Doküman 1 - Explicit Theory

Ansys Workbench LS-DYNA
Module 01: Explicit Theory and

Workbench LS-DYNA
Release 2021 R1
©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

Learning Outcomes for this Course
• After completing this course, Ansys Workbench LS-DYNA, you will be able to
‐ Understand the explicit solution method and apply it to a wide variety of complex nonlinear transient
simulations
‐ Use the robust pre- and postprocessing capabilities of Workbench Mechanical to set up and
postprocess your explicit LS-DYNA simulations
‐ Leverage the robust nonlinear solution capabilities of LS-DYNA for both transient and static
simulations
‐ Postprocess results with Ansys Mechanical and LS-PrePost
‐ Evaluate results from explicit analysis, ensuring accurate and trustworthy simulations
2 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

Learning Outcomes
• After completing this module, you will be able to

‐ Identify those applications uniquely suited to explicit analysis
‐ Understand important explicit analysis fundamentals including the importance of time step, energy
balance and mass scaling
‐ Apply those fundamentals towards achieving efficient solutions with confidence in the results

Module 01 Topics
1. What is LS-DYNA?
2. Typical Applications for Workbench LS-DYNA
3. Implicit and Explicit Methods
4. Time Integration
5. Explicit Time Step
6. Mass Scaling
7. Checking Mass Scaling
8. Mass Scaling Summary
9. Workshop 01.1: Taylor Impact

01.1 What is LS-DYNA?
• A multi-field solver using the Finite Element Method, including

‐ Element Free Galerkin (EFG), Smoothed Particle Hydrodynamics (SPH), Arbitrary Lagrangian Eulerian
(ALE)
‐ Implicit (static and dynamic) and Explicit time integration techniques
‐ Eigenvalue buckling and modal analysis
‐ Coupled analysis with other physics (i.e. thermal, Fluid Structure Interaction,…)
• Originally developed by LSTC (Livermore Software Technology Corporation)
‐ LSTC acquired by Ansys in 2019 and is now Ansys / LST
• LS-DYNA is the combination of multi-field solvers
‐ It reads a keyword input file and writes result files
• Partial integration into Ansys via the Workbench Environment
‐ Continued development focus for further integration

01.1 What is LS-DYNA?
EXPLICIT (short time
dynamics) LS-TaSC
LS-PrePost
Shape, Topology
All LS-DYNA Pre & Incompressible Conservation Optimization
Post Processing Computational Fluid Element/Solution
Dynamics (ICFD) Element (CESE)
Chemistry Thermal
LS-DYNA®
Particle Methods Iso-Geometric
Analysis (IGA)
NVH/Acoustics Electromagnetics Barriers
LS-Opt Occupant
IMPLICIT (static and
Optimization slow dynamics) Validated barriers and
Robustness Occupants
Strongly Coupled SCALABLE Multi-Physics Solver
01.2 Typical Applications for Workbench LS-DYNA
• Drop test of all forms

• Impacts
• Product misuse / severe loadings
• Product failure / fragmentation
• Containment safety and penetration mechanics
• Large plasticity in mechanisms
• Sports equipment design
• Manufacturing processes like machining /
cutting / drawing

01.2 Typical Applications for Workbench LS-DYNA
• Large deformations
‐ Large displacements
‐ Large rotations
‐ Large strains
• Complex contact
‐ Efficient self contact
‐ Abrupt status change
‐ Eroding contact
• Nonlinear Material ✓Short time dynamics

✓ Highly non-linear applications (quasi-static)
‐ Plasticity
‐ Hyperelasticity
‐ Failure
1.3 Implicit and Explicit Methods
Problem Time Magnitude
1 year 10 s 1s 0.1 s 0.01 s 0.001 s 0.0001 s
“Non-linearity”
Creep Static/Dynamic Quasi-Static Drop / impact Ballistics Detonation Hypervelocity

& Blast Impact
IMPLICIT METHODS
EXPLICIT METHODS
LS-DYNA

01.4 Time Integration
Single Degree-of-Freedom System
u(t) - Displacement Inertia fI
Elasticity fS
k p(t)
p(t) m
m
Damping fD
c
Equation of Motion:

• For structural mechanics, in general we must solve the equation of motion:
𝑀 ⋅ ü + 𝐶 ⋅ 𝑢ሶ + 𝐾 ⋅ 𝑢 = 𝑝(𝑡)
• Solve for unknown time dependent quantities:

