State
packs/vv40/examples/nagaraja/ contains one encoded package, cou1/uofa-nagaraja-cou1.jsonld. Morrison ships cou1/ and cou2/, and the divergence between them is described in docs/architecture.md as "the central analytical demonstration".
Why this is now worth doing
The OSF evidence folder (https://osf.io/n4pjz/) turns out to contain exactly the pair that would support a second COU, which was not obvious before the archives were opened:
Ansys_FDAF1717NonCannulated-NoResults.wbpz — the validated constructAnsys_FDAF1717Cannulated-NoResults.wbpz — the cannulated design
That mirrors the paper's own structure: the model is validated against bench testing on the non-cannulated construct, then applied to the cannulated design without new experimental testing. That is a genuine change in context of use with a genuine change in evidentiary support — the same shape as Morrison's CPB/VAD split, but on real vendor artifacts rather than a narrative.
The two archives are also demonstrably distinct rather than near-copies: their .mechdb digests differ, and one carries DesignModeler geometry (.agdb) where the other carries SpaceClaim (.scdoc).
Suggested next step
Encode packs/vv40/examples/nagaraja/cou2/ for the cannulated construct, drawing the risk framing from the paper's own account of extending a validated model to an untested design. Then check whether the weakener profile diverges from cou1 the way Morrison's does; if it does not, that is worth knowing too, and is a cheaper result than it looks.
Worth pairing with snapshots/example-counts.json, which does not currently cover Nagaraja at all (filed separately).
Discovered during
PR #85 (feat(evidence): seal and read Ansys solver artifacts without the solver), while inventorying the three OSF archives. Deliberately out of that PR's scope, which stops at reading and sealing the artifacts.
State
packs/vv40/examples/nagaraja/contains one encoded package,cou1/uofa-nagaraja-cou1.jsonld. Morrison shipscou1/andcou2/, and the divergence between them is described indocs/architecture.mdas "the central analytical demonstration".Why this is now worth doing
The OSF evidence folder (https://osf.io/n4pjz/) turns out to contain exactly the pair that would support a second COU, which was not obvious before the archives were opened:
Ansys_FDAF1717NonCannulated-NoResults.wbpz— the validated constructAnsys_FDAF1717Cannulated-NoResults.wbpz— the cannulated designThat mirrors the paper's own structure: the model is validated against bench testing on the non-cannulated construct, then applied to the cannulated design without new experimental testing. That is a genuine change in context of use with a genuine change in evidentiary support — the same shape as Morrison's CPB/VAD split, but on real vendor artifacts rather than a narrative.
The two archives are also demonstrably distinct rather than near-copies: their
.mechdbdigests differ, and one carries DesignModeler geometry (.agdb) where the other carries SpaceClaim (.scdoc).Suggested next step
Encode
packs/vv40/examples/nagaraja/cou2/for the cannulated construct, drawing the risk framing from the paper's own account of extending a validated model to an untested design. Then check whether the weakener profile diverges from cou1 the way Morrison's does; if it does not, that is worth knowing too, and is a cheaper result than it looks.Worth pairing with
snapshots/example-counts.json, which does not currently cover Nagaraja at all (filed separately).Discovered during
PR #85 (
feat(evidence): seal and read Ansys solver artifacts without the solver), while inventorying the three OSF archives. Deliberately out of that PR's scope, which stops at reading and sealing the artifacts.