ABAQUS NumericalMethods PDF

Numerical Methods
Lecture 3
L3.2
Overview
• Contact Formulation Aspects

• Contact Discretization
• Contact Enforcement Methods
• Contact Tracking
• Incrementation and Newton Iterations
• Summary
Solving Contact Problems with New Capabilities in Abaqus

Contact Formulation Aspects
L3.4
Contact Formulation Aspects
• Contact discretization
• Where is the constraint applied?
• Node-to-surface
• Surface-to-surface
• Constraint enforcement
Contact formulation
• How is the constraint enforced?
• Direct (Lagrange multipliers)
• Penalty method
• Contact tracking (relative sliding)
• How does the constraint evolve?
• Finite sliding
• Small sliding

Contact Discretization
L3.6
• Node-to-surface technique:
• Nodes on one surface (the slave surface) contact the segments on the other
surface (the master surface)
• Contact is enforced at discrete points (slave nodes)
• Surface-to-surface technique
• Contact is enforced in an average sense over a region surrounding each slave
node
• Slave surface is much more than just a collection of nodes

L3.7
• Node-to-surface (N-to-S) contact discretization

• Traditional “point-against-surface” method
• Contact is enforced between a node and surface facets local to the node
• The node is referred to as a “slave” node; the opposing surface is
called the “master” surface
slave
master
These nodes are

free to penetrate

L3.8
• Key implications of the node-to-surface

formulation Master
• Slave nodes cannot penetrate master
surface segments
• Nodes on the master surface can Slave
penetrate slave surface segments
• A gross penetration of a master surface
into a coarsely meshed slave surface
causes solution inaccuracy
• Refinement of the slave surface
improves contact resolution
• Guidelines for master and slave roles
Slave
• The more-refined surface should act as
the slave surface
• The stiffer body should be the master Master
• The active contact region should change

most rapidly on the master surface
(minimizes contact status changes )
L3.9
• While refinement of the slave surface leads to global accuracy, local

contact stress oscillations may still be observed with N-to-S
Uniform pressure load, s =100
• 13% noise in CPRESS solution with N-to-S contact

discretization if top block acts as slave (shown above)
• 31% CPRESS noise if bottom block acts as slave (not shown)
• 2D example
Ideal contact force distribution factors Factors on master nodes assuming
(uniform pressure, linear elements): ideal factors on slave nodes:
1/6 1/3 1/3 1/6 1/6 1/3 1/3 1/6 Slave
1/4 1/2 1/4 Master
1/6×1 + ⅓×⅓ 2×⅓×⅔ 5/18
= 5/18 = 8/18
• A “matched mesh” across the contact interface avoids this noise
L3.10
• Surface-to-surface (S-to-S) contact discretization

• Each contact constraint is formulated based on an integral over the
region surrounding a slave node slave
Tends to involve more master

nodes per constraint, especially if master
the master surface is more
refined than the slave surface
• Still best to have the more-refined surface act as slave

• Better performance and accuracy
• Benefits of the surface-to-surface approach
• Reduced likelihood of large localized penetrations
• Reduced sensitivity of results to master and slave roles
• More accurate contact stresses
• Inherent smoothing (better convergence)
L3.11
• S-to-S discretization often improves

the accuracy of contact stresses
• Related to better distribution of
contact forces among master nodes
• Example: classical Hertz contact
problem:
• The contact pressure contours
are much smoother and the peak
Analytical CPRESSmax = 3.01e+05
contact stress is in very close
agreement with the analytical CPRESSmax = CPRESSmax =
solution using the surface-to- 3.425e+05 3.008e+05
surface approach
Node-to-surface Surface-to-surface

L3.12
• S-to-S discretization reduces likelihood of snagging

Node-to-surface Surface-to-surface
slave master
slave master
Treating the slave surface as a Computing average penetrations and slips

collection of points can trigger snagging over finite regions has a smoothing effect
as slave nodes traverse a corner that avoids snagging

L3.13
• S-to-S discretization reduces the likelihood of master nodes

penetrating a slave surface
master surface
constrained region
slave surface Some penetration

may be observed
at individual nodes;
Surface-to-surface however, large,
Node-to-surface results undetected
results penetrations of
master nodes into
the slave surface
do not occur

L3.14
• S-to-S discretization is much less sensitive to the choice of master and

slave surfaces when the mesh densities are similar
Slave and master surfaces switched

Choosing the slave surface to be the finer mesh will still yield better results; choosing
the master surface to be the more refined surface will tend to increase analysis cost