‐ acceleration, velocity, displacement
‐ derive: strain, stress, reaction forces, …
• We want to know the solution over time!
• For some problems, the solution can be derived analytically
‐ Harmonic analysis: harmonic loads cause harmonic response
• But in general, a numerical approach is necessary.
‐ Discretization in time with time step t
‐ Time integration scheme required to describe the approximation of the solution

Multiple Degree-of-Freedom System
• Equation of Motion: 𝑀 ⋅ ü + 𝐶 ⋅ 𝑢ሶ + 𝐾 ⋅ 𝑢 = 𝑝(𝑡)

‐ Arbitrary function of time
‐ time discretization needed
• Implicit time integration

‐ The equilibrium is set up for new time tn+1
𝑀𝑛+1 ⋅ ü𝑛+1 + 𝐶𝑛+1 ⋅ 𝑢ሶ 𝑛+1 + 𝐾𝑛+1 ⋅ 𝑢𝑛+1 = 𝑝𝑛+1
• Explicit time integration

‐ The equilibrium is set up for current time tn
𝑀𝑛 ⋅ ü𝑛 + 𝐶𝑛 ⋅ 𝑢ሶ 𝑛 + 𝐾𝑛 ⋅ 𝑢𝑛 = 𝑝𝑛
• Implicit: Newmark method / linear acceleration method

ü
ün+1 assumption: linear change in acceleration
ün t t
velocity: u n +1 = u n + un + un +1
2 2
t 2 t 2
displacement: un +1 = un + u n t + un + un +1
3 6
tn tn+1 t
equation of motion at time tn+1: M n +1  ün +1 + Cn +1  u n +1 + K n +1  un +1 = pn +1
6 3 6 6 3 Δt
displacement at time tn+1: ( M n +1 + C n +1 + K n +1 )  u n +1 = pn +1 + M n ( u
2 n
+ 
u n + 2un ) + C n ( u n + 2u n + un )
Δt 2
Δt Δt Δt Δt 2

• Explicit: Central Difference Method

u
exact solution
assumption: linear change in displacement
u n +1/ 2
1
velocity: u n +1/ 2 = (un +1 − un )
t n +1/ 2
1
acceleration: un = (u n +1/ 2 − u n −1/ 2 )
t n
t n −1 t n −1/ 2 tn t n +1/ 2 t n +1 t
t n −1/ 2 t n +1/ 2
t n
equation of motion at time tn: M n  ün + Cn  u n + K n  un = pn
1 1 2 1 1
displacement at new time tn+1: ( M n + C n )  u n +1 = pn − (K n − M n )  u n − ( M n − Cn )  un −1
Δt 2 2 Δt Δt 2 Δt 2 2 Δt

• Characteristics of implicit • Characteristics of explicit

Stiffness matrix Lumped mass matrix
6 1
( 2 M n +1 + K n +1 )  un +1 = pn +1 K n +1 = K n +1 (un +1 ) M n  un +1 = pn M n = diag(mi )
Δt Δt 2
‐ Iteration within time step if stiffness is ‐ No iterations
nonlinear • no convergence necessary
• User intervention required for convergence
‐ Stiffness matrix only on the right side, is
• Automatic and manual adjustment of time step known
‐ Stiffness matrix must be solved as a large
system of equations: ‐ Except for unknown displacements, all values
from previous time step
• linear equation solver like PCG or SPARSE required
‐ CPU time per time step depends on equation ‐ Because of diagonal mass matrix, uncoupled
equations, no matrix operations, no system of
solver equations
• size of system of equations (number of
nodes*DOF/node) • Inversion of lumped mass matrix is trivial
‐ Memory demanding ‐ Low memory requirements
• Make sure to solve in-core (matrix held in memory) • Scalar operations
• Implicit Time Integration
‐ The integration method is always stable, independent of time step size
‐ User effort focused on achieving convergence
• Explicit Time Integration
‐ The integration method is only stable if the time step size is smaller than a critical time step size
(conditional stable). The critical time step size correlates with the highest natural frequency of the
system and reads for linear systems without viscous damping
‐ Because of nonlinearities, critical time step size may change in each cycle
2
t  t crit =
 max
‐ User effort focused on time step ‐ CPU effort depends on

• check small time steps (puts constraints on meshing) • Number of cycles
• limit end time (i.e. for quasi-static) • Number of elements
• number of cycles = tend/Δt • Complexity of element formulation

01.5 Explicit Time Step
• The explicit time integration must capture all frequencies of the FE model
• Modal analysis is time, memory and CPU intensive and must be avoided
‐ Frequencies change continuously during the analysis.
• Time step is adjusted to wave propagation throughout the solution
• The time step must be smaller than the critical time step
‐ A safety factor of 0.9 is used by default, see Analysis Settings
• HOW IS THE TIME STEP DETERMINED??

