ANSYS-Nonlinear Stabilization User Meeting 2019nov20

Twin Cities
ANSYS® User Meeting

November 2019
Stabilization Damping for

Nonlinear Convergence
Agenda
1. Epsilon FEA Introduction

2. Stabilization Overview
3. Stabilization Procedure
4. Stabilization Case Studies
5. Q&A
2
ANSYS User Meeting
Intro to Epsilon
• Epsilon FEA provides engineering analysis (10 yrs!)

• Making Simulation Accurate
– In-depth knowledge of the tools
• ANSYS® Suite of Multi-Physics software
– Experience with industry successes/failures
• Aerospace, Rotating Machinery, Electronics, Manufacturing, Packaging, etc.
– We validate with calibration runs and hand-calcs
• Experienced Assessing Discretization Error
• Making Simulation Affordable

– Low hourly rates and/or fixed-price estimates
– We use specialized experienced engineers
– Detailed statements of work, scope and budget tracking
– Automation (APDL, ACT, Journaling)
3
ANSYS User Meeting
Epsilon’s Customers
• Our customers need load-leveling with:

– Analyst is a team-member, not a black-box
• Interface with same Epsilon analyst to leverage past experiences
– Open and frequent communication
– Any new FEA methods/lessons learned are well communicated
– Schedule/budget fidelity with frequent status updates
• Achieved by using the right person, tools, and technical approach
• Our customers benefit from external expertise

– We infuse up-to-date FEA methods/tools
• Leverage other industries’ FEA innovations
– We share our knowledge, files, and lessons learned!
– We help with tool selection, infrastructure advice
4
ANSYS User Meeting
Stabilization Damping
• Aids solver in converging rigid body motions

– Force imbalance occurs resulting in high/infinite deflection
– Still in static domain (time integration is off!)
– Caused by pivoting, buckling, contact changes, etc.
• Stabilization is useful for analyses with stable beginning and
end states but periods of instability
• STABILIZE command in APDL
– Exposed in Workbench
5
ANSYS User Meeting
Stabilization Vs.
Arc Length Method
• Stabilization is an alternative to Arc Length Method, allows

for simulating instability with Newton-Raphson Method
– See previous user meeting documentation “Nonlinear Convergence”
from November 2010 on our website
– See PADT’s The Focus issue 14 from 2002
– See Unstable Structures ANSYS documentation page
6
ANSYS User Meeting
Stabilization Damping Features
1. Adds numerical viscous damping to affected nodes

– Internal to the solver
2. Damps “pseudo-velocity” of motion without requiring time step
reduction to characterize highly nonlinear activity
3. Allows force-based loads to be used in analyses that would require
displacement-based
4. Can be applied globally (all nodes) or to individual contact regions
– Aids in detecting abrupt changes in contact
– Can also be applied locally by reverting non-stabilized regions to legacy
elements
5. Can be turned on/off between load steps or with restarts
6. Reduces number of iterations by allowing larger time steps
7. Does not preclude the use of any other solver controls/contacts
– Arc Length Method does not support nonlinear contact
7
ANSYS User Meeting
Stabilization Damping
Limitations
1. Cannot simulate negative slope region of
load-displacement response curve
– Snap-through regions, etc.
– Requires global stability in end-state for results to
be viable
2. Possible to overdamp analyses with overly
large time stepping or damping ratios
– Can force convergence to a wildly inaccurate result
3. Damping dissipates energy from the model
– Reduces accuracy, especially for nonlinear
materials
4. Helps with high-strain element distortion
errors, but not ones caused by other (linear)
contacts in the model
8
ANSYS User Meeting
Global Stabilization Damping
Procedure
• Enable global stabilization in the Analysis
Settings
• Reduce or Constant application
– Reduce will start at prescribed stabilization
value and reduce linearly to zero by the end of
solution
– Constant applies stabilization through entire
solution
9
ANSYS User Meeting
Procedure
• Choose Energy or Damping Method
– Energy method sets amount of energy allowed
to be dissipated by damping
– Ratio must be tuned based on load magnitude
and time step duration
We will vary these inputs
– Damping factor is calculated from energy in the case studies
dissipation ratio and average element size
– Alternatively, manually set damping factor
10
ANSYS User Meeting
Procedure
• Set Damping Factor/Dissipation Ratio
– Default energy dissipation ratio of 1E-4
generally useful
– Damping factor has no default value due
to being model-specific, caution when
using
• First substep activation
– Only required for models beginning in an
unstable state, avoid if possible
• Can cause severe overdamping if not properly
tuned
• Set Force Limit
– Checks ratio of stabilization forces to
internal forces
– Does not have any affect on
convergence/bisecting or solving, simply
gives warnings when exceeded
11
ANSYS User Meeting
Contact Stabilization Damping
Procedure
• Stabilization can be
applied at nonlinear
contacts only
• Useful for analyses with
abrupt contact changes
but general global stability
• Set damping factor within
individual contacts
• No energy option, must
We will compare contact
calculate your own damping to global
damping factor damping in the case
studies
12
ANSYS User Meeting
Stabilization Damping Results
• Stabilization Energy Result

