Wellbore Genius

Stage-parallel solver vstage-parallel/1.0.0

Whole-pad parallel solve

Every stage solves at the same time against a predicted stress-shadow field, then the field is rebuilt and the pad re-solves. A pad converges in a few waves instead of one serial step per stage — so you run all 20 stages together instead of chunking them into fives.

Pad

Tighter spacing and taller fractures couple the stages harder, which costs more waves.

Stages: 20

Stage spacing: 200 ft

Fracture height: 200 ft

Zipper pairs (two stages pumped together)

Field-calibrated physics (volume balance, leak-off, height growth)

Using the constants fitted to HFTS-1 and MSEEL. Tune the constants

Cold start (skip the analytic warm start)

Instant pad preview

Preview (approximate)

CPU

—

One shadow field, one relaxation wave — updates as you move the sliders. Approximate by design: no equivalence check runs here, so these bars are never a result you can sign off.

Preview compute

CPU paints instantly, a WebGPU pass replaces it when ready. Best perceived speed.

Checking for a WebGPU adapter — running on CPU until it answers.

The verified run below already matches these settings.

Verified physics tiers

Analytic depletion

Coupling: Reference (fast)

Choose the compute budget explicitly. The verified solve, exports and cross-pad fingerprints carry these tier names, so a fast preview can never masquerade as the high-fidelity result.

Depletion field

Steady stress-relief superposition from the stored drawdown bubbles. Instant and byte-identical to prior releases.

Stage ↔ stress coupling

Self-consistent coupling

Off: Reference (fast). On: Self-consistent (verified).

Anchors each rate partition to the shadow-free net pressure. One geometry solve per stage per wave.

Fiber forward model

Fast analytic dipole + poroelastic heartbeat. The whole fiber responds instantly; use for layout and quick checks.

Causality gate: P arrival at 100 ft is 6.7 ms; modeled elastic energy before that time is exactly zero.

Compare Instant vs Verified

Runs Reference (fast) and Self-consistent (verified) on identical inputs, then shows signed output and convergence deltas. Signed deltas are Verified minus Instant.

Receipt stamps: depletion_fidelity=analytic, stage_coupling_mode=reference, fiber_forward_model=quasi-static

Fracture fidelity tier comparison

Tiers diverge · worst +43.2%

Complex recommended

The same scenario solved on both tiers. The complex tier is an operator-learning surrogate over phase-field reference solutions — instant, and it redistributes fracture volume rather than inventing it. Where the two columns match, the planar answer is the honest one to ship.

Azimuth off σHmax (°)

Bounding contrast (psi)

Expected Pnet (psi)

Cluster spacing (ft)

Target half-height (ft)

Parent depletion (psi)

Escalation score100%

Planar/elliptical versus phase-field-complex tier geometry
Quantity	Planar	Complex	Δ
Half-length (ft)	640	622	-2.8%
Height (ft)	180	258	+43.2%
Avg width (in)	0.140	0.098	-30.0%
Cluster efficiency	0.82	0.49	-39.8%

Front turning

0.9°

Competing strands

2.77

Growth asymmetry

Height moves +43.2% on the complex tier (driven by swarm interaction).

Swarm interaction (100%) Cluster spacing 45 ft vs 90 ft half-height (ratio 0.50). Under ~1× the stress shadows overlap and fractures branch/merge — a single-wing model averages away the physics that decides cluster efficiency.
Height containment (41%) Stress contrast 500 psi vs 420 psi net (ratio 1.19). Below ~1.5× the planar model's imposed height is not honest — height growth must be solved, not assumed.
Depletion skew (12%) Nearest parent is 400 psi depleted. Growth biases toward the drained rock — symmetric wings overstate the child-side half.
Fracture turning (7%) Well azimuth is 12° off σHmax. Beyond ~15° the front curves out of plane — the elliptical-wing model cannot represent the turn.

Last run receipt

converged

0 of 3 modules on a verified tier · solver stage-parallel/1.0.0 · 20 stages · 2 waves

Stage coupling
Reference (fast)
Fast
Depletion fidelity
Not coupled — no parent wells on record
Off
Fiber forward model
Quasi-static
Fast

# solver: stage-parallel/1.0.0
# verdict: converged
# stage_coupling_mode: reference
# depletion_fidelity: analytic
# fiber_forward_model: quasi-static
# fiber_causality: instant-quasi-static

Tier notes · Manual §30 Receipt conventions

Refrac / parent-child depletion

Depleted parents on other pads lower closure stress, so a child or refrac stage grows wider and longer. This state persists between runs and its fingerprint is stamped into every export.

Couple depletion

dep_1b2e9ff1

No depleted parents on record· peak Δσ at a stage 0 psi

No depleted parents recorded — the pad solves as virgin rock.

Use this when the assumptions change — new parent survey, revised drawdown, or a different acreage position.

Analytic depletion

Carries the wells, drawdown, drainage and ramp inputs behind the fingerprint, so a coupled result can be rebuilt elsewhere.

Depletion run history

No coupled runs recorded on this pad yet. Every verified solve and every reset is logged here with the assumptions it used.

Re-running restores the parent wells, drawdown, and drainage exactly as that run had them — including after a reset — so a prior coupled result can be reproduced.

