Some guidelines for implicit analyses in LS-DYNA guidelines for implicit analyses in LS-DYNA ... The...

Some guidelines for implicit analyses in LS-DYNA

Anders [email protected]

mailto:[email protected]

Implicit set-up in LS-DYNA

■ The Implicit user’s guide is available for download fromhttp://www.dynasupport.com/howtos/implicit/some-guidelines-for-implicit-analyses-using-ls-dyna

■ It contains PDF-documents:■ Some guidelines for implicit analyses using LS-DYNA, ■ A Guide to Solving Implicit Mortar Contact Problems in LS-DYNA

■ and LS-DYNA keyword files:■ recommended control card settings for different implicit analysis types, and■ some small examples of different analysis types and features, including restart, rubber

and frequency domain

■ The “Guide to Solving Implicit Mortar Contact Problems” provides some background to the control card settings.

■ The focus of the “Guidelines for implicit analyses” is how to set up the analysis as such, and to guide in choice of elements, contacts, material models etc. The aim is not to describe the control cards settings in detail.

2017-03-28User's guide to Implicit analyses in LS-DYNA

http://www.dynasupport.com/howtos/implicit/some-guidelines-for-implicit-analyses-using-ls-dyna


■ Other sources of useful information regarding implicit analyses in LS-DYNA:■ www.dynasupport.com/howtos/implicit■ http://www.dynaexamples.com/implicit■ www.dynalook.com – search conference papers

■ Dynamore / LSTC also gives courses in implicit analyses:■ Advanced implicit analysis in LS-DYNA (T. Borrvall)■ NVH & Frequency domain analysis in LS-DYNA (Y. Huang / LSTC)■ Introduction to LS-DYNA non-linear implicit analysis (A. Jonsson)

■ see also: https://www.dynamore.se/en/training/seminars

■ Please report bugs and errors in this package! To [email protected] or [email protected]


http://www.dynasupport.com/howtos/implicit

http://www.dynaexamples.com/implicit

http://www.dynalook.com/

https://www.dynamore.se/en/training/seminars

mailto:[email protected]

Implicit set-up in LS-DYNA – Motivation



■ Given a specific task, well defined with respect to geometry, loading and material properties, the user still must make many different choices in order for LS-DYNA work properly:

■ Control cards, telling LS-DYNA how to solve the task■ Element formulations (many different formulations available)■ Contact definitions (many different formulations and options to chose from)■ Material models (over 250 to chose from)

■ The objective of the User’s guide is to reduce the time it takes to make these choices, so that the time can be spent on solving the actual task.


Implicit set-up in LS-DYNA – Control cards

■ Identify analysis type and select appropriate control card include file.■ In many cases, *CONTROL_TERMINATION is the only required additional control card.

■ Use an include file structure! Then the control card include files can be used directly.


Include files Add keywords:(*)

Analysis type control_cards_... database_cards_... *CONTROL_IMPLICIT_...

Linear static linear.key static.key

Non-linear static nonlin.key static.key

Linear buckling (1)static.key BUCKLE

Non-linear postbuckling

arc.key static.key

Eigenfrequency analysis

(1)static.key EIGENVALUE

Linear transient modal dynamics

linear.key dynamic.key MODAL_DYNAMIC

Frequency domain analyses(2)

linear.key

Non-linear implicit dynamics

nonlin.key dynamic.key DYNAMICS

Notes: (*) *CONTROL_TERMINATION must always be added. (1) Can be part of both linear and non-linear analysis. (2) Frequency response functions, steady state dynamics etc.

Implicit set-up in LS-DYNA – Control cards

■ Identify analysis type and select appropriate control card include file.■ In many cases, *CONTROL_TERMINATION is the only required additional control card.

■ Use an include file structure! Then the control card include files can be used directly.


Attached examples


Filename Example ofdoorstiff001.key Linear static analysisbend001.key Non-linear static analysisbolts001.key Non-linear dynamic/static analysis. Bolt pre-

tensioningbuckle001.key Linear buckling analysislimitload001.key Non-linear limit load analysiseigen001.key Eigenfrequency analysiseigen002.key Intermittent eigenfrequency analysistransient001.key Linear transient modal dynamicstransient_nonlin001.key Non-linear transient dynamicsssd_fatigue001.key Steady state dynamics with fatigue evaluationSet_4.key Random vibration analysis with fatigue evaluationTPU.key RubberInsert_and_interference.key

Rubber, press-fit, contacts

Large_deformation.key Rubber, contacts, large deformationspretens001.key Bolt pre-tensioning, first step of restart analysisfull_restart.key Example of full restart, from pretens001.keysmall_restart.key Small restart from pretens001.key

Attached examples


doorstiff001.key limitload001.key

■ A rubber bushing is press fit into a steel casing, and a steel ball is inserted into the bushing


Attached examples - Rubber

Recommended element types

■ For non-linear analyses

■ For linear analyses


Element type Comment LS-DYNA keyword Element formulation Beam *SECTION_BEAM 1

For bolts w. pre-tension

9

For springs, dampers etc.

