LS-DYNA Explicit Dynamics: Keyword File Setup, Element Formulation, and Solver Configuration

LS-DYNA Explicit Dynamics: Keyword File Setup, Element Formulation, and Solver Configuration

LS-DYNA was the first explicit dynamics solver I ever used, and honestly, the keyword file format intimidated me at first. It's just text — no GUI, no tree structure, just cards and parameters. But once I learned the key cards, it became second nature. Let me walk you through the workflow I use for setting up LS-DYNA models, from element formulation to energy balance checks.

LS-DYNA Workflow

Overview

1. Pre-processing: LS-PrePost or external pre-processor (HyperMesh, ANSA)
2. Keyword file: Text file (.k) containing all model definition
3. Solver: LS-DYNA executable (SMP or MPP)
4. Post-processing: LS-PrePost for results visualization

Keyword File Structure

The LS-DYNA keyword file is organized by keywords (cards):

*KEYWORD
*TITLE
Model title
*CONTROL_TERMINATION
   0.1         0         0         0         0         0
*CONTROL_TIMESTEP
   0.0       0.9         0         0         0         0
*DATABASE_GLSTAT
   0.001
*DATABASE_BINARY_D3PLOT
   0.005
*END

Mesh and Elements

Shell Elements

1. *ELEMENT_SHELL
2. Element formulation (ELFORM):
   • 2: Belytschko-Tsay (default, fast, good for most)
   • 5: S/R Hughes-Liu (improved warping, slower)
   • 10: Belytschko-Wong-Chiang (better for warped geometry)
   • 16: Fully integrated shell (no hourglass, slower)
3. Section properties (*SECTION_SHELL):
   • Thickness: t (mm)
   • Integration points: 3-5 through thickness (5 recommended for plasticity)
   • ELFORM: 2 (default)

Solid Elements

1. *ELEMENT_SOLID
2. Element formulation (ELFORM):
   • 1: Constant stress (1-point integration, default, fast)
   • 2: Fully integrated (S/R, no hourglass, slower)
   • 10: Tetrahedron (1-point, for complex geometry)
   • 13: Tetrahedron (10-node, quadratic, accurate but slow)
3. Section properties (*SECTION_SOLID):
   • ELFORM: 1 (default for most)

Beam Elements

1. *ELEMENT_BEAM
2. Element formulation (ELFORM):
   • 1: Hughes-Liu (default, good for large rotation)
   • 2: Belytschko-Schwer (resultant beam)
   • 3: Truss (axial only)
3. Section properties (*SECTION_BEAM):
   • Cross-section: Rectangular, circular, or custom
   • Integration: Through section

Hourglass Control

1. Reduced integration elements (ELFORM 1, 2) have hourglass modes
2. *CONTROL_HOURGLASS:
   • IHQ: Hourglass control type
     • 1: Flanagan-Belytschko (stiffness, default)
     • 2: Flanagan-Belytschko (viscous)
     • 4: Flanagan-Belytschko (stiffness, enhanced)
     • 5: Type 5 (orthogonal stiffness)
   • QH: Hourglass coefficient (0.1 default, 0.05 for more accuracy)
3. Check: Hourglass energy < 5% of internal energy

Material Models

Elastic

1. *MAT_ELASTIC (MAT_001):
   • RO: Density (kg/m³ or tonne/mm³)
   • E: Young's modulus (Pa or MPa)
   • PR: Poisson's ratio
2. Use for: Linear elastic materials (steel, aluminum in elastic range)

Plastic Kinematic

1. *MAT_PLASTIC_KINEMATIC (MAT_003):
   • RO: Density
   • E: Young's modulus
   • PR: Poisson's ratio
   • SIGY: Yield stress
   • ET: Tangent modulus (plastic hardening)
   • BETA: Hardening parameter (0=kinematic, 1=isotropic)
   • SRC: Strain rate parameter (Cowper-Symonds)
   • SRP: Strain rate exponent
2. Use for: General plasticity with strain rate effects

Johnson-Cook

1. *MAT_JOHNSON_COOK (MAT_015):
   • A: Yield stress
   • B: Hardening coefficient
   • n: Hardening exponent
   • C: Strain rate coefficient
   • m: Temperature exponent
   • Tm: Melt temperature
   • Tr: Room temperature
2. Flow stress: σ = (A + Bεⁿ)(1 + C ln(ε̇*))(1 - T*ᵐ)
3. Use for: High strain rate, temperature-dependent metals (crash, impact)

Piecewise Linear Plasticity

1. *MAT_PIECEWISE_LINEAR_PLASTICITY (MAT_024):
   • SIGY: Initial yield stress
   • ET: Tangent modulus
   • LCSS: Load curve for stress-strain
   • C: Strain rate coefficient (Cowper-Symonds)
   • P: Strain rate exponent
2. Use for: Custom stress-strain curve with strain rate

Failure Models

1. *MAT_ADD_EROSION:
   • PFAIL: Failure pressure (spallation)
   • EPSFAIL: Failure principal strain
   • SIGMAX: Maximum principal stress
   • SIGVM: Maximum von Mises stress
2. When criterion is met: Element is deleted from analysis
3. Use for: Fracture, tearing, fragmentation

Contact

Automatic Single Surface

1. *CONTACT_AUTOMATIC_SINGLE_SURFACE:
   • All surfaces can contact themselves and each other
   • Most general contact (like Abaqus general contact)
2. Parameters:
   • SSID: Part set ID (which parts are in contact)
   • FS, FD: Static and dynamic friction coefficients
   • DC: Optional damping
3. Use for: Crash (parts fold and contact themselves)

