Technical Notes¶

This page captures advanced ideas, modelling strategies, and future-ready principles for modern mine planning. It is written for technical decision-makers, data scientists, and optimisation professionals who want a structured, expert reference for building integrated mine value chain systems.

Why These Notes Matter¶

Mine planning is evolving from periodic schedule generation to a continuous decision-support system that integrates operations, economics, geology, and sustainability.

This page is not a simple overview. It is intended to:

define the architectural elements of advanced mine planning systems
identify the modelling assumptions that shape high-value results
highlight the research and practical gaps between current tools and future-ready solutions

Modern mine planning must combine:

economic value and processing variability
geotechnical risk and operating flexibility
uncertainty propagation across the value chain
ESG constraints as core planning requirements
real-time operational feedback

Digital Twin for Mining Systems¶

A digital twin is a live, bi-directional representation of the mine system, not a one-time model.

Core components¶

dynamic block models and geological uncertainty
fleet telemetry and equipment state
processing plant throughput and recovery response
geotechnical stability and excavation geometry
optimisation engines and operational business rules
ERP / scheduling / sensor feedback loops

Why it matters¶

A mature twin enables:

continuous plan updates instead of annual re-plans
early detection of value erosion
constrained optimisation over both mine and plant
rapid evaluation of scenarios and what-if changes

Expert note¶

The highest leverage comes from connecting stochastic mine models with real-time data streams. This allows decisions to move fluidly between expected-value planning and risk-aware execution.

Explainable AI in Mine Planning¶

Explainable AI is essential for trust, adoption, and regulatory compliance in mining.

Key questions it must answer¶

Why was this block or stope chosen?
What features drive cut-off grade or blending decisions?
How sensitive is the schedule to changes in price or recovery?
Which constraints are binding and why?

Recommended techniques¶

SHAP values for feature attribution
counterfactual scenarios and decision rules
sensitivity analysis of objective and constraints
rule extraction from machine learning models

Expert note¶

Transparency is only useful when it covers both the predictive model and the optimisation layer. Planners need to know not only what the grade model says, but how that prediction propagates through the scheduling and processing decisions.

Reinforcement Learning for Scheduling¶

Reinforcement learning can treat mine scheduling as a sequential decision problem, bringing adaptive control to a traditionally static discipline.

Model structure¶

State: mine progress, inventory, fleet availability, market signals
Action: mine, process, delay, re-sequence, or reassign resources
Reward: NPV adjusted for penalties, ESG costs, and operational risk

Practical applications¶

adaptive cut-off grade policies
dynamic haulage and dispatch decisions
responsive production sequencing
autonomous asset utilisation

Expert note¶

Hybrid solutions are most effective. Use RL to generate flexible policies, then validate and constrain them with mathematical optimisation so feasibility, safety, and traceability are preserved.

Geometallurgical Optimisation¶

Processing performance is variable and should be modelled explicitly rather than averaged away.

Integrated workflow¶

classify ore types by metallurgy and recovery response
link block model attributes to plant behaviour
optimise mining and milling together
evaluate product quality, throughput, and revenue simultaneously

Why this is advanced¶

Most current systems assume fixed recovery. Integrated geometallurgy allows planners to capture real value by aligning mining, blending, and processing decisions.

Expert note¶

Effective systems support advanced recovery functions and multi-objective trade-offs, enabling planners to quantify the value of flexibility versus immediate throughput.

Graph and Hypergraph Models¶

Graph-based representations are powerful for encoding precedence, connectivity, and dependencies.

Graph use cases¶

open-pit precedence and ultimate pit limit via closure graphs
underground stope access and sequencing networks
logistics and stockpile flow as network optimisation problems

Hypergraph opportunities¶

Hypergraphs extend traditional graphs by capturing interactions among multiple entities simultaneously.

Potential applications:

stope groups with shared access and ventilation dependencies
complex blending and multi-material constraints
equipment interactions with combined resource requirements

Expert note¶

Graph modelling should be treated as the solver structure itself. Encoding adjacency, precedence, and resource interactions explicitly improves both model fidelity and optimisation performance.

ESG-Constrained Planning¶

Environmental, Social, and Governance objectives are now core planning requirements.

Common ESG dimensions¶

carbon emissions and energy source mix
water consumption and discharge limits
rehabilitation and closure progress
community impact and social licence

Planning guidance¶

include ESG metrics in objectives and constraints
define hard limits for compliance-critical factors
use scenario analysis to quantify trade-offs
embed ESG performance into decision-making workflows

Expert note¶

The best systems evaluate ESG and economic outcomes together. This makes every schedule inherently sustainable, rather than retrofitting sustainability after the fact.

Expert Notes and Deployment Guidance¶

Implementation principles¶

build modular data and model pipelines
version models and scenarios for auditability
prioritise explainability over marginal solver gains
deploy optimisation as part of an operational workflow, not an isolated batch process

What distinguishes advanced systems¶

real-time integration of planning and execution
stochastic and risk-aware optimisation
explicit coupling of mine, plant, and logistics
transparent, traceable schedule reasoning

Final takeaway¶

The future of mine planning is not a single algorithm. It is a unified system that combines digital twins, explainable AI, adaptive control, geometallurgical realism, and ESG-aware optimisation into one cohesive decision platform.