L3.15
• S-to-S discretization will generate multiple constraints at corners when

appropriate
Node-to-surface
• Single constraint in “average”
normal direction at corner
• Not stable
• Leads to large penetrations
and snagging
• Workaround: 2 contact pairs
Surface-to-surface
• Two constraints are generated at
corner (even if 1 contact pair)
• See arrows near corner
• Accurate and stable
• No smoothing of surface normals

L3.16
• S-to-S discretization takes into consideration shell and membrane

thicknesses when performing contact calculations
• N-to-S only considers this effect for the small sliding formulation
Thickness ignored
Thickness taken
into account

L3.17
• S-to-S discretization is fundamentally sound for situations in which

quadratic elements underlie the slave surface
• N-to-S struggles with some quadratic element types
q
• Related to (1) discrete treatment of slave q q
surface and (2) consistent force distribution
for element type q
1
pA
3
• Workarounds (with pros and cons)
Zero force at
• C3D10M, supplementary constraints, etc. corner nodes
Uniaxial pressure loading of 5.0
Slave:
C3D10
Master:
C3D8 Node-to-surface Surface-to-surface

L3.18
• S-to-S discretization has a greater tendency to generate unsymmetric

stiffness terms where the master and slave surface are not
approximately parallel to each other
• Use of the unsymmetric solver is sometimes necessary to avoid
convergence difficulties
*STEP, UNSYMM=YES

L3.19
• S-to-S discretization works best when the contacting surfaces have

nearly opposing normals
• Works well for many cases involving corners

L3.20
• Surface-to-surface discretization, however, has difficulty resolving

point-to-surface contact
Slave
Point-to-surface Surface-to-surface formulation:
contact
• Penetrations are averaged over
finite regions
Master
• Contact normal is not necessarily
equal to master surface normal
Slave
Surface-to-surface
contact
Master

L3.21
• Modeling suggestion for point-to-surface contact in Abaqus/Standard:

• Supplement general contact or surface-to-surface contact pairs with
node-to-surface contact pairs involving significant feature edges
Contact surfaces
Holder
Clip
leadingEdge
*Contact Pair, type=SURFACE TO SURFACE
Clip, Holder
*Contact Pair, type=NODE TO SURFACE
leadingEdge, Clip

L3.22
• Sequence of deformed plots

Contact Constraint Enforcement
L3.24
Constraint Enforcement
• Node-to-surface
Contact formulation
• Penalty method
• Finite sliding
• Small sliding

L3.25
• Strict enforcement
• Intuitively desirable (no approximation)
• Can be achieved with Lagrange multiplier method in Abaqus/Standard
• Drawbacks
• Can make it challenging for Newton iterations to converge
• Any overlapping constraints are problematic for equation solver
• Lagrange multipliers add to solver cost
Physically “hard” pressure vs. penetration behavior
p, contact pressure
Any pressure
possible when in
-h contact
No pressure
h<0 h=0
h, penetration
No penetration; Constraint enforced;
no constraint required positive contact pressure

L3.26
• Direct enforcement
• Lagrange multiplier method
• Constraint equations and Lagrange multipliers are added to the system
of equations
Unconstrained system of equations
K u = f
Constraint equations added
K CT u f
=
C 0 l 0
Vector of Lagrange multiplier degrees of

freedom (constraint forces)
• One per constraint
L3.27
• Penalty method
• The penalty method is a stiff approximation of hard contact
p, contact pressure p, contact pressure
Any pressure
possible when in
contact  k, penalty stiffness
No pressure No pressure
h, penetration h, penetration
Strictly-enforced hard contact Penalty method approximation of hard contact
K+Kp u = f

L3.28
• Pros and cons of penalty method

• Advantages:
• Significantly improved convergence rates
• Better equation solver performance
• No Lagrange multiplier degree of freedom unless contact
stiffness is very high
• Good treatment of overlapping constraints
• Disadvantages:
• Small amount of penetration
• Typically insignificant
• May need to adjust the penalty stiffness relative to the default
setting in some cases

