OpenRadioss: the open-source crash and impact solver

OpenRadioss is the open-source release of Altair Radioss, an explicit dynamics solver with roughly thirty years of industrial use behind it. Altair published it under the AGPL-3.0 in September 2022.

That is a more significant event than most open-source releases. Radioss is a production crash solver — used in automotive safety development, where results inform designs that are then physically crash-tested against regulatory requirements. Until this release there was no open-source code in that class at all: explicit capability existed in general-purpose FEA packages, but a validated crash code did not.

Explicit dynamics covers events measured in milliseconds, where inertia dominates and the structure is deforming violently. Why that needs a different solver is worth reading first; in short, implicit FEA cannot practically handle it.

• Vehicle crash — full-vehicle and component-level impact, the problem the code was built around.
• Drop test — consumer products, enclosures, packaging, transport cases.
• Impact and ballistics — projectile penetration, bird strike, debris impact.
• Blast — structural response to explosive loading.
• Metal forming — stamping and deep drawing, where large plastic deformation and contact are the whole problem.

The distinction between “has an explicit solver” and “is a crash code” comes down to three things, and they are where thirty years of industrial use actually lives.

Material models. Real materials are strain-rate dependent: steel and especially polymers behave differently when deformed in milliseconds than in a slow tensile test. A crash code carries a large library of rate-dependent models with failure criteria, plus composites, foams and honeycombs.

Contact. In an impact the solver spends most of its effort determining what is touching what, as a structure folds onto itself and surfaces come into self-contact. Robust contact algorithms are what stop a crash model from failing rather than deforming.

Element failure and deletion. Elements that have accumulated enough damage must be removed so a tear propagates. Doing that stably, without the energy bookkeeping falling apart, is genuinely hard.

For comparison, several general-purpose FEA codes carry an explicit option that is perfectly adequate for a drop test. None of them is a crash code in this sense — worth knowing before choosing one for the job.

OpenRadioss is AGPL-3.0, which differs from the GPL that most of the other solvers in this space use. The AGPL extends copyleft obligations to software delivered over a network: if you modify it and offer it as a hosted service, the modified source must be made available to users of that service.

For running it on your own machines this changes nothing. For anyone planning to build a cloud simulation service on top of it, it is a real consideration and one for a lawyer rather than a forum post. It is also a reasonable read of Altair’s intent: open for engineers to use, less immediately convenient to repackage as a competing SaaS.

The input is a Radioss starter and engine deck — text files defining the model, materials, contacts and the run. The format is well documented and it is not gentle: crash models are large, and the setup surface is correspondingly wide.

Explicit work also demands discipline the solver cannot supply. Watching the smallest element, because one sliver sets the time step for the entire model. Keeping mass scaling modest enough not to change the physics. Checking energy balance before reading any result at all. These are the habits that separate a credible impact model from an animation.