can plot Energy dissipation
per element
• Allows checking for excessive
damped energy as well as
identifying damped areas
• Not compatible with contact-
only damping
13
ANSYS User Meeting
Case Study 1: Snap-Through
Skipping Analysis
• Analysis begins in stable state, then as force
overcomes frictional contact at latch,
experiences a short period of instability
before coming to a new stable state resting
on fixed block
• No actual buckling “snap through” is
occurring, but similar style force-
displacement curve
• Representative of most model instabilities,
such as due to buckling, material failure,
abrupt contact changes, etc.
• Stabilization can be turned off for first
substep
• Stabilization will help skip over the region
where the main member becomes unloaded
and pseudo-velocity becomes very high, as a
result of a very high force controlled loading
14
ANSYS User Meeting
Case Study 1: Snap-Through
Skipping Analysis
• For baseline comparison, re-solved

– Since the end-state is known, we can
solve for this directly by skipping the
first phase
• This will allow us to determine the
effect of damping energy loss on
results to a known control result
• Instabilities are localized to contact
areas, so contact-only damping can
also be used
15
ANSYS User Meeting
Case Study 1: Damping Effects
on Accuracy
• Example of a
“missed” second
contact detection
16
ANSYS User Meeting
on Accuracy
• Example of a
“missed” initial
contact detection
17
ANSYS User Meeting
on Accuracy
10 substeps Reduce Constant
% difference % difference
Stabilization in stress Stabilization in stress
Deformation Energy # of from Deformation Energy # of from
Method Value Stress (psi) (in) (BTU) iterations nominal Stress (psi) (in) (BTU) iterations nominal
Damping 0.05 2.49E+06 10.4 4.23E-03 84 496.65 2.46E+06 10.161 9.76E-03 90 488.85
Damping 0.1 2.46E+06 10.233 8.27E-03 94 489.28 FAILED FAILED FAILED 255 FAILED
Damping 0.2 2.45E+06 10.069 1.72E-02 87 484.86 4.21E+05 1.0024 5.64E-04 68 0.69
Damping 0.25 4.21E+05 1.0024 3.15E-04 64 0.69 4.21E+05 1.0024 7.05E-04 72 0.69
Damping 0.5 4.21E+05 1.0024 6.30E-04 68 0.69 4.22E+05 1.0024 1.84E-03 106 0.81
Damping 0.8 FAILED FAILED FAILED 96 FAILED 2.42E+06 10.213 1.85E+00 241 479.86
Energy 1.00E-04 2.45E+06 10.068 5.42E-03 74 484.96 2.51E+06 10.47 1.22E-02 87 499.43
Energy 5.00E-04 4.21E+05 1.0024 4.14E-04 64 0.69 FAILED FAILED FAILED 186 FAILED
Energy 1.00E-03 4.22E+05 1.0024 6.18E-04 148 0.91 4.22E+05 1.0024 2.88E-03 95 0.82
Energy 1.00E-02 2.42E+06 10.225 2.11E+00 281 479.96 4.21E+05 1.0024 1.54E-02 93 0.78
Nominal Values
Stress (psi) Deformation (in) # of iterations
4.18E+05 1.0024 63
• Solving with 10 initial substeps

• Damping that is either too low or too high results in
unconverged solves or missed contact detection
• Damping values do not correlate with number of iterations
18
ANSYS User Meeting
on Accuracy
4 substeps Reduce Constant
% difference % difference
Deformation Energy # of from Deformation Energy # of from
Method Value Stress (psi) (in) (BTU) iterations nominal Stress (psi) (in) (BTU) iterations nominal
Damping 0.05 FAILED FAILED FAILED 217 FAILED FAILED FAILED FAILED 169 FAILED
Damping 0.2 FAILED FAILED FAILED 172 FAILED 4.21E+05 1.0024 1.32E-03 111 0.81
Damping 0.8 4.22E+05 1.0024 1.48E-03 114 0.82 2.45E+06 10.15 6.49E-01 160 484.88
Energy 1.00E-04 FAILED FAILED FAILED 218 FAILED FAILED FAILED FAILED 195 FAILED
Energy 5.00E-04 2.45E+06 10.118 4.87E-01 198 486.87 4.22E+05 1.0024 6.63E-03 78 0.83
Energy 1.00E-03 4.22E+05 1.0024 1.03E-02 171 0.91 2.42E+06 10.214 2.15E+00 255 478.98
Energy 1.00E-02 4.21E+05 1.0024 1.67E-02 39 0.72 2.38E+06 10.052 1.9046 124 468.55
Nominal Values
4.18E+05 1.0024 63
• Solve with 4 initial substeps rather than 10