6

Shell 1st order *SECTION_SHELL 162nd order 23

Solid 1st order hex *SECTION_SOLID -22nd order tet 162nd order hex 23

Element type Comment LS-DYNA keyword Element formulationBeam For linear static

analyses*SECTION_BEAM 13(1)

Shell 1st order *SECTION_SHELL 18 (20, 21, 99)(2)

Solid 1st order hex *SECTION_SOLID 18 (99)Notes: (1) The only allowed beam type for linear static analyses. (2) All shell and solid element formulations from non-linear analysis can also be applied in linear static analyses

Materials

■ Some comments on material models:■ From R9, *MAT_PIECEWISE_LINEAR_PLASTICITY works well also for

implicit analyses. ■ It is not recommended to use *MAT_PLASTIC_KINEMATIC with BETA = 1 if

isotropic hardening is desired. Use *MAT_PIECEWISE_LINEAR_PLASTICITYinstead.

■ Avoid using *MAT_ELASTIC with ν ≈ 0.5 for modelling rubber materials. Use *MAT_HYPERELASTIC_RUBBER or *MAT_SIMPLIFIED_RUBBER_FOAMinstead.

■ For linear static analyses, it is recommended to use *MAT_ELASTIC if stresses are of interest.

■ For modelling of creep effects, *MAT_THERMO_ELASTO_VISCOPLASTIC_CREEP (MAT_188) is recommended.


Contacts

■ For tied contact, there are many options available.

■ It is strongly recommended to check which nodes get tied and not! Use the pre-processor’s contact check functionality. Also check messages printed in d3hsp. Nodes that are not tied will be reported.


*CONTACT_TIED_ Constraint formulation

DOFs Movenodes(1)

Momenttransferred(2)

NODES_TO_SURFACE Kinematic 1 – 3 YesNODES_TO_SURFACE_OFFSET Penalty 1 – 3 No NoNODES_TO_SURFACE_CONSTRAINED_OFFSET Kinematic 1 - 3 No Yes

SHELL_EDGE_TO_SURFACE Kinematic All(3) YesSHELL_EDGE_TO_SURFACE_OFFSET Penalty All(3) No NoSHELL_EDGE_TO_SURFACE_BEAM_OFFSET Penalty All(3) No YesSHELL_EDGE_TO_SURFACE_CONSTRAINED_OFFSET Kinematic All(3) No YesNotes: (1) Only slave nodes within the search distance will be moved. (2) Due to separation between slave nodes and master segments. Only applies to the _OFFSET_ option. (3) All available degrees of freedom, depending on element type of the slave node.

Contacts

■ For sliding contact in implicit analyses, the recommendation is to use the Mortar contacts:■ *CONTACT_AUTOMATIC_SURFACE_TO_SURFACE_MORTAR_ID,

■ *CONTACT_AUTOMATIC_SINGLE_SURFACE_MORTAR_ID.

■ Normally requires a minimum of input:■ define what is in contact using parts or part sets, and■ specify coefficient of friction.

■ Penetration information will be reported in mes0* - files. ■ If “too large” penetrations are noted during the simulation, contact stiffness can be

increased using the SFS- and IGAP-parameters.


*CONTACT_AUTOMATIC_SINGLE_SURFACE_MORTAR_ID$# cid title

1010Global contact all$# ssid msid sstyp mstyp sboxid mboxid spr mpr

0 0 5 0 0 0 0 0$# fs fd dc vc vdc penchk bt dt

0.15 $# sfs sfm sst mst sfst sfmt fsf vsf

Contacts

■ For Mortar contact, IGAP controls the progressive contact stiffness increase

■ Mortar contact can also handle press-fit, using IGNORE = 3 (“small” penetrations) or 4 (finite penetrations), and MPAR1 to specify time to resolve the initial penetrations, and MPAR2 to specify search distance for penetrating nodes.

■ Use the softer part (less stiff material, or coarser mesh) as slave.


Troubleshooting convergence problems

■ Start out with the recommended control card settings.■ Check the model integrity.

■ Element quality, element types, un-intended cracks in the mesh, element connectivity (beams attached directly to solids by common node(s)) etc.

■ Many pre-processors have built-in tools for model checking

■ Check that consistent units are used (for materials, loadings, accelerations etc.)

■ Check material models■ Avoid use of *MAT_ELASTIC for rubbers (at least for finite deformations)■ User MAT_24 instead of MAT_3 with BETA = 1■ Check last segment of hardening curve! Since

LS-DYNA will extrapolate, negative last segment slope is not recommended.

