Wellbore Genius
Single-pane view comparing the live coupledNewtonStep kernel against stored site DFITs, the analytical KGD/PKN/Carter registry benchmarks (leak-off coverage included), and offset-history drift checks. Mirrors the data that powers the merge-blocking CI gate.
2 curves committed to source. Δ ≡ tolerance − rms (positive ⇒ headroom, negative ⇒ regression).
| Site | Kind | rms [log] | max [log] | tol [log] | Δ [log] | n | Status |
|---|---|---|---|---|---|---|---|
KGD canonical (μ=0.1 Pa·s, E'=25 GPa, q/h=1e-3 m²/s) site-kgd-canonical | kgd-width | 7.02e-2 | 7.56e-2 | 1.15e-1 | +4.48e-2 | 2 | PASS |
PKN canonical (h=30 m, μ=0.1 Pa·s, E'=25 GPa, q=0.05 m³/s) site-pkn-canonical | pkn-length | 0.00e+0 | 0.00e+0 | 2.00e-1 | +2.00e-1 | 2 | PASS |
Live kernel residuals against the analytical references on the locked early-time windows. Carter is the leak-off entry; its tolerance was sweep-tightened to 1e-5 log-units and is gated by a dedicated step in
regression-gate.yml.| Benchmark | Runner | rms [log] | max [log] | tol [log] | n | Status |
|---|---|---|---|---|---|---|
kgd-width live-kernel | coupledNewtonStep + MK moving-tip + reinit i=8 (Phase 7 inc 15) | 7.02e-2 | 7.56e-2 | 1.15e-1 | 2 | PASS |
pkn-length live-kernel | coupledNewtonStep (Phase 3.5 kernel, height-contained) | 0.00e+0 | 0.00e+0 | 1.50e-1 | 2 | PASS |
carter-falloff live-kernel | coupledNewtonStep + Carter sink mass-balance recovery (Phase 7 inc 14) | 6.95e-8 | 9.82e-8 | 1.00e-5 | 2 | PASS |
Same canonical KGD inputs run on three scenarios (coarse / refined grid; refined + larger seed offset). Compares terminal-time width, half-length and reinit |∇φ| against the baseline scenario. Tolerances: Δw ≤ 0.0002 m, Δa ≤ 0.05 m, Δ|∇φ| ≤ 0.05.
| Baseline | Variant | t [s] | Δw [m] | Δa [m] | Δ|∇φ| | Status |
|---|---|---|---|---|---|---|
| Coarse grid | Refined grid | 2.00 | 6.67e-4 | 2.22e-16 | 2.22e-16 | FAIL |
| Coarse grid | Refined + larger seed | 2.00 | 6.67e-4 | 2.22e-16 | 2.22e-16 | FAIL |
Pending field-data coverage
Coverage gaps the field-validation report will flip from "pending" to "live" as site campaigns land. Tracked in lockstep with the `leak-off` and `proppant-transport` rows of the solver scoreboard.
- PendingLeak-off — site DFITs with PDL + filter-cake history MethodCarter √t is fully validated (rms ≈ 7.2e-8 on [1, 2] s). The PDL amplification + dynamic filter-cake terms still need field benchmarks per reservoir class before they can be promoted out of beta.
- PendingProppant transport — offset-well placement history MethodSingle-phase advection + Eulerian-Granular two-phase scaffold are wired through the live kernel. Site validation against placement logs / production-attribution offsets is the open item shared with the leak-off row.
CI summary (plain text)
Identical to the strings emitted by
formatSiteValidationReport, formatDfitBenchmarkReport and formatOffsetHistoryReport. Safe to paste into PR comments or incident tickets.DFIT site validation @ 2026-07-29T03:56:23.368Z — PASS (2 sites) [PASS] site-kgd-canonical kind=kgd-width rms=7.02e-2 max=7.56e-2 tol=1.15e-1 Δ=+4.48e-2 n=2 [PASS] site-pkn-canonical kind=pkn-length rms=0.00e+0 max=0.00e+0 tol=2.00e-1 Δ=+2.00e-1 n=2 DFIT registry @ 2026-07-29T03:56:23.368Z — PASS (3 benchmarks) [PASS] kgd-width rms=7.02e-2 max=7.56e-2 tol=1.15e-1 n=2 [PASS] pkn-length rms=0.00e+0 max=0.00e+0 tol=1.50e-1 n=2 [PASS] carter-falloff rms=6.95e-8 max=9.82e-8 tol=1.00e-5 n=2 Offset-history drift: FAIL (tol: Δw≤0.0002m, Δa≤0.05m, Δ|∇φ|≤0.05) Coarse grid vs Refined grid @ t=2s: Δw=6.67e-4m, Δa=2.22e-16m, Δ|∇φ|=2.22e-16 → DRIFT Coarse grid vs Refined + larger seed @ t=2s: Δw=6.67e-4m, Δa=2.22e-16m, Δ|∇φ|=2.22e-16 → DRIFT