Well data management
Import and manage LAS 2.0/3.0 and CSV datasets with full curve metadata and unit validation, organised into projects by field, basin and operator.
Petrophysics · Intelligence · Insight
PetroARX
Interpretation Platform
A desktop platform for well-log interpretation, built in Python and PyQt5. It takes a LAS file from raw curves through quality control, formation evaluation and machine-learning analytics to a defensible net-pay answer — in one workspace.
10+ modules
LAS 2.0 / 3.0 · CSV
Multi-well
AI analytics
3D wellbore
13
Interpretation tabs, dashboard to ELAN
150+
Public data sources in the downloader
92.4%
Facies classification accuracy, 6 classes
3
Operating systems from one codebase
Walkthrough
See a full interpretation run
A recorded session through the platform — loading a LAS file, running quality control, computing shale volume, porosity and saturation, then reading the net-pay result.
Walkthrough recording coming soon
The screen recording is still being produced. Every screen it will cover is already captured in the interface gallery below.
Browse the interface instead
Drop the recording at assets/video/petroarx-demo.mp4 and this player picks it up
automatically — no other change needed.
Capabilities
What the platform does
Ten core modules, reachable from a single project workspace. Curves loaded once are available to every calculation downstream.
Quality control engine
Automated QC for missing values, outliers and spikes using IQR, Z-score and Isolation Forest, with before/after comparison and a depth-indexed issues register.
Multi-track log viewer
Configurable depth-track plots, triple-combo presets, crossplots, histograms, pair plots and violin plots, all synchronised to the active depth range.
3D wellbore visualisation
Trajectories rendered as colour tubes with log ribbons, formation tops and perforations, under full camera, lighting and cross-section control.
Petrophysical modules
Integrated shale volume, porosity, water saturation, permeability and net-pay workflows, each with its own parameter panel and result summary.
AI analytics
Facies classification, missing-log prediction and ML-driven quality control, trained inside the application with cross-validation and uncertainty reporting.
Advanced interpretation
Pore-pressure prediction, geomechanics analysis and image-log processing, plus multi-mineral ELAN inversion with a least-squares solver.
Open data downloader
A curated, searchable index of 150+ public seismic, well-log, geological, satellite and research datasets, filterable by category, access type and region.
Embedded Python console
The active well is exposed as a DataFrame so any workflow the interface does not cover can be scripted in place and run against live data.
Reporting and export
Export cleaned curves, computed logs, QC registers and figures as PDF, PNG, SVG or CSV, with an export-report action on every analysis tab.
Workflow
From raw LAS to net pay
The tab order mirrors the interpretation sequence, so the workspace itself carries the method.
IngestLAS 2.0/3.0 and CSV, with unit validation
InspectCoverage, null fraction, per-curve statistics
Quality controlStatistical and Isolation Forest passes
VisualiseMulti-track, crossplots, 3D trajectory
Shale volumeLarionov, Clavier, Steiber or linear
PorosityDensity, neutron and sonic, shale corrected
SaturationArchie from Rt and PHIE
Net payCutoffs, pay flags and zone summaries
Interface
Every module, as it ships
Screens captured from PetroARX v1.0 running a real supercombo log suite over a 3,918–4,146 m interval. Select any screen to enlarge it.
Project workspace
Project dashboard
Active well, depth range, data-integrity score and a dynamic log-availability radar, with quick actions into every major module.
Project setup
New project
Project identity is captured first — field, basin, country and operator — so every well imported afterwards inherits that context.
Dataset
Dataset overview and statistics
Depth range, sample count, curve count and average null percentage, backed by per-curve min, max, mean and standard deviation plus a coverage matrix.
Dataset
Well data table and distributions
The full curve table with column rename and undo, alongside histogram and boxplot views of whichever curve is selected.
Open data
Open data downloader
A searchable index of 150+ verified public sources across seismic, well log, geological, satellite and research categories, filterable by access type and region.
Quality control
Statistical quality control
Missing values, outliers, spikes and negatives are flagged per curve, with before-and-after histograms, boxplots and a depth-indexed issues table.
Quality control
Machine-learning quality control
Isolation Forest anomaly detection across multi-curve features, reporting an anomaly score against depth, per-feature contributions and an exportable anomaly register.
Log display
Multi-track log plot
Caliper, gamma ray, density, resistivity, neutron and magnetic resonance curves on synchronised depth tracks, from triple-combo or custom presets.
