Recent Developments in LS-DYNA®

18th Annual HP Symposium Dilip Bhalsod April 3, 2007 Livermore Software Technology Corporation

Outline of talk
• Introduction • Recent developments
– – – – – – – – – – Shell seatbelts Particle method for airbags Sensors Contacts Selective mass scaling Implicit Incompressible flow Electromagnetic forming Hot forming Conclusions


LS-DYNA development
• Our developments are concentrated on three products:
– LS-Dyna – LS-Opt – LS-PrePost

• LS-PrePost and LS-Opt are part of the LSDyna distribution and do not require license keys.

Development goals
• Combine multi-physics capabilities in a scalable code for solving highly nonlinear transient problems to enable the solution of coupled multi-physics and multi-stage problems in one run
• • • • • • • • • • • Full 2D & 3D capabilities Explicit Solver Implicit Solver Heat Transfer ALE, EFG, SPH, particle methods Navier-Stokes Fluids (version 980) Radiation transport (version 980) Electromagnetics (version 980) Acoustics Interfaces for users, i.e., elements, materials, loads, etc. Interfaces with other software, Madymo, USA, etc.

Development goals-implicit
• Springback for sheet metal stamping • Static initialization of crash models • Dynamic springback simulation after crash simulation
– Reliable measurements between numerical and physical results can be more easily obtained

• An embedded linear capability to automatically solve for normal modes, attachment modes, and constraint modes
– Include infinitesimal motions superimposed on rigid bodies for NVH and durability modeling

• Eigenvalue analysis to check the rigid body modes in the crash models
– Identify inadvertent constraints

one model for crash. durability. and NVH modeling • Explicit durability and NVH modeling go mainstream in MD Nastran • No optional added cost LSTC developed features in LSDYNA 6 . durability. and fuel tank. • Manufacturing simulation results from LS-DYNA used in crash. LS-DYNA.LSTC’s vision • In automotive. to enable the complete modeling of crash including airbags. NVH shared and maintained across analysis groups • One scalable multi-physics code. occupants.

LSTC’s vision • • • • LS-DYNA specific pre-processing. post-processing. with no added charges. LS-PrePost. LS-OPT is considered a critical enabling technology • 7 . Unrestricted open databases Focus on large distributed memory low-cost clusters running large simulations As processor costs decrease and cluster sizes increase. LS-OPT. and optimization. LS-DYNA software prices per processor will proportionally decrease to keep simulation costs affordable Optimization technology will automate engineering design calculations.

000 elements. • MPP version is recommended if more than 4 processors are used per run • Model sizes continue to grow faster than Processor speed 8 . typical large simulation models typically contain 1.000000 to 4. • FEA dummies are preferred over rigid body dummies in crash simulations • 12-32 processors are used in runs that complete within 12-24 hours • Calculations give digit-to-digit repeatability for a fixed domain decomposition.000.Current state of explicit • Currently.

Near future for explicit • Model sizes of 10.000. such as THUMS.000 elements • 128-512 processors in overnight runs • Human dummy models. will increase model sizes even further • Honeycomb barriers will be modeled by shell elements • Number of processors will increase 5-10 times • Optimization software use in crash analysis will become widespread 9 .

ALE.Final goal for explicit simulations • Simulation results accepted in place of prototype testing – What is required? • Strict modeling guidelines for analysts. NVH. etc. Particle methods Sensors and control systems Complete compatibility with NASTRAN • Manufacturing simulations (in LS-DYNA. Durability. etc.) providing the initial conditions for crash simulations 10 . Moldflow. and a single comprehensive model for crash. • Continued software improvements – – – – – Constitutive models Contact FSI with SPH.

it is safe to predict that serial and SMP implicit solvers used in large scale nonlinear simulations will also become obsolete within the next 5 years.Parallel computing • In less than one decade from 1998-2006 the use of explicit codes has undergone a radical transformation – From 100% serial and SMP licensed CPU’s for crash to 90% MPP with the remaining 10% of CPU’s typically running smaller models on 1-8 processors – Today serial and SMP explicit codes are becoming obsolete and will eventually be phased out • What about implicit? – More difficult to create an MPP version – Requires more expensive hardware so there is less customer pressure to create MPP versions – However. 11 .

Recent version 971 developments 12 .