■ Check (and correct) unintended initial penetrations in contacts.■ Use INGORE = 3 or 4 of the Mortar contacts to handle press-fit.

■ Check tie contacts. Ensure that the intended slave nodes are tied (not too many, not to few)


■ Use a moderate (not too small) initial time step. ■ Negative eigenvalue warnings are most likely due to rigid body modes or

elements that gets severely distorted (for example rubber) during deformation.■ Check for rigid body modes by eigenvalue analysis (*CONTROL_IMPLICIT_EIGENVALUE)■ Use implicit dynamics (or in some cases inertia relief) to properly handle rigid body

modes.

■ Use the d3iter – file! Looking at the non-converged states can give a lot of information, if for example loose parts in the structure are “flying away”. Scale up the displacement, or fringe plot it, in order to identify areas where large changes in displacement take place.■ Set RESPLT = 1 on Card 4 of *DATABASE_EXTENT_BINARY to be able to fringe plot the

residual in the d3iter – files. This can pinpoint where in the model convergence is the hardest.

■ Relax penalty stiffness of contacts. ■ The Mortar contacts will print penetration messages, and based on this information the

user can judge what is acceptable in terms of relative or absolute penetration distance.



■ Is the physical problem in fact unstable? Will collapse / buckling / bifurcation occur? ■ Try switching from load control to displacement control ■ Try using the arc-length solver ■ Turn on dynamics

■ For analyses of rubber (or other incompressible materials), turn off the geometric stiffness effect (set IGS = 2 on *CONTROL_IMPLICIT_GENERAL).

■ For (dynamic) simulations with (very) small time steps, for example 1.E-4, modify the DTMIN settings of *CONTROL_TERMINATION and *CONTROL_IMPLICIT_AUTO (to for example 1.E-6) This allows the implicit solver to re-try step with a smaller time step if convergence fails.

■ Switching to full-Newton can be efficient for highly non-linear problems■ Set ILIMIT=1 and increase MAXREF (to 30 – 60) on *CONTROL_IMPLICIT_SOLUTION

■ Activating the non-symmetrical equation solver (set LCPAK = 3 on *CONTROL_IMPLICIT_SOLVER) may help in some cases, such as follower loads, or “snap-through” deformation.



■ Check for “too easy” convergence. Look for “noisy” contact forces, or unsatisfied global equilibrium (for example if spcforc doesn’t add up to the applied force)■ Tighten tolerances slightly, set for example DCTOL = 5.E-4 on *CONTROL_IMPLICIT_SOLUTION

■ If the global displacements are large, switch to DNORM = 1 (possibly also increase DCTOL)



Example of contact forces from “too easy” convergence. For Displacement > 100 mm, the contact force using the recommended convergence tolerances oscillates very much in this case. Switching to DNORM = 1 resolves this.

Restart in implicit LS-DYNA

■ LS-DYNA has two different types of restart:■ Small deck restart, allowing only minor modifications of the model, such as termination

times, load curves, output intervals■ Full deck restart, allows complete re-definition of the model.

■ How to setup multiple load cases for small deck restart, for example bolt pre-tensioning followed by X, Y and Z-loading:■ Define all load cases in the main run file, using different load curve IDs.■ Specify zero entries in all load curves except one (for example X-loading).■ Set termination time to (for example) 1.0 in the main run file, so that only the pre-

tensioning is performed.■ The first load case will be a very simple restart, just changing *CONTROL_TERMINATION.■ To run the other load cases, change *CONTROL_TERMINATION and use *CHANGE_CURVE_DEFINITION and *DEFINE_CURVE to re-define the load curves.




*LOAD_NODE_POINT80610 1 100101 -982.80611 2 100102 -982.80612 3 100103 98.2

*DEFINE_CURVE_TITLELoad case 1 curve

1001010.,0.1.,0.2.,1.3.,1.*DEFINE_CURVE_TITLELoad case 2 curve

1001020.,0.1.,0.2.,0.3.,0.*DEFINE_CURVE_TITLELoad case 3 curve

1001030.,0.1.,0.2.,0.3.,0



*CONTROL_TERMINATION2.0

*CHANGE_CURVE_DEFINITION100101100102

*DEFINE_CURVE_TITLELoad case 1 curve (zero)

1001010.,0.1.,0.2.,0.3.,0.*DEFINE_CURVE_TITLELoad case 2 curve (active)

1001020.,0.1.,0.2.,1.3.,1.

Thank you!

Your LS-DYNA distributor and more


Some guidelines for implicit analyses in LS-DYNA guidelines for implicit analyses in LS-DYNA ... The...

Documents

Transcript of Some guidelines for implicit analyses in LS-DYNA guidelines for implicit analyses in LS-DYNA ... The...