L3.29
• Default penalty stiffness

• Abaqus tries to find a “happy medium” between
• Penalty stiffness too low:
• Excessive penetrations
• Penalty stiffness too high in Abaqus/Standard:
• Convergence rates degrade
• Lagrange multiplier degrees of freedom are needed to avoid ill-
conditioning
• Penalty stiffness too high in Abaqus/Explicit:
• Significant reduction in stable time increment
• The default penalty stiffness is based on a representative stiffness of the
underlying elements
• A scale factor is applied to this representative stiffness to set the
default penalty stiffness; its magnitude is higher in Abaqus/Standard
than in Abaqus/Explicit
L3.30
• For cases in which the default penalty stiffness is not suitable,

options to scale the penalty stiffness are available
• Order-of-magnitude changes are recommended
• If the scale factor > 100, Abaqus will automatically invoke a
variant of the method that uses Lagrange multipliers to avoid
ill-conditioning issues
Keyword interface
*SURFACE INTERACTION
*SURFACE BEHAVIOR, PENALTY
penalty stiffness, clearance offset, scale factor (all optional)
:
*STEP
:
*CONTACT CONTROLS, STIFFNESS SCALE FACTOR=value
Step dependent Multiplicative!

(careful!)

L3.31
• Penalty stiffness magnitude

• Stiff or blocky problems
• The default penalty stiffness generally produces results that are
comparable in accuracy with those obtained with the direct method
• Usually requires less memory and CPU time
• Bending-dominated problems
• The default penalty stiffness can often be scaled back by two orders
of magnitude without any significant loss of accuracy
• Scaling back the penalty stiffness for bending-dominated problems
sometimes increases the convergence rate

L3.32
• Example
Constraint Maximum Max. Mises Solver

# Iters.
method penetration stress FLOPs
0.4% of
Default
collar elem. 6.166E4 50 2.8E10
penalty
dimension
Lagrange
0 6.173E4 57 3.6E10
multiplier

Relative Sliding between Bodies
L3.34
• Node-to-surface
Contact formulation
• Penalty method
• Finite sliding
• Small-sliding approximation

L3.35
• Abaqus offers finite- and small-sliding

versions of S-to-S and N-to-S contact 201 206
formulations 202
205
203 204
• Finite-sliding formulation: General 101 102 103
104 105 106
applicability
• Point of interaction on master
surface is updated using the
true representation of the
Master surface
master surface
• Small-sliding formulation:
Approximation intended to reduce
102
solution cost; limited applicability Possible path of
slave node 102
• Planar representation of master
surface per slave node based
on the initial configuration
• Only available for contact pairs
102
(and not self contact or GC) Master “slide plane”
for slave node 102
L3.36
Small-Sliding Approximation
• Every slave node interacts with its own local slide plane
• In 2D/axisymmetric it is depicted as a line
• Assumes that the relative motion per slave node remains
small compared to the local curvature and facet sizes of
the master surface
• Advantage: Less nonlinearity

• Potential for reduced cost per iteration and finding a
converged solution in fewer iterations
• Disadvantage: The results can be nonphysical if the

relative tangential motion does not remain small
• It is the user’s responsibility to ensure that the
assumption is not violated

L3.37
Small Sliding Approximation
• Example of nonphysical behavior

with a small-sliding formulation
• The rectangular body acts as the
slave
• The slide planes are represented by
white lines in the animation
• The slide planes translate with the
punch as it moves to the right
• Confusing, nonphysical contact
behavior occurs due to the linear
approximation of the master surface
per slave node
• Lesson:
• Use the finite-sliding formulation
if you do not want to worry about
whether small-sliding assumptions
are appropriate!
L3.38
Small Sliding Approximation
• Invoking small-sliding (contact pairs only):
*CONTACT PAIR, SMALL SLIDING

L3.39
Formulation Summary
• Good formulation characteristics (for accuracy, robustness, and generality)

• Accurate representation of surface geometry
• Slave surface: not just a collection of points S-to-S
• Master surface: not approximated as flat per slave node finite-sliding
• Geometric corrections: reduce discretization error corr. for S-to-S
• Distribution of nodal forces consistent with underlying element formulation
• Ability to satisfy “patch tests” for contact S-to-S
• Continuity in contact forces upon sliding S-to-S
• Individual constraint stresses should oppose penetration (sliding)
S-to-S
• Nontrivial aspect for some quadratic element types
• Avoid “over-constraints” and “under-constraints” master-slave roles
• Generally, number of contact constraints in an active contact region
should equal number of nodes of the more-refined surface in that region
• Small amount of numerical “softening” penalty method
• Robust contact search algorithm to avoid missing contacts, etc. finite-sliding
L3.40
Formulation Summary
• Available formulations for general contact and contact pairs
Modeling Approach
Formulation Aspect General Contact Contact Pairs
Default: Node-to-surface (N-to-S)
Contact Discretization Surface-to-surface
Optional: Surface-to-surface (S-to-S)
N-to-S default: Direct enforcement
Contact Enforcement Penalty enforcement
S-to-S default: Penalty enforcement
Default: Finite sliding
Relative Sliding Finite sliding
Optional: Small sliding
Refers to defaults for keyword input file

