Written and maintained by CASRAI Editorial Board
Last updated
Short answer
LS-DYNA is an advanced, general-purpose multiphysics simulation program developed around explicit and implicit finite element methods. Originally authored as DYNA3D by Dr. John O. Hallquist at Lawrence Livermore National Laboratory and later expanded by LSTC and Ansys, it specializes in modeling transient, highly nonlinear mechanical events including structural impacts, high-velocity collisions, material failure, and fluid-structure interactions.
Definition and scope
LS-DYNA is a nonlinear transient dynamic finite element analysis (FEA) software suite designed to simulate structural deformation, severe impact, and coupled multiphysics behavior. Developed in 1976 as DYNA3D by Dr. John O. Hallquist at Lawrence Livermore National Laboratory (LLNL), early versions were released into the public domain in 1978. Hallquist founded the Livermore Software Technology Corporation (LSTC) in 1987 to commercialize the code as LS-DYNA, which was subsequently acquired by Ansys in 2019.
The operational scope of LS-DYNA centers on high-strain-rate phenomena where inertia, shock wave propagation, and large geometric distortions dominate physical behavior. Unlike traditional implicit structural codes that often diverge under extreme mesh distortion or contact changes, LS-DYNA solves dynamic equations across small time increments without factorizing a global stiffness matrix. Models are constructed using ASCII keyword decks bounded by *KEYWORD and *END cards, structured into parameter blocks such as *CONTROL, *PART, *SECTION, *MAT, *CONTACT, and *DATABASE. Strict control over these parameter cards is essential to maintain computational reproducibility across simulation batches.
How it works or how it is used in practice
LS-DYNA operates primarily through explicit time integration via the central difference method. At each time step tn, nodal accelerations are computed directly from the lumped diagonal mass matrix M and force balance: M ün = Fext,n – Fint,n. Because M is diagonal, no global matrix inversion is required, allowing efficient scaling across millions of elements. The integration scheme is conditionally stable, governed by the Courant-Friedrichs-Lewy (CFL) limit: Δt ≤ Lmin / c, where Lmin is the characteristic element length and c = √(E / ρ) is the dilatational wave speed. When tiny elements restrict the time step, researchers apply selective mass scaling via *CONTROL_TIMESTEP to maintain practical runtimes while monitoring kinetic energy additions.
The software also features an implicit solver using Newton-Raphson iterations for static equilibrium, gravity loading, and sheet metal springback. Researchers frequently switch solvers in sequence, executing an explicit forming step followed by an implicit springback recovery. Over 300 constitutive models are supported, including linear elasticity (*MAT_001), rate-dependent piecewise plasticity (*MAT_024), Johnson-Cook fracture for high-velocity ballistic impacts (*MAT_015), Mooney-Rivlin hyperelasticity (*MAT_027), and concrete damage models (*MAT_072R3). Penalty-based contact algorithms, such as *CONTACT_AUTOMATIC_SINGLE_SURFACE, track self-contact and folding without requiring predefined contact pairs.
Multiphysics capabilities include Incompressible CFD (ICFD) for free-surface fluid-structure interaction, CESE for supersonic shock waves, and electro-magnetics (EM) for magnetic metal forming and battery failure modeling. To resolve extreme deformation where standard meshes tangle, LS-DYNA employs meshless methods like Smoothed Particle Hydrodynamics (SPH) and the Discrete Element Method (DEM). While researchers select Abaqus for general implicit mechanical assemblies or COMSOL Multiphysics for continuous field equations, LS-DYNA remains the reference standard for high-energy structural collisions.
Types or related concepts
Selecting an appropriate LS-DYNA formulation depends on event duration, deformation magnitude, and phase interactions. The following table compares primary solver formulations verified against official Ansys LS-DYNA manuals:
| Formulation | Governing Integration Scheme | Stability Limit | Matrix Inversion | Primary Research Domain |
|---|---|---|---|---|
| Explicit Dynamics | Central difference integration; diagonal lumped mass | Conditionally stable; bounded by acoustic CFL limit | None required; solved element by element | Automotive crashworthiness, bird strike, ballistic impact, drop testing |
| Implicit Structural | Newton-Raphson equilibrium iterations | Unconditionally stable; governed by convergence | Requires global stiffness matrix factorization | Pre-loading, modal analysis, thermal equilibrium, sheet metal springback |
| Arbitrary Lagrangian-Eulerian (ALE) | Lagrangian deformation followed by Eulerian advection | Governed by element size and fluid advection velocity | None; explicit advection step | Fluid sloshing, underwater explosions, high-speed fluid-structure interaction |
| Smoothed Particle Hydrodynamics (SPH) | Mesh-free Lagrangian continuum particle smoothing | Conditionally stable by smoothing length and sound speed | None; neighbor particle search | Hypervelocity penetration, fragmentation, fluid jetting, material shredding |
Where it matters for labs, clinics or supply decisions
In computational mechanics laboratories, LS-DYNA bridges physical testing and virtual prototyping. Experimental data from drop towers, gas guns, and split-Hopkinson pressure bars calibrate nonlinear material cards (such as *MAT_024), establishing verified baselines before structural simulations.
