nodal heat fluxes by consistent integration over the element
volume. See ref. KJB, Section 7.2.3 for details.

Chapter 6: Static and implicit dynamic analysis
6. Static and implicit dynamic analysis

This chapter presents the formulations and algorithms used to solve
static and dynamic problems using Solution 601. This includes
convergence checking and the available solvers. Most flags or
constants that need to be input in this chapter are in the NXSTRAT
bulk data entry.
Information about the progress of the solution is always output
to the .f06 file. A shorter summarized output is provided in the .log
file.
6.1 Linear static analysis

 In linear analysis using Solution 601, the finite element system
ref. KJB equilibrium equations to be solved are
Sections 8.2.1,
8.2.2 and 8.2.3 KU = R
 A direct sparse solver, iterative multigrid solver or 3D iterative

solver can be used to solve this system of equations, see Section
6.5.

 The equation solvers assume that the system stiffness matrix is
symmetric.
 The equation solvers assume that the system stiffness matrix is

positive definite. This requirement can be summarized as follows:
The Rayleigh quotient
T K 
  
T 
must be greater than zero for any displacement vector  . Since

    is equal to twice the strain energy stored in the system (for
T    ), this is equivalent to the requirement that the strain
energy stored in the finite element system when subjected to any
displacement vector  must be greater than zero.

6.1: Linear static analysis
 Hence, the finite element system must be properly supported, so

that the system cannot undergo any rigid body displacements or
rotations.
 It also follows that no part of the total finite element system

must represent a mechanism, see ref. KJB, Fig. 8.7, p. 704, for such
a case.
 Nodal point degrees of freedom for which there is no stiffness

must be restrained. A degree of freedom does not carry any
stiffness if all of the elements connected to the nodal point do not
carry stiffness into that degree of freedom. In this case the degree
of freedom must be restrained using boundary conditions.
Note that nodal degrees of freedom which are not connected to
any elements and are not used as dependent nodes in constraint
equations are automatically deleted by the program.
 The elements joining into a nodal point and contributing to the

stiffness of the nodal point degrees of freedom must all be defined
properly and in a physically correct manner. For example, if there
are program input errors that yield a zero or negative Young's
modulus or an incorrect nodal point numbering for an element, the
stiffness at a system degree of freedom may be zero or negative.
 More details on the solvers available in Solution 601 are

provided in Section 6.5.
6.2 Nonlinear static analysis

 In nonlinear static analysis the equilibrium equations to be
ref. KJB solved are:
Section 8.4
t t
R  t t F  0
where t+ΔtR is the vector of externally applied nodal loads at time

(load) step t+Δt, and t+ΔtF is the force vector equivalent (in the
virtual work sense) to the element stresses at time t+Δt.
 The nonlinearity may come from the material properties, the

kinematic assumptions, deformation dependent loading, the
presence of contact, or the use of special features such as the

Advanced Nonlinear TMG 201 400

Uploaded by

Copyright:

Available Formats

Advanced Nonlinear TMG 201 400

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advanced Nonlinear TMG 201 400

Uploaded by

Copyright:

Available Formats

Chapter 3: Material models and formulations

Shaded area is area

Fig. 3.4-13: Evolution of strain memory surface in

Isotropic hardening rules: The isotropic hardening rules are

where 0 y and EP are material constants, and t t

192 Advanced Nonlinear Solution  Theory and Modeling Guide

Pairs of (e P ,  y ) are given. It is allowed for  y to either increase

Advanced Nonlinear Solution  Theory and Modeling Guide 193

where Q and b are material constants. Q can be positive to

Memory-exponential hardening (exponential hardening with strain

In this model, the yield surface size  y depends on the strain

memory size t t q (see above for a description of the strain

where t t q is the size of the strain memory surface.

194 Advanced Nonlinear Solution  Theory and Modeling Guide

Kinematic hardening rule: The kinematic hardening rule is:

Armstrong-Fredrick nonlinear kinematic hardening:

The back stress is expressed as a sum of partial back stresses

where t t α ( m ) is partial back stress number m. All of the partial

where h ( m ) and  ( m ) are material constants (these material

It is allowed to use one partial back stress, with   0 . Then

Stress  plastic strain curve for uniaxial cycling

In uniaxial cycling, a typical stress – plastic strain curve is

Advanced Nonlinear Solution  Theory and Modeling Guide 195

EP  h      y  on segment A-B. The size of the plastic

Fig. 3.4-15: Stable uniaxial plastic cycle using

Stress  plastic strain curve for initial loading

During initial loading, it can be shown that

The corresponding stress  plastic strain curve is shown in Fig.

196 Advanced Nonlinear Solution  Theory and Modeling Guide

isotropic hardening, and that the yield stress is  y . Hence the

Fig 3.4-16: Stress - plastic strain curve during initial loading,

Combination of isotropic and kinematic hardening rules: By

Stress-strain integration: When plasticity is detected, the

Constitutive tensor: The constitutive tensor (stress-strain matrix)

Formulations: When used with the small displacement/small

Advanced Nonlinear Solution  Theory and Modeling Guide 197

If geometrically nonlinear effects are to be included, the large

Mixed displacement-pressure formulation: Plane strain,

Thermal strains: If a thermal load is applied to the structure, the

Modeling of rupture: Rupture conditions can also be included

3.4.2.2 Specification of input

Parameters ISO, KIN, RUP in this entry reference data defined in

198 Advanced Nonlinear Solution  Theory and Modeling Guide

incremental plastic strains. RTOL is a tolerance used to assess

Parameter TYPE=BLIN, MTLIN, EXP, MEMEXP specifies the

In general, given a plastic-bilinear material with given yield stress

In general, given a plastic-multilinear material with Young's

this material can be converted into an equivalent plastic-cyclic

Advanced Nonlinear Solution  Theory and Modeling Guide 199

using the formulas

aeps(i)  strain(i)  (stress(i) / E) , sr(i)  stress(i)

Note that aeps at the elastic limit will always be zero.

3.5 Temperature-dependent elastic material models

 The thermal orthotropic model is available for the 2-D solid,

 Both models can be used with small displacement/small strain

200 Advanced Nonlinear Solution  Theory and Modeling Guide

Fig. 3.5-1: Variation of material properties for thermo-elastic model

 The calculation of thermal strains is described in Section 3.1.6.

 Note that if the material constants are all temperature