Engineering tools,
ready to use.

Supervised learning and clustering with parameters you choose, scored the way a held-out well would score them. Linear, regularised, tree-ensemble and neural-network families; k-means, four agglomerative linkages and DBSCAN with resampling stability. Every leaderboard is built with nested cross-validation, so the reported figure is not the maximum of the scores that chose the model — the naive figure is shown beside it, and the gap between them is labelled. Forecasts must beat an Arps decline baseline before they are reported as results, and piecewise-constant models are refused for extrapolation because beyond the training range their prediction is a horizontal line. An interpretation layer states plainly whether the result is strong enough to act on; the written summary is checked so that every figure in the prose traces back to a computed finding.

Arps DCA — exponential, hyperbolic, harmonic, and linear decline. Input production history, get b-factor, Di, and EUR automatically. Filter by well or field.

Field-level gas & oil well surveillance dashboard. For gas wells: Coleman-Turner critical rate, Arps DCA, liquid loading detection, intermittent producer detection. For oil wells: water cut & WOR trends, Reciprocal Rate Method & Ershaghi X-plot for reserves/EUR, Chan diagnostic (coning vs channeling auto-classification). Monitors well health states — Stable, Warning, Critical, Shut-in — with intervention recommendations and autonomous AI alerts. Password-protected.

Three coupled physics layers in one canvas: reservoir material balance (gas p/Z multi-tank, or oil by the generalised Schilthuis equation with solution-gas drive and Corey relative permeability), well inflow (Darcy + Nodal calibrated from multi-rate tests; Vogel/composite and Beggs–Brill lift for oil), and surface gathering network (Weymouth pipes + Cullender-Smith VLP propagating backpressure from sales meter to wellhead, with manifolds and station or wellhead compression). Build the field topology by dragging components onto the canvas; upload real history CSVs to calibrate each component; forecast the whole system end-to-end. No empirical decline curves — every rate is physics. 100% client-side, sub-second forecast for 50-well networks.

Two balances in one tool, selected by fluid system. Gas: upload pressure–production history, fit P/Z, auto-detect the drive mechanism, and project OGIP and recovery factor at your abandonment pressure — with Cole plot and Fetkovich aquifer auto-fit for water drive, Roach correction for geopressured / HPHT cases, and Wichert–Aziz handling for sour gas. Oil: Havlena–Odeh straight line through the origin for OOIP, gas-cap sizing against an independent volumetric N, Campbell plot to separate water influx from an in-place error, and a four-way drive index split. An energy plot on both paths tracks how the drive mechanism shifts across the production history. Published worked examples ship loaded, so the answers can be checked rather than believed. All client-side — no data leaves your browser.

Complete production system analysis: IPR (Linear, Vogel, Fetkovich, Backpressure) meets VLP (Beggs-Brill, Hagedorn-Brown) at the operating point. Full PVT correlations (Standing, Dranchuk-Abou-Kassem, Lee-Gonzalez-Eakin), choke models (Gilbert, Achong), pressure traverse, and sensitivity sweeps across tubing ID, wellhead pressure, choke size, reservoir pressure, water cut, and GOR. Diagnose whether the well is IPR-dominated, VLP-dominated, or balanced. New — injection mode (water or gas): injectivity vs tubing-intake operating point, with an explicit injectivity index, fracture-pressure limit, and maximum matrix-injection rate. 100% client-side.

Full buildup/well-test interpretation from raw high-resolution gauge data: automatic QC (spikes, gaps, frozen intervals), rate-schedule clock reconciliation, Bourdet derivative with superposition time, flow-regime identification with slope guides, auto-window Horner/superposition semilog, and matching against a 15-model analytical catalog (dual porosity, faults, channels, U-shaped and closed boundaries, radial composite, fractures) ranked by AICc with explicit non-uniqueness warnings and parameter uncertainty. LIT deliverability, AOF, IPR/VLP operating point, and volumetrics close the loop. Gas (real-gas pseudopressure) and oil. Validated against classic published worked examples. 100% client-side computation — test data never leaves the browser. Password-protected (shared sign-in with GWIS / Nodal).

Upload production time-series (CSV or Excel) and the tool recovers contacted in-place plus permeability and skin — all from flowing data. Flowing material balance returns OGIP/OOIP across gas, oil and gas-oil (solution-gas-drive two-phase MBE) systems; Blasingame split uses the transient infinite-acting period for k and boundary-dominated flow for skin, with radial superposition time so variable-rate data analyses correctly. Real-gas pseudo-pressure m(p) and pseudo-time throughout; honest non-convergence when the depletion signal is too weak rather than a fabricated number. 100% client-side.

Six-method rock typing in one dashboard: Winland R35, FZI / HFU (Amaefule), PGS (Permadi–Susilo / Wibowo–Permadi / Hakiki), Modified Lorenz Plot (Gunter), Leverett J-function, and ML log-based classifier (kNN with leave-one-out CV). Upload core data, get pore-type populations, heterogeneity scores, flow-unit detection per well, and AI-generated interpretation narratives. Each method ships with an expert guide modal explaining theory, equations, and trade-offs. 100% client-side computation. Companion tool to the Rock Typing Reimagined blog article. Password-protected (shared sign-in with GWIS / Nodal).

Systematic Enhanced Oil Recovery candidate selection using established screening criteria from Taber, Martin & Seright (1983) SPE-12069, NPC (1984), and Green & Willhite (1998) SPE Textbook Vol. 6. Eight EOR methods scored against eight reservoir parameters — CO₂ Miscible, N₂ Miscible, HC Miscible, WAG, Steam Flooding, In-Situ Combustion, Polymer Flooding, Surfactant/ASP. Each parameter scored against published acceptable and ideal ranges. Disqualification logic auto-applied when any parameter falls outside absolute bounds. Expandable parameter breakdown per method with individual scores. 100% client-side.

Probabilistic OGIP/OOIP estimation with Monte Carlo simulation. Define input parameter distributions (Triangular, Normal, Lognormal, Uniform) for area, net pay, porosity, water saturation, formation volume factor, and recovery factor — then get P10/P50/P90 reserves with histogram, CDF, and tornado sensitivity. Supports multi-zone aggregation (independent LHS sampling per zone). 10,000 trials by default, configurable up to 100k. 100% client-side, deterministic seeding for reproducibility.