• Analysis time step, ∆t, depends on element size!

‐ Time step must be limited to prevent a traveling wave from passing through more than 1 element
during a time step
‐ Small elements control the time step used to advance the solution in time
‐ Uniform element size must be controlled
• Element size controlled by user throughout the mesh
‐ Not automatically dependent on the geometry
• Implicit analyses usually have a static region of stress concentration where mesh is refined (strongly dependent on
geometry)
• Explicit analyses exhibit stress waves propagating through mesh, changing the regions of stress concentration
‐ Mesh refinement is usually used to improve efficiency
• Mesh transitions should be smooth for maximum accuracy
• Mesh refinement does not necessarily imply a smaller element size!

• Total solution time depends on the time step and number of elements
t simu = tcycle  ncycle 

  nelem   t simu
 t  
t simu = (nelem  telem )      t   t simu
 t  
t simu − computation time for simulation
𝑛𝑐𝑦𝑐𝑙𝑒 − 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑦𝑐𝑙𝑒𝑠
tcycle − computation time for 1 cycle 𝑡 − 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑒𝑛𝑑 𝑡𝑖𝑚𝑒
telem − computation time for 1 element Δ𝑡 − 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 𝑠𝑡𝑒𝑝
nelem − number of elements
• How to estimate and control the simulation time step …?

• Courant-Friedrichs-Lewy (CFL) Stability Criteria

‐ L/c is the time it takes an elastic stress wave to traverse the element
2 𝐿
t  t crit
= Δ𝑡 = 0.9
 max
𝑐
L - characteristic element length

c - speed of sound in material  dependent on element type
0.9 - timestep scale factor
Solid elements: Shell elements: Beam elements:

L
L
L E Young‘s modulus
 Poisson‘s ratio
𝐸(1 − 𝜈) 𝐸 𝐸
𝑐= 𝑐= 𝑐=  Density
𝜌(1 + 𝜈)(1 − 2𝜈) 𝜌(1 − 𝜈 2 ) 𝜌

• Charcteristic Element Length, L – Shell Elements

‐ Many possibilities to choose from, edge lengths or diagonals, maximums or minimums?
‐ Using maximum of edge lengths will provide smallest L
l4 lc=min(l1, l2, l3, l4)
d2 d1 lc=max(l1, l2, l3, l4) 𝐴
l3 l1 𝐿=
lc=min(d1, d2) max(𝑙1 ,𝑙2, 𝑙3, 𝑙4 )
l2 lc=max(d1, d2)
• Characteristic Element Length - Solids
𝑉𝑜𝑙𝑢𝑚𝑒
𝐿=
𝐴𝑟𝑒𝑎max 𝑠𝑖𝑑𝑒

• Material Speed of Sound, c = √(E/ρ)

‐ E = Young’s modulus
‐ ρ = material density Material Speed of Sound (m/s)
‐ Thus, c increases with stiffness and decreases Steel 5240
with density Aluminum 5328
‐ Larger c = lower ∆t Titanium 5220
• To increase timestep and reduce Plexiglass 2598
analysis time Water 1478
‐ Increase density (Mass Scaling) Air 331
‐ Lower modulus
‐ Increase element size

• Time Step larger than CFL criteria • Time Step within CFL criteria
‐ Solution instabilities

• Example, Rectangular Shell Elements Rigid Ground Plane –

Does not affect time step!
‐ 100 mm x 50 mm
‐ E = 2E+05 MPa Flexible Body
‐ ν = 0.3 (Steel)
‐ ρ = 7.85E-09 tonne/mm3
• L = A/Lengthmax edge = 50
2𝐸+05 100 mm
• c= √ = 5.29E+06 mm/s
(7.85𝐸−09)(1−0.3^2) 50 mm
• ∆t = 0.9 L/c = 8.505E-06 s
• No need to perform this calculation
manually…..