Automatic Surface to Surface

1. *CONTACT_AUTOMATIC_SURFACE_TO_SURFACE:
   • Define master and slave surfaces
   • More controlled than single surface
2. Parameters:
   • SSID: Slave surface
   • MSID: Master surface
   • FS, FD: Friction coefficients
3. Use for: Known contact pairs (impact, forming)

Contact Formulation

1. Penalty method (default):
   • Contact force = kpenalty × penetration
   • kpenalty: Auto-calculated from material stiffness
   • Can scale: SFS (slave), SFM (master)
2. Soft constraint:
   • For soft-on-hard contact (rubber on steel)
   • Penalty based on segment area, not material stiffness
3. Constraint method:
   • Lagrange multiplier (exact, but can oscillate)

Contact Tips

1. Use *CONTACT_AUTOMATIC_SINGLE_SURFACE for crash
2. Use *CONTACT_AUTOMATIC_SURFACE_TO_SURFACE for known pairs
3. Check contact penetration in post-processing
4. If penetration is excessive: increase SFS (penalty scale factor)
5. If contact is unstable: add contact damping (DC)

Time Step Control

Stable Time Step

1. *CONTROL_TIMESTEP:
   • DT2LC: Time step scale factor (0.9 default, 0.67 for safety)
   • TSSFAC: Scale factor for shell/solid elements
2. Stable time step:
   • Δtcrit = Lmin / c
   • Lmin: Smallest element characteristic length
   • c: Wave speed = √(E/ρ)
3. Example: Steel, Lmin = 1mm
   • c = √(200×10⁹/7850) = 5048 m/s
   • Δtcrit = 0.001/5048 = 1.98×10⁻⁷ s ≈ 0.2 μs
4. For 100 ms simulation: ~500,000 cycles

Mass Scaling

1. *CONTROL_TIMESTEP:
   • DT2MS: Mass scaling target time step
   • Positive: Add mass to elements below target dt
   • Negative: Add mass and report
2. Example: DT2MS = 1×10⁻⁷
   • All elements with dt < 1×10⁻⁷ get mass added
   • Check: Mass increase < 5% total (quasi-static)
3. *DATABASE_MASS:
   • Report mass scaling information
   • Check mass added per part

Time Step Output

1. *DATABASE_GLSTAT:
   • Reports time step, kinetic energy, internal energy per cycle
2. Monitor:
   • Time step: Should be stable (not dropping significantly)
   • If time step drops: Element is distorting (check mesh quality)

Output Control

Binary Output

1. *DATABASE_BINARY_D3PLOT:
   • DT: Time interval between plot states (e.g., 0.005s)
   • Number of states: T_end / DT
2. d3plot files contain:
   • Deformation
   • Stress and strain
   • Element deletion (erosion)
   • Contact information

ASCII Output

1. *DATABASE_GLSTAT:
   • Global statistics per cycle
   • Time, KE, IE, hourglass, total energy
2. *DATABASE_MATSUM:
   • Per-part energy summary
3. *DATABASE_RCFORC:
   • Resultant contact forces
4. *DATABASE_SPCFORC:
   • Reaction forces at constraints
5. *DATABASE_NODFOR:
   • Nodal forces

Energy Balance

Energy Components

1. Kinetic energy (KE): 0.5 × Σ(m × v²)
2. Internal energy (IE): Strain + plastic energy
3. Hourglass energy (HG): Artificial energy from hourglass control
4. Contact energy (CE): Energy from contact sliding (friction)
5. Total energy (TE): KE + IE + HG + CE
6. External work (EW): Work done by external forces

Energy Checks

1. Conservation: TE ≈ EW (total energy = external work)
2. Hourglass: HG < 5% of IE
3. Mass scaling: Check mass increase < 5% (quasi-static)
4. Contact energy: CE should be reasonable (friction work)
5. If TE ≠ EW: Energy is being created or destroyed (check for errors)

Running LS-DYNA

SMP vs. MPP

1. SMP (Shared Memory Parallel):
   • Single machine, multiple cores
   • Up to 8-16 cores
   • Simpler setup
2. MPP (Massively Parallel Processing):
   • Distributed memory (cluster)
   • Hundreds to thousands of cores
   • Domain decomposition
   • Required for large models (> 1M elements)

Command Line

ls-dyna_smp_d i=inputfile.k ncpu=8 memory=2G

Or for MPP:

mpirun -np 64 ls-dyna_mpp_d i=inputfile.k memory=2G

Memory

1. Memory: Specified in words (1 word = 8 bytes on 64-bit)
2. memory=2G: 2 billion words = 16 GB
3. Estimate: ~1 million words per 1000 elements (varies by model)
4. If memory is insufficient: LS-DYNA reports required memory

Verification Checklist

• [ ] Element formulation is appropriate (ELFORM 2 for shells, 1 for solids)
• [ ] Hourglass control is active (check HG energy < 5% IE)
• [ ] Material model matches material behavior (Johnson-Cook for high rate)
• [ ] Contact includes all possible contact pairs
• [ ] Time step scale factor is appropriate (0.9 default)
• [ ] Mass scaling is checked (< 5% mass increase for quasi-static)
• [ ] Energy balance is satisfied (TE ≈ EW)
• [ ] Element deletion is used appropriately (failure criteria)
• [ ] Output frequency captures key events
• [ ] Results are mesh-independent (refine and compare)

Wrapping Up

LS-DYNA's keyword file format gives you control over everything, which is both its strength and its weakness. There's no GUI holding your hand — if you get a card wrong, the solver might run but give you bad results. The two things I check on every LS-DYNA run: hourglass energy (keep it under 5% of internal energy) and mass scaling (keep added mass under 5% for quasi-static). If either of those is too high, your results aren't trustworthy regardless of how good everything else looks.