Loading depletion diagnostics…

Separate-pad depletion is excluded from this solve
none

No separate-pad depletion on record — results are virgin-rock.

Preview bars and the verified solve agree on depletion: there is nothing to include.

20-stage pad, 2 waves — 6.7× faster (modelled)

Wall-clock steps

serial chain needs 20

Speedup vs serial

6.7×

modelled, verified equivalent

Measured wall time

—

measured after hydration

Waves

37 stage solves queued

Worst half-length gap

0.15%

1.1 ft vs serial

Equivalence gate

Verified equivalent to the serial run — 20 stages in 2 waves, 6.7x faster.

verified-equivalent

coupling mode: reference

residual monotone: yes

order independent: yes

run verdict: converged

Stage coupling convergence

Reference (fast)

Per-stage record for the selected coupling tier. Residuals are net-pressure movement between internal fixed-point passes; derate is the final rate partition used by the verified result.

1 pass / stage

20 stages

Reference mode anchors each stage to the fixed shadow-free net pressure, so there is no internal residual trail. One pass and the final derate are still shown here for comparison with the self-consistent tier.

Per-stage stage-coupling convergence record for the selected mode
Stage	Status	Iterations	Final derate	Final residual (psi)	Residual trail (psi)
S1	single-pass	1	1.0000	—	Not applicable — fixed shadow-free anchor
S2	single-pass	1	0.9359	—	Not applicable — fixed shadow-free anchor
S3	single-pass	1	0.9155	—	Not applicable — fixed shadow-free anchor
S4	single-pass	1	0.9088	—	Not applicable — fixed shadow-free anchor
S5	single-pass	1	0.9040	—	Not applicable — fixed shadow-free anchor
S6	single-pass	1	0.9009	—	Not applicable — fixed shadow-free anchor
S7	single-pass	1	0.8987	—	Not applicable — fixed shadow-free anchor
S8	single-pass	1	0.8971	—	Not applicable — fixed shadow-free anchor
S9	single-pass	1	0.8959	—	Not applicable — fixed shadow-free anchor
S10	single-pass	1	0.8949	—	Not applicable — fixed shadow-free anchor
S11	single-pass	1	0.8941	—	Not applicable — fixed shadow-free anchor
S12	single-pass	1	0.8934	—	Not applicable — fixed shadow-free anchor
S13	single-pass	1	0.8929	—	Not applicable — fixed shadow-free anchor
S14	single-pass	1	0.8924	—	Not applicable — fixed shadow-free anchor
S15	single-pass	1	0.8920	—	Not applicable — fixed shadow-free anchor
S16	single-pass	1	0.8916	—	Not applicable — fixed shadow-free anchor
S17	single-pass	1	0.8913	—	Not applicable — fixed shadow-free anchor
S18	single-pass	1	0.8910	—	Not applicable — fixed shadow-free anchor
S19	single-pass	1	0.8908	—	Not applicable — fixed shadow-free anchor
S20	single-pass	1	0.8906	—	Not applicable — fixed shadow-free anchor

Wave trace

Each wave re-solves only the stages still moving; the freeze front sweeps down the pad as the interference field settles.

Residual and freeze progress for each relaxation wave
Wave	Stages solved	Frozen	Max residual (psi)	L2 residual (psi)
0	20	3 / 20	0.613	2.204
1	17	20 / 20	0.166	0.198

Stages

Parallel result beside the serial reference, stage by stage.

Per-stage parallel versus serial results
Stage	Δσ (psi)	Rate derate	Half-length (ft)	vs serial (ft)	Within tol.
S1	0.0	1.000	778	0.0	yes
S2	5.3	0.936	738	0.0	yes
S3	7.1	0.915	726	-1.1	yes
S4	7.7	0.909	722	0.0	yes
S5	8.2	0.904	719	0.1	yes
S6	8.5	0.901	717	0.1	yes
S7	8.7	0.899	715	0.1	yes
S8	8.8	0.897	714	0.1	yes
S9	8.9	0.896	713	0.1	yes
S10	9.0	0.895	713	0.2	yes
S11	9.1	0.894	712	0.2	yes
S12	9.2	0.893	712	0.2	yes
S13	9.2	0.893	712	0.2	yes
S14	9.3	0.892	711	0.2	yes
S15	9.3	0.892	711	0.2	yes
S16	9.3	0.892	711	0.2	yes
S17	9.4	0.891	711	0.2	yes
S18	9.4	0.891	710	0.2	yes
S19	9.4	0.891	710	0.2	yes
S20	9.4	0.891	710	0.2	yes

Run this pad on the fleet

Each wave becomes a block of queue tasks — only the stages still moving are queued, so a frozen stage never costs a second solve. Same lanes, leases, retries and live progress as any other batch.

2 wave(s)

37 stage runs

est. 1.85 credits

serial chain: 400 steps

2 wave(s) · 37 stage runs · est. 1.85 credits · deepest dependency chain 2 vs 400 sequential serial steps

Receipts

Every export carries the solver version, the equivalence verdict and the fingerprint divergence override (if one is in force), so a reviewer can tell whether the fast run was trustworthy and comparable.

Depletion run history

Separate-pad depletion is excluded from this solvenone

Separate-pad depletion is excluded from this solve
none