• These defaults were established prior to implementation of
surface-to-surface discretization and penalty methods
• These are not the defaults for contact pairs created in
Abaqus/CAE based on initial proximity

L3.41
Formulation Summary
• Common issues when converting contact pair models to general contact

• Problems due to initial overclosures
• Shell or membrane thickness is accounted for by default
• Finite-sliding, node-to-surface contact pairs do not account for
thickness
• Initial overclosures are resolved with strain-free adjustments by
default
• Treated as interference fits by default with contact pairs
• All exposed surfaces are typically considered
• In some models contact pairs are not defined on penetrated regions

Incrementation and Newton Iterations
(Abaqus/Standard)
L3.43
• Newton method: iterative method used to solve nonlinear problems

Given:
• Starting displacement, u0
• Desired load, “P”
Remaining
• Ability to evaluate f(u) and K(u)
“residual
I0 = f(u0) Find:
force”
• Displacement solution, us, such
that f(us) = P
Iteration 1
• System of eqs.
K0 Du = P – I0 magnified
• Du = ca (see fig.)
• New estimate
ua = u0 + ca Iteration 2
• System of eqs.
Ka Du = P – Ia
• Du = cb (see fig.)
• New estimate
ub = u a + c b
L3.44
• The Newton method, however, is not guaranteed to converge

• Example in which Newton iterations diverge:
Load Goal: find this point
P
Diverging!
Applied load
Displacement
Starting point
Load applied in 1 increment
• Increase the likelihood of convergence by decreasing load increment

• Use multiple load increments to achieve desired total load
Load Load
P
P1
Displacement Displacement
Half load in 1st increment Remaining load in 2nd incr.

L3.45
• Abaqus automatically adjusts the load increment size

• Goal: Find converged solution robustly and efficiently with respect to the
number of iterations
• Basic idea: Track convergence rate in order to determine when to
increase or decrease load increment size
• User suggests increment size; Abaqus tries to optimize it
Slow convergence Reduce

or divergence increment size
Convergence in few Increase

iterations increment size

L3.46
• Contact causes kinks in the load vs. displacement curve

• There is a slope discontinuity upon change in contact status
• As a result, contact changes interrupt overall convergence rate tracking
Undeformed shape
3. Compress tip
2. Contact
rigid surface
1. Bend beam
P
Deformed shape
(Mises stress contours)
Challenging for
Newton method!

L3.47
• “Severe discontinuity iterations” (SDIs)

• An SDI is an iteration during which contact constraints change state
• Open/closed, stick/slip
• Idea is to filter out contact effects from convergence rate tracking
• The logic to adjust the increment size treats SDIs separately
Status (.sta) file Separate iteration
for the previous counts for SDIs and
example (beam non-SDIs
contact) 1st attempt did not
convergereduce Dt
2nd attempt at first

incr. converges
Increase Dt due to
fast convergence
DP  (Dt/T) Pfinal Converged incr. with

contact activated
Total step
time=1.0 Trend toward larger
Dt after contact is
established
L3.48
• “Hard” contact pressure vs. overclosure:

Contact
pressure Compliant
Non-compliant
“Hard” contact
Gap distance Penetration

distance
Penetration for
“open” contacts
Tensile stress for
“closed” contacts
• Default behavior: SDIs do not block convergence

• “Convert SDI”: small penetrations/tensile stresses trigger contact status
changes (and SDIs) but do not necessarily block convergence
• Without “Convert SDI”
• Contact status changes block convergence

L3.49
• Schematic of behavior within an increment (default behavior)
Begin
increment
Identify initially Form and Identify changes

1 active contact 2 solve system 3 in contact
constraints of equations constraint status
Newton
Yes iterations
Determine if Check if
End
5 tending toward 4 solution has
increment
No convergence No converged Yes
(Reduce increment (At least one (Within

size and try again) convergence criterion convergence
is not satisfied) tolerances)

L3.50
• Default contact algorithm (refer to flowchart on previous page):

1 Determine the initial contact state at each point (closed or open)
• For first increment of a step, based on initial model state; otherwise,
based on solution extrapolation (if any)
2 Form the system of equations with contact constraints imposed, then

pass through the equation solver
3 Are contact pressures and clearances consistent with the assumed

contact state?
• Contact status changes (open/closed or stick/slip) often cause
significant changes to the system of equations
• Iterations with contact status changes are flagged as severe
discontinuity iterations (SDIs)

L3.51
4 Has convergence been achieved?

