SafeLight
Failure and post-buckling material modelling
The problem
Cracks in ultra-high-strength steel are hard to predict
Cracking around holes, edges and welds in ultra-high-strength steel is difficult to predict under crash loading — and it is exactly where a lightweight structure fails first.
SafeLight calibrates failure strain as a function of triaxiality, from specimens whose geometry is chosen to sit at different stress states.
Why the usual indicators fall shortTensile strain, forming limits, element size, stress triaxiality
Plastic strain from a standard tensile test, or a forming limit diagram, is not an adequate indicator. The answer depends heavily on element size and holds only for monotonic loading, and the maximum plastic strain at failure itself varies with element size and sheet thickness.
Making a car both safe and light means using the material to its full extent, which means modelling its real strength past first yield and through post-buckling rather than stopping where a simpler indicator says it should have failed.
How much strain a material survives depends on how it is being pulled. Sheared, stretched in one direction, or stretched in two at once, the same material fails at very different strains — the difference is what stress triaxiality measures, and it is why one number from a tensile test cannot stand in for failure everywhere in a component.
Product overview
A failure and post-buckling material model for crash, calibrated per material from specimens loaded at three different stress states.
Status
Material model · in commercial use since 2020 · calibration required per material
Evidence boundary
Calibration is required per material. Simulation footage is computed output, not a physical validation test.
Capabilities
- Loading-state sensitivity
- Element size and thickness
- Post-buckling force response
- Computed component response
Where two models part
Two failure strains on the same spot-welded steel assembly. The curves are indistinguishable through peak load and separate only once failure initiates.
What the clip shows
A hot-stamped 22MnB5 hat profile with tailored soft zones, resistance spot welded to a DP780 cover plate and crushed along its axis. The assembly is run twice, the two models differing only in the local failure strain assigned to them. Both reach 103.53 kN at 7.8 ms and stay on top of each other until 9.2 ms, when failure initiates; after that they are as much as 41.0 kN apart. Equivalent plastic strain is contoured on both. Forming, welding and crush were simulated as one chain under CiSMA.
- Peak, both cases
- 103.53 kN
- Curves separate
- 9.2 ms
- Largest difference
- 41.0 kN
What happens when the material varies
Crush the same cast component three times, changing nothing but the failure strain, and peak force moves 52.4 kN across the whole measured spread.
What the clip shows
Force–time curves for one ribbed HPDC aluminium component crushed along its axis, run at the low end, the mean and the high end of the measured failure-strain distribution. The three curves rise together to roughly 850 kN, separate only after peak load, and decay through the fold. Peak force spans 843.3 to 895.7 kN.
- −3σ
- 843.3 kN
- Mean
- 856.4 kN
- +3σ
- 895.7 kN
Method and sources
How the model is calibrated
Failure initiation is calibrated with the Stepwise Modelling Method, a Luleå University of Technology method that Aerobase uses and did not invent. Specimens are loaded at three different stress states, and full-field digital image correlation measures the strain as each one fails.
Runs in
- LS-DYNA Windows, Linux
- OpenRadioss Windows, Linux
“The model has been calibrated using uniaxial, shear, and plane strain test results. The model has been validated in bending and compression tests.”
How it was testedMaterial, stress states, campaign status
- Material
- A recycled secondary AlSi10MnMg(Fe) — so named because its iron content exceeds the limit for the primary alloy. Characterised at 2 mm wall thickness; bending validation on a 5 mm plate.
- Stress states
- Three, not a continuous sweep: shear, uniaxial through a hole, and plane strain through a notch. Three specimens each.
- Aluminium — completed
- The HPDC aluminium characterisation behind this model is finished. It ran at Luleå University of Technology under FlexCrash, which closed in August 2026.
What this does not establishCalibration per material · computed, not measured · nothing about spot welding
- Calibration is required per material. A model calibrated on this alloy does not transfer to another without new specimens.
- The component runs on this page are computed. They have no measured counterpart, and the spread across them describes material scatter rather than agreement with a physical test.
- Nothing here bears on resistance spot welding, which is a separate simulator with its own evidence.