01 / EXPLOREEDA
Understand the data
Explore distributions, outliers, domains and spatial relationships while checking drillhole integrity, assays and sample support.
QA/QC • Statistics • Spatial EDAView detailsThe Earth tells a story. We help you read it
Turning geological complexity into clearer decisions
OBK connects data exploration, neural geological modeling, resource estimation and classification in one software platform—then extends that geological understanding into geometallurgy and geotechnics.
From observations to orebody knowledge
OBK / WORKSPACE
REPORT OUTPUT
A central Z section from the geological model report. Switch plane and diagnostic below.
The connected platform
One geological thread, from data to decision
Every stage carries context forward. OBK brings the work of understanding, modeling and evaluating a deposit into a coherent spatial framework.
02 / MODELML
Model the geology
Learn geological domains from interval data, chemistry and local structural fields, then predict directly onto the block model.
Domains • Continuity • ANISOView details03 / ESTIMATEGRADES
Estimate the resource
Carry geological domains and directional continuity into block-scale grade estimation, with traceable assumptions and diagnostics.
Domaining • Support • DiagnosticsView details04 / CLASSIFYCONFIDENCE
Explain confidence
Bring drillhole spacing, geological continuity and estimation quality into a reviewable resource classification, with simulation-based uncertainty where available.
Evidence • Criteria • AuditabilityView details05 / SIMULATETBS
Explore possible orebodies
Simulate multiple lithology and grade realizations conditioned to drillhole data, then compare spatial outcomes.
Lithology • Grades • TBSView details06 / DHSADRILLING
Plan the drilling
Test drillhole spacing against the acceptable uncertainty for current and expected production.
Spacing • Production • RiskView detailsInside the geological model report
Same data. Deeper insight
Explore assay distributions from the supplied neural geological model report alongside clearly labeled previews of the statistical analyses in OBK.
67,132valid FEGL assays
24sample categories in FEGL
6 viewsEDA displays and methods
01 / EXPLORE · EDA STUDIOOBSERVED DATA
Report dataColor reveals the domains behind the distribution
67,132 valid FEGL values · category colors follow the reportFEGL counts from the geological model report, with sample categories distinguished by color and a median marker at 45.10.
Analysis with visible limits
The histograms and box plots use supplied report data. PCA, clustering and regression panels are concept previews; those results are not computed from this report.
02 / MODEL · ANISO
Geology lives in continuity
ANISO is OBK's internal dynamic anisotropy engine. It uses the structural orientation supplied for each valid block to steer geological continuity locally, while honoring fault boundaries and compatible geochemical signatures. That direction field carries structural context into the block-scale modeling workflow.
The supplied 3D example shows folded layers, mineralized zones and drillhole traces. Switch to the directional concept to see how local orientation follows the structure. Dip and dip direction are not invented where structural support is missing.
Internal OBK engineANISO / FOLDED GEOLOGICAL CONTINUITYOBK-supplied visual reference
↗ Directional continuity⌁ Local structural change◫ Block-level orientation
02 / MODEL · THE CONNECTION
From structural context to block predictions
ANISO describes how geological orientation changes across the deposit where structural observations support it. The modeling workflow combines that local context with logged drillhole intervals, chemistry and spatial position. This helps the neural model learn patterns within a changing geological framework.
At block centers, the network can produce categorical probabilities for lithology or domains and continuous values for grades or other properties. Those predictions then need geological review in three dimensions and testing on drillholes held out from training. The neural explainer and X, Y and Z sections below show how to inspect the result.
02 / MODEL · NEURAL NETWORK
One network. Different kinds of answers
The network learns relationships in labeled intervals, chemistry, position and supported directional features, then predicts at block centers. Switch between categorical and continuous targets to see both kinds of output.
Prediction target
01 / OBSERVATIONSDrillhole intervals
Logged domains, assays, position and supported structural inputs.
Labeled examples + spatial features02 / LEARN + INFER
Connections illustrate learned relationships; their movement is a teaching aid, not actual model weights.
03 / BLOCK OUTPUTSPredictions at block centersPredicted numeric values
Categorical variable Domains / lithology
Hematitite72%
Siliceous Itabirite19%
Goethitic Itabirite9%
Class label + probabilitiesContinuous variables Assays / properties
Fe58.4%
Density3.42t/m³
Conceptual values for illustration; accuracy must be assessed on held-out data.
Multitask mode combines categorical probabilities and continuous values when configured for both targets.
TRAIN ON OBSERVATIONS → PREDICT BLOCKS02 / MODEL
Read the orebody from every direction
These central X, Y and Z sections come from the supplied neural geological model report. Read them together to examine domain geometry, contacts and continuity from three directions. The viewer at the top of the page also lets you switch plane and diagnostic.
Test the interpretation
Compare the predicted categories with observed intervals and evaluate generalization on held-out drillholes. These orthogonal sections illustrate the modeled geometry; independent validation establishes predictive performance.
03–06 / Resource workflow
From block grades to drilling decisions
Move from geological interpretation to estimates, reviewable classification, conditional realizations and production-aware drill planning.
03 / ESTIMATEGRADES
Estimate within geological boundaries
Carry the interpreted domains and supported directions of continuity into block-scale grade estimation. Document the samples used, search configuration, block support and choices made for each domain.
Review grade trends against composites, inspect estimated blocks near drillholes and compare global summaries. These checks describe the support and behavior of a deterministic estimate; spatial uncertainty is evaluated through conditional simulation in step 05.