• By default, Abaqus quantifies the contact incompatibilities
associated with SDIs; these incompatibilities must be sufficiently
small to achieve convergence with respect to the contact state
• Also have to ensure that the force residuals and solution corrections
are sufficiently small to achieve equilibrium
• If the contact state and equilibrium conditions satisfy their respective
convergence criteria, the increment is complete
5 If convergence is not achieved, is it likely to be achieved?
• Abaqus evaluates trends, such as the number of contact status
changes in successive iterations, to determine whether or not to
continue iterating or cut back the increment size
• If convergence is likely, update the contact constraints based on 3
and the stiffness, and re-solve the system of equations; otherwise,
try again with a smaller increment size

Lecture 3 Summary
L3.53
Review of Topics Discussed in this Lecture
• Title: Numerical methods

• Contact Formulation Aspects
• Contact Discretization
• Contact Enforcement Methods
• Contact Tracking
• Incrementation and Newton Iterations

ABAQUS NumericalMethods PDF

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

ABAQUS NumericalMethods PDF

Caricato da

Copyright:

Formati disponibili

Numerical Methods

• Contact Formulation Aspects

Solving Contact Problems with New Capabilities in Abaqus

Contact Formulation Aspects

Solving Contact Problems with New Capabilities in Abaqus

Solving Contact Problems with New Capabilities in Abaqus

• Node-to-surface (N-to-S) contact discretization

These nodes are

Solving Contact Problems with New Capabilities in Abaqus

• Key implications of the node-to-surface

• The active contact region should change

• While refinement of the slave surface leads to global accuracy, local

• 13% noise in CPRESS solution with N-to-S contact

• Surface-to-surface (S-to-S) contact discretization

Tends to involve more master

• Still best to have the more-refined surface act as slave

• S-to-S discretization often improves

Solving Contact Problems with New Capabilities in Abaqus

• S-to-S discretization reduces likelihood of snagging

Treating the slave surface as a Computing average penetrations and slips

Solving Contact Problems with New Capabilities in Abaqus

• S-to-S discretization reduces the likelihood of master nodes

slave surface Some penetration

Solving Contact Problems with New Capabilities in Abaqus

• S-to-S discretization is much less sensitive to the choice of master and

Slave and master surfaces switched

Solving Contact Problems with New Capabilities in Abaqus

• S-to-S discretization will generate multiple constraints at corners when

Solving Contact Problems with New Capabilities in Abaqus

• S-to-S discretization takes into consideration shell and membrane

Solving Contact Problems with New Capabilities in Abaqus

• S-to-S discretization is fundamentally sound for situations in which

Solving Contact Problems with New Capabilities in Abaqus

• S-to-S discretization has a greater tendency to generate unsymmetric

Solving Contact Problems with New Capabilities in Abaqus

• S-to-S discretization works best when the contacting surfaces have

Solving Contact Problems with New Capabilities in Abaqus

• Surface-to-surface discretization, however, has difficulty resolving

Solving Contact Problems with New Capabilities in Abaqus

• Modeling suggestion for point-to-surface contact in Abaqus/Standard:

Solving Contact Problems with New Capabilities in Abaqus

• Sequence of deformed plots

Solving Contact Problems with New Capabilities in Abaqus

Solving Contact Problems with New Capabilities in Abaqus

Solving Contact Problems with New Capabilities in Abaqus

Constraint equations added

Vector of Lagrange multiplier degrees of

p, contact pressure p, contact pressure

Solving Contact Problems with New Capabilities in Abaqus

• Pros and cons of penalty method

Solving Contact Problems with New Capabilities in Abaqus

• Default penalty stiffness

• For cases in which the default penalty stiffness is not suitable,

Step dependent Multiplicative!

Solving Contact Problems with New Capabilities in Abaqus

• Penalty stiffness magnitude

Solving Contact Problems with New Capabilities in Abaqus

Constraint Maximum Max. Mises Solver

Solving Contact Problems with New Capabilities in Abaqus

Relative Sliding between Bodies

Solving Contact Problems with New Capabilities in Abaqus

Relative Sliding between Bodies

• Abaqus offers finite- and small-sliding