3D
3D well trajectory viewer
The deviation path rendered as a colour tube carrying a gamma ray ribbon, with formation tops, a depth scale bar and full camera and lighting control.
Crossplots
Advanced crossplot studio
Density–neutron, Pickett, M-N and Buckles templates with sandstone, limestone and dolomite lines, gas-effect shading and a gamma ray colour axis.
Crossplots
Pair plot and violin matrix
Cross-curve relationships across the whole log suite in one matrix, with histogram and violin alternatives for single-curve distributions.
Formation evaluation
Shale volume
Sand and shale lines picked automatically by percentile, from the histogram peak or by hand, then transformed and summarised as net sand against high shale.
Formation evaluation
Shale volume model comparison
Linear, Larionov Tertiary, Larionov Older, Clavier and Steiber overlaid on the same interval, so the choice of transform becomes an evidenced decision rather than a default.
Formation evaluation
Porosity calculation
Density, neutron and sonic porosity derived from matrix and fluid properties, with optional shale correction and a combined effective-porosity track.
Formation evaluation
Water saturation
Archie saturation from resistivity and effective porosity with tortuosity, cementation and saturation exponents, plus a data-quality panel that states its own assumptions.
Formation evaluation
Formation evaluation and net pay
Gamma ray, shale volume, porosity and saturation computed together over the chosen interval, resolved into a pay-zone summary table and a net-pay flag overview.
Machine learning
Facies classification
A Random Forest, XGBoost and SVM ensemble across six facies classes, with configurable train/test splits, balanced class weights, early stopping and a live training log.
Machine learning
Missing log prediction
An XGBoost regressor trained on complete wells reconstructs absent curves, reporting actual against predicted tracks, residual analysis and a 95% uncertainty band.
Extensibility
Embedded Python console
The active well is exposed as a DataFrame and a data service, so bespoke calculations run in place against live project data.
Documentation
In-app information center
Workflow guidance, technology notes and keyboard shortcuts, versioned against the build so the documentation cannot drift from the release.
Reference dataset: a nine-curve supercombo suite (CALI, GR, RHOB, RT, TNPH, NPHI, TCMR, CMFF) over 1,495 samples between 3,918.05 m and 4,145.74 m.
AI analytics
Machine learning, inside the interpretation loop
Models are trained in the application against the loaded project, not bolted on afterwards — so predictions, their cross-validated scores and their uncertainty all travel with the well.
Facies classification
A Random Forest, XGBoost and SVM ensemble assigns electrofacies across six classes, with balanced class weights, early stopping and per-fold validation scoring.
92.4%Accuracy
0.91F1 score
6Facies classes
Ensemble (RF + XGBoost + SVM) · 4,251 samples · 80/20 split
Missing log prediction
An XGBoost regressor trained on complete wells reconstructs absent curves, returning actual against predicted tracks, residual diagnostics and a 95% confidence band.
5Training wells
15,368Training samples
5-foldCross-validation
XGBoost Regressor · 7 features · 95% prediction interval
Anomaly-based log QC
Isolation Forest scores every sample against multi-curve features, ranking anomalies by depth and reporting which curves drove each detection.
95%QC score
69Anomalies flagged
5%Contamination
Isolation Forest · multi-curve features · CSV export
Engineering
How it is built
A modular PyQt5 application over the scientific Python stack, packaged to run identically on Windows, Linux and macOS.
Architecture
- GUI frameworkPyQt5
- Layout toolingQt Designer
- PatternMVC, modular
- Scripting surfaceEmbedded Python
- PlatformsWindows · Linux · macOS
Data layer
- LAS parsinglasio
- Tabular datapandas
- NumericsNumPy
- StatisticsSciPy
- FormatsLAS 2.0/3.0 · CSV · XLSX
Visualisation
- Log tracksMatplotlib
- Statistical plotsseaborn
- Interactive chartsPlotly
- 3D renderingMatplotlib 3D, Phong
- Figure exportPNG · PDF · SVG
Machine learning
- Core libraryscikit-learn
- Gradient boostingXGBoost
- Anomaly detectionIsolation Forest
- ValidationK-fold cross-validation
- Inversion solverLeast-squares (NNLS)
Explore PetroARX
Read the user guide, browse the development repository, or get in touch for a walkthrough of the platform against your own log suite.