• LS-DYNA offers a time step estimation

‐ LSDYNA Pre Toolbar: CFL Time Step
‐ Helpful to identify locations to remesh
‐ Time step also reflected in Solution Information
during solution
For a simulation with End Time, tend = 0.001 s, the

number of time steps (cycles) will be tend / ∆t = 118

• Time Step - Summary

‐ The CFL Time Step is updated throughout the analysis as elements deform

• Depends upon element type, size and linear material properties; use realistic density and stiffness t  0.9  L 
E
• Elements in tension tend to increase the time step
• Elements in compression tend to decrease the time step t end
‐ Explicit doesn’t perform iterations, instead number of time steps depends on end time n=
t
• Large end time will produce long solution times
‐ With a non-uniform mesh, a few elements can drastically affect the time step; uniform mesh is
preferred!
‐ Time step is reduced if
• Element size is reduced (mesh refinement)
• Density is reduced
• Stiffness is increased (young’s modulus)
‐ Rigid elements don’t affect the step size
For a simulation with End Time, tend = 0.001 s, the

number of time steps (cycles) will be tend / ∆t = 1127

01.6 Mass Scaling
• Best way to control time step size is with a

uniform mesh; what if a uniform mesh is not
possible?
‐ Complex geometry meshed with tetrahedrons or mixture
of brick/tets or quads/triangles
• Mass Scaling
‐ User defines the desired time step size (Analysis Settings)
• Turn on Automatic Mass Scaling
• Default “Time Step Size” is 1E-07 s
‐ Values larger than critical time step would cause unstable
solution
‐ Solver adjusts the element time step by increasing
element mass (density) for those elements with a CFL
time step less than the prescribed time step size

01.6 Mass Scaling
• Example Mass Scaling calculation for shell elements

‐ Density adjustment for any shell element with CFL time step smaller than solution time step
Ln,min E
t specified = and c=
c  (1 −  2 )
( t specified )2  E
 n =
Ln  (1 −  2 )
2
‐ From the earlier example with uniform element edge Lmin = 50 mm

• E = 2E+05 MPa
• ν = 0.3
• ρ = 7.85E-09 tonne/mm3
• Resulting time step was 8.505E-06 s
‐ For the non-uniform mesh at right with Lmin = 6.12 mm
• If desired time step is 3.0E-06 s instead of 8.87E-07 s
• ρn = 5.28E-08 tonne/mm3 for each element that doesn’t meet the desired time step
• IS THIS AN ACCEPTABLE INCREASE IN MASS?

01.6 Mass Scaling
• Mass scaling is used to speed up analysis

‐ Pure numerical approach
‐ Ensures a constant time step size – end of calculation can be estimated
‐ BUT, it modifies the physics of the problem!
• Mass scaling increases inertia forces
‐ i.e. 𝐹 = 𝑚𝑢ሷ
‐ But not body loads (e.g. gravity)
• Mass scaling must be limited:
‐ due to non-physical behavior
• Check the mass scaling in post processing and make sure the results are valid

01.7 Checking Mass Scaling
• Explicit solutions do not perform equilibrium

iterations to enforce a force balance in the system
‐ Explicit relies on energy balance in the system
• From the example shown thus far
‐ Original Part mass = 1.57E-02 tonne
‐ Initial part velocity = 4428.7 mm/s
‐ Initial kinetic energy in the system should be:
• K.E. = 1/2mV2 = 1.54E+05 mJ
‐ After solution, Solution Information reports Energy Summary
• Initial K.E. = 3.35E+05 mJ!!
• Mass scaling was too much!

‐ Too much mass added
‐ Problem physics have changed

• In the previous example,

‐ The desired time step 3.0E-06 was too large
‐ The original mesh size was too small; too many elements violated the desired CFL criteria
‐ Too much mass was added to the model; Kinetic Energy doubled!
• Options are:
‐ Increase mesh size and achieve more uniformity in element size via additional meshing controls
‐ Decrease the desired time step (solution time will increase)

• Other ways to check for mass scaling

‐ Solution Information Solver Output (below)
‐ Added Mass Result Tracker in Solution Information (right)
• Offers ability to track added mass per part

01.8 Mass Scaling Summary
• Mass scaling is used to speed up analysis, but it may change the physics of the problem
• Mass scaling increases inertia forces
• Mass scaling must be limited:
‐ Due to non-physical behavior
‐ May not be appropriate for cases involving high speed ballistic impact and failure
• Check the mass scaling in postprocessing and make sure the results are valid
‐ Look at full model energy summary – is initial kinetic energy correct?
‐ Total added mass and ratio from Solution Output may not be an indicator
• Example: a heavy train at rest and a vehicle impacting; ratio may still be small in this case
‐ Look at added mass per part
‐ rule of thumb for parts of interest:
• dynamic analysis: A few percent (up to 5 %)
• quasi static analysis: 20%, but problem dependent
‐ Solve again with reduced time step; If results don’t change, mass scaling effect is low

01.9 Workshop 01.1: Taylor Impact
• Goal: To become familiar with the Ansys LS-DYNA workflow by performing a Taylor
Impact test
‐ cylinder made of elasto-plastic steel hits a rigid wall with a defined initial velocity