• Damping that is either too low or too high results in
unconverged solves or missed contact detection
• Very few damping values lead to convergence ANSYS User Meeting
19
on Accuracy
% difference in
Deformation # of stress from
Method Value Stress (psi) (in) iterations nominal
Contact Damping, 10 substeps 0.05 4.21E+05 1.0023 47 0.78
Contact Damping, 10 substeps 0.5FAILED FAILED 183FAILED
Contact Damping, 1 substep 0.05 4.28E+05 1.0023 14 2.33

Contact Damping, 1 substep 0.1FAILED FAILED 32FAILED
Nominal Values
4.18E+05 1.0024 63
• Contact damping of 0.1 provides most accurate stress value at expense of

number of iterations
• Single Substep has low iterations, but higher error
• Using program controlled time stepping, initial contact is ignored but
secondary contact detected
20
ANSYS User Meeting
Case Study 2: No Initial Contact
• Analysis of a tubular frame, held with

a simple pin support at bottom
corner to allow rotations (as well as
planar symmetry)
• Vertical force at upper rear corner to
push frame into nearby wall with
frictional contact
• Frame is not in initial contact with
wall, and force magnitude is
significant enough to raise the
pseudo-velocity beyond manageable
substep sizes
• Since frame begins from an unstable
position, stabilization must be turned
on for first substep
– Alternatively, contact stabilization can be
used to slow the initial contact
21
ANSYS User Meeting
Case Study 2: No Initial Contact
• For baseline
comparison, solved
without stabilization by
rotating frame into
initial contact
• Check for effects of time
step length as well as
plasticity
22
ANSYS User Meeting
Case Study 2: Time Stepping
Effects on Accuracy
Linear Materials Reduce Constant
% %
difference difference
Time Deformation Energy # of from Deformation Energy # of from
Method Stepping Stress (psi) (in) (BTU) iterations nominal Stress (psi) (in) (BTU) iterations nominal
Global Damping = 0.1 1s, 10 step 1.09E+06 3.4117 1.09E-03 145 6.05 1.09E+06 3.4152 1.87E-03 130 5.98
Global Damping = 0.1 1s, 1 step 9.33E+05 13 1.19E-01 65 -8.90 9.35E+05 13.061 0.116 64 -8.76
Global Damping = 0.1 4s, 1 step FAILED FAILED FAILED 30 FAILED FAILED FAILED FAILED 30 FAILED
Global Energy Ratio = 1E-4 1s, 10 step 1.09E+06 3.4149 1.22E-02 103 6.04 1.09E+06 3.4149 1.05E-02 104 6.04
Global Energy Ratio = 1E-4 1s, 1 step 1.09E+06 3.4161 3.60E-01 146 6.03 6.48E+05 2.18 0.72431 12 -36.75
Global Energy Ratio = 1E-4 4s, 1 step 1.09E+06 3.4161 3.60E-01 146 6.03 6.48E+05 2.18 7.24E-01 12 -36.75
Contact Damping = 0.1 1s, 10 step N/A N/A N/A N/A N/A 1.03E+06 3.442 N/A 102 0.91
Nominal Values
1.02E+06 3.413959 112
• With program controlled time stepping (one substep), contact is often missed, or damped too highly
• Note that Energy Dissipation Ratio method is unaffected by time step lengths, but Damping Factor is
• Significantly less accurate than previous case study, likely due to extended sliding contact
• Contact Damping significantly more accurate stresses than global damping, but less accurate deformation
– Due to maximum stress being far from contact area
23
ANSYS User Meeting
Effects on Accuracy
• Example of an
overdamped
analysis
• High
stabilization
energy and low
accuracy
24
ANSYS User Meeting
Effects on Accuracy
% %
Global Damping = 0.1 1s, 1 step 1.51E+05 17.692 9.18E-02 756 6.28 1.51E+05 17.693 0.11354 698 6.33
Global Energy Ratio = 1E-4 1s, 4 step FAILED FAILED FAILED 53 FAILED FAILED FAILED FAILED 71 FAILED
Nominal Values
1.42E+05 17.69236 585
• Bilinear material properties required a higher Energy Dissipation Ratio (1E-

