ข้ามสารบัญไปยังเนื้อหา
อัปเดต 2026-09-07
Public validation dossier (VC-001..073+)
EPANET · FluidFlow cross-checks
A published, growing record of 70+ numbered verification cases (VC-001 through VC-073 and counting) cross-checking DekEn's engines against EPANET, FluidFlow, and independent closed-form/exact references — so a claim of accuracy points at a specific, inspectable case rather than an assertion.
เมื่อไรจึงใช้
- You want to see proof a specific engine (gas transient, slurry, marching two-phase, column separation) was actually checked against an external reference, not just unit-tested internally.
- You're citing DekEn's credibility to a client or reviewer and need a traceable case number, not 'we tested it'.
- You're adding new physics and need the pattern for anchoring it — every new engine capability picks up its own dossier entry.
คำนวณอย่างไร
Each case pairs a DekEn solve with an independent oracle — published exact solutions, a second numerical implementation with zero shared code (self-validated separately), or a real reference tool (EPANET/FluidFlow) — and reports the measured discrepancy against an a-priori error band set BEFORE the check runs.
ข้อควรระวัง
noteAn a-priori error band is set before the comparison is run, specifically so a band can't be widened after the fact to make a borderline result 'pass' — treat any case whose band was chosen after seeing the result as suspect.
noteIndependent oracles are audited too: building one new-physics oracle caught a wrong grouping in the reference paper's OWN printed equation — cross-checking runs both directions, the external reference isn't automatically ground truth either.
ที่เกี่ยวข้อง