Seatbelt shell element Definition of slipring and retractor 13 .

14 .

will be read.Extended formats • Customer requests for extended formats are increasing since 8 character IDs are too restrictive – All keyword formats are being optionally extended • • • • 2 cards will be read for each existing card I10 > I20.0 – Current formats are being kept and all changes will be backwards compatible. A 2-5 year lead time is expected.0 > E20. I8 > I16. 15 . E10. – New formats are being added to 971_R3 for early release since preprocessors need to be updated.0 Mixed default and extended formats will be read Structured input will uniformly go to I20 and E20. • All old input files. both structured and keyword.

velocity and displacement sensor for suspension.g. tire • Force and torque sensor for suspension • Acceleration.Sensors for Automotive Safety • There are over 110 sensors in the average vehicle today. vehicle and occupant • Occupant classification sensor for restraint system • Strain gauge for steel and seat 16 . • Pressure sensor for brake.. e.

contact. Multiple status switch is possible. The output signal is the numerical value of a physical sensor. The output signal is the logical result. 17 . airbag. based on the logical result of a set of SENSOR_SWITCH. true or false.….*Sensor • *SENSOR_DEFINE defines a physical sensor. • *SENSOR_CONTROL controls a function. • *SENSOR_SWITCH compares the sensor signal with a given criterion.


*Contact_force_transducer • Scalability of LS-DYNA to hundred’s of processors is limited by contact – Scalability for 1-2 contact definitions covering the entire vehicle is excellent – Scalability for 300+ contact definitions diminishes significantly as the number of processors increase • 300 contact definitions are used to obtain detailed reaction force information. 19 .

20 . – Problem is that these transducers cannot separate the forces in two-sided contact • In version 971_R2 a master surface is also accepted such that the reaction force is accumulated by the interactions of the slave and master surfaces – It is now possible to move to fewer contact interfaces and improve scalability.*Contact_force_transducer • Force transducers measure contact forces within a vehicle by accumulating all forces acting on a segment set during contact.

Smooth contact with CAD • Contact with CAD surface (IGES and VDA formats) has been available in LS-DYNA for ~15 years but rarely used – Advantages: • Smooth surface with eliminates sudden change in surface normal vectors – Contact force calculations and work piece stresses are more accurate – Springback predictions are more accurate – Disadvantages: • Contact search and treatment is time-consuming • Not robust due to poor quality CAD data 21 .

22 . Reduces the contact noise and produces a smoother result with less mesh sensitivity – More accurate residual stresses provide for springback results which are comparable to CAD springback results • • Less sensitivity to contact penalty parameters Implemented for MPI.Smooth contact with FEA • • • • A smooth curve-fitted surface represents the contact surface Eliminates sudden changes in the surface normal across the element Provides a more accurate representation of the physical surface.



e.Conventional mass scaling • Lumped nodal masses are scaled (increased) to increase stable time step in explicit finite element analysis • Severe scaling unavoidably introduces unwanted. i. • Dummy interacting with steering wheel finely meshed with solid elements 25 . inertia effects in the structure under consideration – this will in practice limit the amount by which the time step can be increased • Crash models are often forced to run at small step sizes due to critical components where mass scaling will lead to wrong results. nonphysical..

but leaves the low frequency domain relatively unaffected – severe mass scaling can be performed without deteriorating the accuracy of results 26 .Selective mass scaling • Mass is increased under the constraint that the rigid body translational behavior is preserved • Lowers the high frequencies which allows for a larger stable time step.

usage • Selective mass scaling can be performed on the entire model or a subset of parts • Selective mass scaling and conventional mass scaling can run together in the same model • Selective mass scaling is activated using a single parameter on *CONTROL_TIME STEP 27 .Selective mass scaling .

7 hours with selective mass scaling dt=1.1 microseconds • Simulation times to 50 ms • 5.e-07 • 0.e-06 28 .5 hours with mass scaled such that dt=1.Steering wheel example • Steering wheel impacted by body block • Solid elements dt< 0.

Steering wheel example • Conventional mass scaling with dt=2.e-07 SMS SMS CMS CMS 29 .e-06 dt=2.e-06 dt=1.e-06 Red: Cyan: Blue: Green: dt=1.e-06 dt=2.

