Caliber
Calibration campaigns, from test plan to material card
Product overview
A model is only as good as the measurements behind it. Caliber is the campaign that produces them: Aerobase defines the test matrix, has the specimens made and tested at partner facilities, fits the model to the results, and validates it against tests it was not fitted to.
Status
Engineering service · for SafeLight and PHASES subscribers · testing at partner facilities
Evidence boundary
Aerobase designs and interprets the campaign. The physical testing is done at partner facilities — Aerobase does not operate a laboratory of its own.
Capabilities
- Test-matrix design
- Specimen manufacture
- Thermo-mechanical testing
- Digital image correlation
- Dilatometry and metallography
- Parameter identification
- Validation testing
- Traceable material pedigree
Two castings off the same tool do not fail at the same strain, so the model predicts a probability rather than a single threshold.
What the clip shows
The measured probability of failure against equivalent plastic strain, for recycled high-pressure die-cast AlSi10MnMg(Fe). One curve carries the mean and both three-sigma bounds. The bounds are stated as mean ± 3 standard deviations rather than as a percentile, because they are computed from a handful of specimens per stress state and a sample that size cannot support a probability quoted to three decimals. The scatter is not noise — it comes from skin-core gradients with fine-grained surface layers up to about 150 µm, skin that does not form consistently, and casting defects including cold flakes, cold shots, shrinkage pores and laminations. Because the material varies this much, a design is checked against the level of risk it has to tolerate rather than against one number.
Five stress states, three to four tests each. The mean local failure strain against triaxiality, and the ±3σ band the repeats put around it.
What the clip shows
Specimen geometry decides the stress state the material sees: a shear coupon, a plate with a hole, a notched plate. Digital image correlation follows the strain field across each surface until it localises and the specimen fails, and the failure strain is read at that spot, at that instant, over a 0.1 mm length scale. Plotted against stress triaxiality the measured mean is not a tidy falling curve — it drops, rises again at 0.48 and falls sharply at 0.58 — and three to four tests at each state put a band around it, from mean − 3σ to mean + 3σ. That band is the point for recycled steel, where variation in Mn, Si, Cr, Cu, Sn and other residual elements changes the phase transformation and with it the failure strain. The curve is the martensite condition; ferrite/pearlite and bainite are separate curves in the same study. Tests at 0.001 s⁻¹ on hot-stamped steel, under CiSMA.
The same measurement on recycled die-cast aluminium: three stress states, and the band around each.
What the clip shows
The same procedure run on recycled high-pressure die-cast aluminium, at three stress states rather than five. The mean failure strain falls from about 10% in shear to 5% in plane strain, and the bounds — mean ± 3 standard deviations from three to four tests — sit either side of it — widest in shear, where the material has most room to vary. Shown as a percentage: the source axis is labelled dimensionless, which cannot be right at these magnitudes, and the same deck puts the die-cast mean failure strain on a percent scale. Tests at 0.001 s⁻¹, under FlexCrash.
How a campaign runs
Plan. Test. Calibrate. Validate.
Aerobase defines the engineering question and the evidence that would answer it before a single specimen is machined. The physical testing runs at partner facilities — Luleå University of Technology, RISE and Swerim — which Aerobase designs and interprets but does not own.
01
Plan
Fix the engineering question, the material, the measurements and the acceptance evidence — before specimens are ordered or a fitting route is chosen. Deciding afterwards what would have counted as success is how campaigns produce data nobody can use.
02
Test
Machine the specimens and run the agreed programme at partner facilities. Every measurement stays tied to its specimen, its conditions and the question it was taken to answer.
03
Calibrate
Identify the model parameters by inverse solution: a cost function comparing computed response against measurement is minimised, each parameter either free or fixed inside declared bounds. Raw data, processed data and the resulting parameter set are kept together with their pedigree.
04
Validate
Compare the fitted model against tests it was not fitted to, under controlled boundary conditions. Until that has been done, a model is calibrated and not validated — they are different words and only one of them is earned by fitting.
“Sophisticated models need to be calibrated with equally sophisticated measurements.”
What gets measuredDIC · thermo-mechanical · dilatometry · stepwise modelling
- Digital image correlation
- A tensile machine with digital speckle photography measures the strain field across the specimen surface, and the failure strain is read from it directly — the local strain in the element that fails, at the instant it fails. An extensometer averages over a gauge length and cannot see that; failure initiates somewhere specific, and the number the model needs is the one at that spot. The length scale it is measured over is declared, because a failure strain without one is not a number a model can use.
- Gleeble thermo-mechanical testing
- Stress–strain curves from room temperature to above 1100 °C, at strain rates up to 50 per second, for models that have to hold across a forming or heat-treatment cycle. Also used to reproduce a thermal cycle on a specimen so a validation test can be run against it.
- Dilatometry and metallography
- Transformation kinetics are measured by dilatometry and calorimetry, combined with metallography to check the result against what the microstructure actually did. Dilatometry is named separately from the Gleeble work on purpose: much of this data comes from a deformation dilatometer, which is a different instrument.
- Stepwise Modelling Method
- Failure initiation is calibrated with the Stepwise Modelling Method, a Luleå University of Technology method that Aerobase uses and did not invent, from specimens loaded at several stress states with DIC following each one to failure.
What this does not establishPartner facilities, not our lab · per material · fitted is not validated
- Aerobase designs and interprets the campaign. The physical testing is done at partner facilities, and Aerobase does not operate a laboratory of its own.
- A campaign result belongs to the material, the model and the measurements defined for it. It does not transfer to another material without new specimens.
- Fitting a model to measurements makes it calibrated. It becomes validated only once it has been compared against tests it was not fitted to.
Sources
- “Crashworthiness Under Material Variability: Probabilistic Modelling of Recycled Aluminium HPDC & Recycled Steels”, Aerobase Innovations AB — CHS² 2026, 10th International Conference on Hot Sheet Metal Forming of High-Performance Steel, Plzeň, 1–3 June 2026. The source for all three clips.
- FlexCrash — Horizon Europe grant agreement 101069674. The recycled HPDC aluminium work.
- CiSMA — Horizon Europe grant agreement 101177798. The recycled electric-arc-furnace steel work.