3) for convergence
• Similar accuracy to linear case, contact damping less effective at increasing
accuracy ANSYS User Meeting
25
Effects on Accuracy
• Bilinear
Deformation
26
ANSYS User Meeting
Case Study 3: Inherently
Unstable Models
• Same tubular frame as
Case Study 2, with wall
removed
• Model has a large
pivot/rigid body
response to applied load
– no nonlinear contacts
• Unstable equilibrium
point does exist
27
ANSYS User Meeting
Case Study 3: Inherently
Unstable Models
• Can be solved
without stabilization
by applying load in
equivalent vector to
deformed shape
• Still requires use of
weak springs for
inherent instability
(very low reaction
force)
28
ANSYS User Meeting
Effects on Accuracy
% %
Global Damping = 0.1 1s, 10 step 9.31E+05 13.053 2.05E-01 100 0.01 9.31E+05 13.053 2.09E-01 91 -0.06
Global Damping = 0.1 1s, 4 step 9.31E+05 13.037 1.95E-01 78 -0.02 9.29E+05 13.05 2.03E-01 79 -0.18
Global Damping = 0.1 4s, 10 step 9.30E+05 13.049 1.04E-01 95 -0.13 9.30E+05 13.049 1.06E-01 94 -0.14
Global Energy Ratio = 1E-4 1s, 4 step 9.31E+05 13.043 6.51E-01 62 -0.01 9.30E+05 13.015 8.82E-01 52 -0.17
Global Energy Ratio = 1E-4 1s, 1 step FAILED FAILED FAILED 81 FAILED 5.04E+05 0.83794 0.36625 4 -45.82
Global Energy Ratio = 1E-4 4s, 1 step FAILED FAILED FAILED 81 FAILED 5.04E+05 0.83794 0.36625 4 -45.82
Nominal Values
9.31E+05 13.05192775 50
• Significant variations in accuracy depending on time stepping with little

correlation
• Reduce method significantly more accurate than Constant in all cases
• Overdamped case exists using program controlled time stepping
29
ANSYS User Meeting
Effects on Accuracy
• Example of an
overdamped
analysis
• High
stabilization
energy and low
accuracy
30
ANSYS User Meeting
Effects on Accuracy
% %
Global Damping = 0.1 1s, 1 step 1.45E+05 21.689 2.73E-01 393 0.11 1.45E+05 21.69 0.2967 384 0.11
Nominal Values
1.45E+05 21.65429474 310
• Bilinear materials converged more often due to necessary bisecting
31
ANSYS User Meeting
Effects on Accuracy
• Bilinear
Deformation
32
ANSYS User Meeting
Conclusions
1. Viable solution to momentarily unstable models

2. Ideal to use local contact damping when possible
as opposed to global damping
3. Stabilization Energy dissipated does not
necessarily correlate to lost accuracy
4. Reduce method generally more accurate than
Constant and less likely to produce
overdamped/unconverged analyses
5. Energy dissipation can result in either over or
under conservative stress predictions
6. As with all analyses, sanity checks very important
33
ANSYS User Meeting
Input / Questions
34
ANSYS User Meeting
… within Epsilon

ANSYS-Nonlinear Stabilization User Meeting 2019nov20

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

ANSYS-Nonlinear Stabilization User Meeting 2019nov20

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

ANSYS-Nonlinear Stabilization User Meeting 2019nov20

Uploaded by

Copyright:

Available Formats

Twin Cities

ANSYS® User Meeting

Stabilization Damping for

1. Epsilon FEA Introduction

• Epsilon FEA provides engineering analysis (10 yrs!)

• Making Simulation Affordable

• Our customers need load-leveling with:

• Our customers benefit from external expertise

• Aids solver in converging rigid body motions

• Stabilization is an alternative to Arc Length Method, allows

1. Adds numerical viscous damping to affected nodes

• Stabilization Energy Result

• For baseline comparison, re-solved

• Solving with 10 initial substeps

• Solve with 4 initial substeps rather than 10

Contact Damping, 1 substep 0.05 4.28E+05 1.0023 14 2.33

• Contact damping of 0.1 provides most accurate stress value at expense of

• Analysis of a tubular frame, held with

• Bilinear material properties required a higher Energy Dissipation Ratio (1E-

• Significant variations in accuracy depending on time stepping with little

1. Viable solution to momentarily unstable models

You might also like