Grade model + validationAssumptions and diagnostics stay visible
04 / CLASSIFYCONFIDENCE
Make classification reviewable
Bring drillhole spacing, geological continuity, sample support and estimation diagnostics into explicit classification criteria. Review where the evidence supports each resource category, including transitions near sparse drilling or complex geology.
When conditional simulations are available, examine production-scale variability as additional evidence. Define classification thresholds for the deposit and intended mining scale, and keep the reasoning traceable to the underlying data.
Spatial evidence + criteriaProject-defined thresholds
05 / SIMULATETBS
Explore possible orebodies
Conditional simulation with Turning Bands (TBS) is a geostatistical workflow. TBS builds 2D or 3D correlated random fields from fast one-dimensional simulations along many randomly oriented lines; the realizations are then conditioned to the observed drillhole data.
Generate many plausible block models for continuous grades and categorical lithology. For lithology, conditioned Gaussian fields can be converted to rock types through truncation rules. Compare the realizations to quantify spatial variability and uncertainty at block and production scales.
Multiple conditional realizationsSame drillhole observations · different plausible outcomes
Concept animation · observations stay fixed
DOMAINS A B C DGRADE LOW HIGH
06 / DHSADRILLING
Drill for the confidence production needs
Drillhole Spacing Analysis (DHSA) tests alternative drillhole grids through conditional realizations. For each grid, assess uncertainty over the volume mined in each production increment, rather than judging spacing from a single block.
Compare the mesh required for current production with that for expected production. Set an acceptable uncertainty for each production horizon, then identify spacing and infill that meet its criterion.
Drilling grid × production volume × acceptable uncertaintyEvaluate each production horizon separately
Final output / OBK report
One report. The full orebody story
At the end of the workflow, OBK will generate a comprehensive report summarizing the main findings, assumptions and diagnostics from EDA, geological modeling, estimation, resource classification, conditional simulation and DHSA.
It will extend the visual and analytical structure of the current neural geological model report into a more complete document, using OBK's forest, sage, sand and ivory palette throughout.
Planned final deliverable · illustrative layoutOBK / Orebody KnowledgeComprehensive report
Report overviewFrom data to drilling decisions
Six connected chapters, pairing figures and conclusions with methods, assumptions, validation and next checks.
Executive summary / illustrativeVisual preview · no project values shown
01–02Data & geology
03–04Resource & confidence
05–06Risk & drilling
01
EDAData checks, distributions and spatial patterns
02
Geological modelANISO, neural outputs and model validation
03
EstimationInput support, block grades and diagnostics
04
ClassificationCriteria, evidence and resource categories
05
SimulationConditional realizations and spatial uncertainty
06
DHSADrill spacing, production and uncertainty limits
Summary · methods · figures · assumptions · diagnostics
Built for scrutiny
Evidence at every step
A result is only useful when its assumptions, support and limits are visible. OBK connects geological reasoning to diagnostics that can be challenged and refined.
Data integrity
Review input intervals, trajectories, exclusions and sampling support before modeling.
Geological fit
Compare conditioned blocks and interpreted solids with observed drillholes as distinct checks.
Predictive performance
Use held-out drillholes and spatial diagnostics to test generalization, without treating a fit target as a guarantee.
Resource confidence
Trace classification criteria back to spatial support and continuity, using simulation-derived uncertainty when available.
More applications for neural modeling
Beyond lithology and grade
Neural modeling in OBK can target more than rock types and grades. With suitable data and validation, the same geological and spatial context can support geometallurgical and geomechanical models for processing and mine design.
Geometallurgy
Connect geological domains, mineralogy, chemistry and processing testwork to model how the ore responds through each concentration stage.
Spatial models can include:
- Mineral distribution and liberation relevant to processing
- Ore and gangue percentages in concentrates from successive concentration stages
- Mass recovery and metallurgical recovery, supported by testwork and a defined process route
MineralogyConcentrate compositionMass & metallurgical recovery
Geotechnics
Integrate drillhole logs, geomechanical tests, geological domains and drilling measurements to map rock-mass behavior and support mine design.
Potential model targets:
- Rock Quality Designation (RQD) and geomechanical classes
- Rock strength (R) and compactness where measured or characterized
- Measurement While Drilling (MWD) indicators, calibrated against geotechnical observations
RQD & classesStrength & compactnessMWD
Orebody Knowledge
Geology first. Decisions with depth
OBK brings geological interpretation, spatial modeling and machine learning into one connected software platform. We believe the shape, continuity and uncertainty of an orebody should be understood before they are reduced to a resource number.
Meet the people behind OBK
Resource Geology · Orebody Modeling
André Andrade
Senior Resource Geologist
MSc in Geosciences and Environment
André works across geological modeling, mineral resource estimation, classification and geometallurgy, with experience in iron ore, gold, tin and rare earths. At OBK, he connects geological interpretation with data-driven modeling.
View LinkedIn profile
Resource Estimation · Simulation
Flávio Amarante
Specialist Resource Geologist
MSc in Geostatistics
Flávio brings more than 10 years of experience in mineral resource estimation. At OBK, he focuses on estimation and conditional simulation, with particular attention to geological uncertainty and classification.
View LinkedIn profileLET'S CONNECT / OBK
Let's read your orebody together
Every deposit has its own geometry, data and decisions. Tell us where you are in the workflow, and we can explore how OBK fits your geological and mining questions.
OBK / GEOLOGICAL SOLIDILLUSTRATIVE IMAGE
01 / PROJECT CONVERSATION
Tell us what you are working on
Share the context that matters to your team. Choose as many areas as you like.