• • Tank test simulation – resident gas in the tank/bag can be considered 3 unit systems currently supported for ease of input. 30 .PARTICLE METHOD • • • • • • Very similar input as Control volume input Venting and Porosity supported – Blocked and unblocked Internal baffles defined through sid2 – Flow through internal baffles can be monitored Multiple inflators Switching from Particle method to CV 2 step procedure – Extract mdot and Temp. from gasguide stations – Use of extracted data in step 1 as input to individual chambers.

NIDn ANn B1 B2 Bn VX1 VX2 VXn STYP2 PATM NID1 BLOCK TATM NID2 HCONV C23 TEND NID3 TSW Cair C1 C2 Cn VY1 VY2 VYn Pair LCM1 LCM2 LCMn VZ1 VZ2 VZn Tair LCT1 LCT2 LCTn INFO1 INFO2 INFOn INFG1 INFG2.PARTICLE METHOD *AIRBAG_PARTICLE SID1 STYP1 SID2 NP UNIT VIS IAIR NGAS NORIF Extra card if IAIR=1: XMair Aair Bair NGAS cards: XM1 A1 XM2 A2 . INFGn initial air data gas component data inlet orifice data 31 . XMn An NORIF cards: NID1 AN1 NID2 AN1 .

Particle method for airbag deployment Example: flat bag after 5ms compared to results from Eulerian and uniform pressure models particle method multi-material Eulerian and fluid-structure interaction uniform pressure 32 .

Particle method for airbag deployment Example: airbag center deflection compared to results from Eulerian and uniform pressure models bag center deflection (mm) Eulerian particle UP time (ms) 33 .

Tank Test using Particle method 34 .

Tank Test using Particle method 35 .

Control Volume – Particle Method .ALE 36 .

Control Volume – Particle Method .ALE 37 .

CPU minutes for 3 methods ALE Particle Method Control volume 0 50 100 150 38 .

Particle method with curtain bag 39 .

Particle method with curtain bag with switch to cv at 20ms 40 .

Incompressible flow (980) • Explicit-implicit solver • Lagrangian treatment of most interfaces • Automatic meshing and adaptive remeshing • FSI with large deformations • MPP implementation 41 .

Flow Past a Cylinder (Re=200) Surface and Volume Adaptivity Based on Velocity Gradient Error (Prescribed error = 1%) 42 .

final mesh size ~2M 43 . Initial mesh size: ~200K elements. error=1.Flow around a structure Max.5%.

This motion has an effect on the fields and the currents. changing the temperature. The forces induce motion and deformation of the conductors. These currents generate magnetic fields. The currents generate Joule heating in the conductors.LS-DYNA electromagnetism • • • • • • Introduction of electrical currents in solid conductors. The magnetic fields coupled with the currents generate Lorentz forces on the conductors. 44 . and thus some mechanical as well as electromagnetic properties (conductivity for example). as well as induced currents. electric fields.

Magnetic Metal Forming Coil Holder (insulator) Coil Workpiece Die 45 .


Example • Model of jet turbine fan blades – 2 sets of fan blades on a common shaft – 24 fan blades on each – Lots of eigenvalues with multiplicity 48 • Requested the smallest 100 eigenvalues. 70M nonzeroes in lower triangle of K. 47 . 2.06M dofs. • 700K nodes.

48 .

Implicit followed by Explicit 49 .

Unconnected parts 50 .

Fully connected parts 51 .

Performance 20000 18000 16000 CPU Seconds 14000 12000 10000 8000 6000 4000 2000 0 0 5 10 # of Processes 52 15 20 .

MPP Eigensolver Speed up 18 16 14 12 10 8 6 4 2 0 0 5 10 15 20 53 Speed-up Ideal .




requiring multiple-stages. multi-physics.Conclusions • Version 971 has become a major release with many new capabilities that focus on the automotive industry – – – – MPP implicit ALE is at production level for out-of-position occupant analysis EFG is implemented for scalable MPP applications Particle method for airbags • • Significant progress has been made to include most crash capabilities within the implicit solver LSTC’s software development goal continues to be the implementation within one scalable code of all capabilities required to solve problems that involve. Full compatibility with linear structure models leading to a one model environment 57 • . running on large clusters of processors.


Sign up to vote on this title
UsefulNot useful