In biomedical and clinical biomechanics, LS-DYNA is the standard solver for evaluating traumatic brain injury (TBI) and orthopedic impacts. Research groups deploy biofidelic human models—such as THUMS and GHBMC—alongside virtual Anthropomorphic Test Devices (ATDs) to capture occupant injury metrics. Device teams also simulate stent expansion and catheter delivery using nonlinear contact and hyperelastic cards.
Laboratory computing infrastructure demands key operational choices:
- SMP vs. MPP Architecture: Shared Memory Parallel (SMP) operates across threads on a single workstation node. Massively Parallel Processing (MPP) uses domain decomposition and MPI communication across distributed cluster nodes, scaling across hundreds of cores.
- Simulation Governance: Because transient models generate large binary outputs (such as
d3plotfield files andglstatenergy logs), laboratories should implement a software management plan to standardize input keyword decks, scripts, and retention policies. - Stochastic Sensitivity: Due to sensitivity to initial mesh alignments, engineers couple LS-DYNA with tools like LS-OPT or run Monte Carlo simulation workflows to evaluate manufacturing tolerances and probabilistic failure thresholds.
Common misunderstandings
- Explicit integration is always faster: Explicit steps execute quickly because no matrix is inverted, but conditional stability demands tiny time steps (often microseconds). For long-duration quasi-static processes, an implicit solver with large time steps is far more efficient.
- Ignoring zero-energy hourglass modes: Under-integrated elements run quickly and avoid shear locking, but can produce zero-energy hourglass deformation. Engineers must monitor
*DATABASE_GLSTAT; if hourglass energy exceeds 5% to 10% of internal energy, hourglass stiffness controls or fully integrated elements must be applied. - Unchecked mass scaling: Applying mass scaling via
*CONTROL_TIMESTEPartificially increases density in tiny elements to expand time steps. Excessive mass scaling adds non-physical inertia, distorting acceleration pulses and falsifying dynamic responses. Added mass should remain under 1% of total model mass. - Assuming LS-DYNA is solely for automotive crash: While dominant in crashworthiness, modern LS-DYNA integrates comprehensive solvers for incompressible CFD, compressible shock aerodynamics, electromagnetics, and discrete particle physics.
Frequently asked questions
What is the primary difference between explicit and implicit solvers in LS-DYNA?
The explicit solver advances equations directly using a diagonal mass matrix without matrix inversion, excelling at high-speed, transient impact events. The implicit solver resolves static or dynamic equilibrium by assembling and inverting global stiffness matrices, ideal for long-duration loading and springback.
How does LS-DYNA calculate the Courant time step limit?
For each element, LS-DYNA computes the critical time step as the ratio of characteristic length to acoustic dilatational sound speed (c = √(E/ρ)). The solver identifies the minimum value across all elements and applies a scale factor (typically 0.90) to guarantee numerical stability.
What key output files are produced during an LS-DYNA run?
Essential outputs include d3plot (binary state files for 3D visual post-processing in LS-PrePost), glstat (global energy balance tracking kinetic, internal, and hourglass energy), matsum (individual material energy histories), and messag (solver diagnostics and warning logs).
When is Smoothed Particle Hydrodynamics (SPH) preferred over standard finite elements?
SPH is preferred during severe structural fragmentation, perforation, or fluid jetting where conventional Lagrangian grids suffer from severe mesh tangling, element inversion, or negative Jacobian calculation errors.
Sources
- Ansys, Inc. "Ansys LS-DYNA: Multiphysics Solver & Explicit Dynamics Software." Product Overview and Documentation Portal. URL: https://www.ansys.com/products/structures/ansys-ls-dyna (Accessed November 9, 2026).
- Ansys, Inc. / Livermore Software Technology Corporation (LSTC). "LS-DYNA Keyword User's Manual (Volume I: Core Keywords, Volume II: Material Models, Volume III: Multiphysics) and Theory Manual." Documentation Portal. URL: https://lsdyna.ansys.com/manuals/ (Accessed November 9, 2026).
- Hallquist, John O., Lawrence Livermore National Laboratory (LLNL) & Department of Energy (DOE) OSTI technical records. "DYNA3D: An Explicit, Finite Element and Finite Difference Code for Dynamic Analysis of Solids in Three Dimensions." Historical and Technical Documentation. URL: https://www.osti.gov/biblio/1989989 and https://lsdyna.ansys.com/ (Accessed November 9, 2026).