End of presentation
01.7 Workshop 01.2: Postprocessing
• Goal: To become familiar with the various files that LS-DYNA writes and the alternate
postprocessing tools they provide

• Other ways to check for mass scaling
• Full model mass increase is reported in

‐ Solution Information (messag file) and d3hsp
‐ ASCII result file glstat (also % increase)
• Part mass increase is reported in

‐ ASCII result file matsum
‐ Binary result file d3plot
• Nodal mass increase is reported in


01.6 Mass Scaling
• Percentage nodal mass increase

‐ LSPP: Post/FriComp/Misc/mass scaling
‐ Typically, maximum nodal mass increase is much
larger than per part or per model, 100 or 200%
may be ok for a few small elements
• How to verify if mass increase is reasonable

or not?
‐ At least running again with reduced time step can
confirm the analysis. If results don’t change, mass
scaling effect is low.

01.6 Mass Scaling
d3hsp:
• Time step size for mass scaling 100 smallest timesteps

------------------------------------------------
element number part timestep
‐ Smallest time steps are listed in d3hsp solid 3300 1 1.6023E-07
‐ Search for “100 smallest” to find the listing “100 …
solid 2068 1 1.6023E-07
smallest timesteps” solid 2889 1 1.6945E-07

solid 2801 1 1.6945E-07
• Typically, a time step at the list’s end is suitable
solid 2537 1 1.6945E-07
• Keep in mind that a small percentage of Solution Information:

mass increase for the full model may lead calculation with mass scaling for minimum dt
to a large increase for certain parts and added mass =
physical mass=
1.0043E-06
3.2471E-04
elements! ratio = 3.0930E-03
1 t 0.0000E+00 dt 1.80E-07 flush i/o buffers
1 t 0.0000E+00 dt 1.80E-07 write d3plot file
1 t 0.0000E+00 dt 1.80E-07 write intfor file
problem cycle = 56
time = 9.9000E-06
added mass = 1.0043E-06
percentage increase = 3.0930E-01
…

Doküman 1 - Explicit Theory

Uploaded by

Copyright:

Available Formats

Doküman 1 - Explicit Theory

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

Doküman 1 - Explicit Theory

Uploaded by

Copyright:

Available Formats

Ansys Workbench LS-DYNA

Module 01: Explicit Theory and

©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

2 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• After completing this module, you will be able to

3 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

4 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• A multi-field solver using the Finite Element Method, including

5 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

NVH/Acoustics Electromagnetics Barriers

• Drop test of all forms

7 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Nonlinear Material ✓Short time dynamics

Creep Static/Dynamic Quasi-Static Drop / impact Ballistics Detonation Hypervelocity

9 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

10 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• For structural mechanics, in general we must solve the equation of motion:

• Solve for unknown time dependent quantities:

11 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

Multiple Degree-of-Freedom System

• Equation of Motion: 𝑀 ⋅ ü + 𝐶 ⋅ 𝑢ሶ + 𝐾 ⋅ 𝑢 = 𝑝(𝑡)

• Implicit time integration

𝑀𝑛+1 ⋅ ü𝑛+1 + 𝐶𝑛+1 ⋅ 𝑢ሶ 𝑛+1 + 𝐾𝑛+1 ⋅ 𝑢𝑛+1 = 𝑝𝑛+1

• Explicit time integration

• Implicit: Newmark method / linear acceleration method

ün+1 assumption: linear change in acceleration

equation of motion at time tn+1: M n +1  ün +1 + Cn +1  u n +1 + K n +1  un +1 = pn +1

13 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Explicit: Central Difference Method

equation of motion at time tn: M n  ün + Cn  u n + K n  un = pn

14 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Characteristics of implicit • Characteristics of explicit

‐ User effort focused on time step ‐ CPU effort depends on

16 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

17 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Analysis time step, ∆t, depends on element size!

18 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

t simu = tcycle  ncycle 

nelem − number of elements

• How to estimate and control the simulation time step …?

19 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Courant-Friedrichs-Lewy (CFL) Stability Criteria

L - characteristic element length

0.9 - timestep scale factor

Solid elements: Shell elements: Beam elements:

20 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Charcteristic Element Length, L – Shell Elements

21 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Material Speed of Sound, c = √(E/ρ)

22 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

23 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Example, Rectangular Shell Elements Rigid Ground Plane –

24 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• LS-DYNA offers a time step estimation

For a simulation with End Time, tend = 0.001 s, the

25 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

• Time Step - Summary

For a simulation with End Time, tend = 0.001 s, the

26 ©2021 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.