Verification & Validation
The Op3 V&V suite contains 140 active falsification gates
across 15 test modules. All tests must pass before any commit reaches
main.
Test summary
Module |
Tests |
Topic |
|---|---|---|
|
4 |
Analytical closed-form references (Euler-Bernoulli, Rayleigh) |
|
4 |
Cross-path internal consistency |
|
5 |
Physical sensitivity invariants |
|
12 |
Damping, energy, reciprocity, units, coordinates, orthogonality |
|
9 |
PISA conic + 6x6 stiffness invariants |
|
10 |
Hardin-Drnevich + Vucetic-Dobry |
|
8 |
HSsmall constitutive wrapper + CSV round-trip |
|
8 |
Mode D dissipation-weighted formulation |
|
6 |
OpenFAST runner infrastructure |
|
3 |
Static condensation, SACS round-trip, Mode B/C |
|
13 |
MC propagation, PCE, Bayesian |
|
15 |
Framework integration, project I/O, visualization wiring |
|
6 |
Snapshot byte-identical reproduction |
|
4 |
Published-source frequency calibration |
|
33 |
Eigenvalue + pushover + transient on 11 examples |
Calibration regression
Example |
Op3 f1 (Hz) |
Reference (Hz) |
Source |
Error |
|---|---|---|---|---|
NREL 5 MW fixed |
0.316 |
0.324 |
Jonkman 2009 NREL/TP-500-38060 |
-2.5% |
NREL 5 MW OC3 monopile |
0.275 |
0.277 |
Jonkman & Musial 2010 NREL/TP-500-47535 |
-0.4% |
SiteA 4 MW class tripod |
0.235 |
0.244 |
PhD Ch. 5 field OMA |
-3.7% |
IEA 15 MW monopile |
0.188 |
0.170 |
Gaertner 2020 NREL/TP-5000-75698 |
+10.6% |
Standards conformance
DNV-ST-0126 audit (9 clauses x 4 examples = 36 checks): 35/36 PASS. The single failure is the SiteA tripod 1P frequency separation (6.8% < 10% required), which is a real engineering finding.
IEC 61400-3 audit (12 clauses x 4 examples = 48 checks): structural and foundation provisions all PASS; DLC coverage is partial as documented in the OpenFAST runner section.
Cross-validation against published benchmarks
In addition to the internal V&V tests above, Op3 has been cross-validated against 39 independent benchmarks from 20+ published sources, achieving a 92% verification rate (35/38 in-scope).
See Cross-Validation Against Published Benchmarks for the complete cross-validation report including OptumGX bearing capacity factors (0.8–7.8% error vs Vulpe 2015), foundation stiffness (3–26% vs Doherty 2005 / OxCaisson), and the Bothkennar field trial comparison (-21% vs measured Kr).