By Dr Ali Soofastaei, CEO, Innovative AI

Near-real-time digital integration is transforming mine-to-mill optimisation, enabling mining operations to connect every blast, haul, and mill decision into a single, data-informed control loop that drives efficiency and profitability.

Mining companies have long understood that the connection between the mine and the mill determines the profitability of an operation. While geological endowment sets the stage, it is the quality of integration between upstream mining activities and downstream processing that dictates throughput, recovery, energy consumption and, ultimately, shareholder value. The mine-to-mill (M2M) concept, first popularised in the 1990s, has evolved significantly. What began as a push for better alignment of blast fragmentation and mill performance is now changing into a digitally orchestrated system that can operate in near real time.

This transformation is driven by the convergence of new sensing technologies, advanced process control, and digital standards that enable interoperability. The result is a practical framework for closed-loop optimisation that allows every blast design, haul assignment and stockpile reclaim decision to be evaluated for its downstream economic impact.

Why M2M, and why now?

Traditional M2M programs showed that aligning blasting, crushing and milling could unlock step-change improvements: higher throughput, lower energy per tonne and better recoveries. What has changed in recent years is the digital and instrumentation layer. Today, real-time particle size vision systems, cyclone tomography, mill impact sensors, and artificial intelligence (AI)–powered flotation cameras provide continuous observability across the value chain.

On the computing side, integrated extraction simulators and plant digital twins enable both physics-based and data-driven models to be incorporated into daily decision-making loops. Combined with industrial interoperability standards, these technologies make seamless integration achievable at scale. This means M2M is no longer a once-a-year study; it can now function as an operational control philosophy.

What ‘near real time’ really means

In the context of process optimisation, ‘near real time’ refers to the ability to sense, compute, decide and act within the settling time of the process itself. For grinding and flotation circuits, this means seconds to minutes, not hours. Typical latency budgets include sensor acquisition in seconds, edge preprocessing within seconds, model execution within seconds, advanced process control (APC) and model predictive control optimisation cycles within tens of seconds, and actuation of set points within seconds. This rapid cycle ensures that corrective actions – such as adjusting mill charge, cyclone split or froth strategy – occur before variability propagates downstream, protecting both throughput and recovery.

A reference architecture for M2M in 2025

The modern M2M architecture can be structured in four layers:

• Level 0–2 (process): Instrumentation and control systems, including fragmentation vision, cyclone sensors, mill fill detectors and flotation froth imaging.
• Level 3 (operations management): Historians and streaming analytics platforms, plus fleet management systems that capture ore attributes and coordinate routing.
• Level 4 (enterprise): Digital twins, simulation platforms and planning suites that integrate economic objectives with operational settings.
• Security and governance: Secure telemetry, cybersecurity frameworks, and clear boundaries to ensure reliable, auditable data flows.

Together, these layers create a continuous chain from shovel to concentrator to corporate dashboards.

Modelling: physics meets data

A critical feature of modern M2M is the blending of physics-based and machine learning (ML) models. Population balance models remain the backbone for grinding and classification, providing explainability and robustness. ML surrogates and soft sensors fill the gaps, enabling faster updates between lab assays and online analysers, especially for grind size estimations and froth grade prediction. Integrated flow sheets combine blasting, crushing, grinding and flotation into one computational graph, enabling predictive scenario testing before implementing changes.

This hybrid approach mitigates the risks of relying solely on ML, which can drift – or solely on physics, which can be slow and data hungry.

Closed-loop optimisation: from pit to plant

The essence of M2M is closing the loop. A modern strategy might look like this:

• Drill and blast – plant: Blast designs are tailored not only for fragmentation, but also for downstream recovery economics. Vision systems feed fragmentation data directly into mill charge and cyclone set points.
• Ore routing and blending: Truck dispatch systems and stockpile models ensure ore chemistry, hardness, and particle size distributions are aligned with real-time plant needs.
• Grinding and classification: Advanced controllers manipulate feed rates, water addition, and mill speeds to achieve target grind sizes at minimal energy.
• Flotation: AI-enabled froth vision guides air rates, reagent dosing, and froth pull to operate near the grade-recovery frontier. This continuous orchestration ensures that every operational decision is judged against its impact on plant KPIs and financial outcomes.
• Feedback to mine planning: Recovery and penalty insights are fed into next-day blast and haul schedules, completing the mine-to-market loop.

Executive Outcomes and KPI Trees

For executives, the benefits of M2M integration can be represented in a KPI tree: earnings before interest, taxes, depreciation and amortisation (EBITDA) driven by recovery × throughput × (price – costs). Recovery levers include grind size stability, froth selectivity and grade curves. Throughput levers include crusher particle size distribution, SAG power and cyclone capacity. Cost levers include energy per megawatt hours, grinding media and liner wear, and reagent use.

These metrics translate directly into investor-facing performance. Case experiences across multiple operations have demonstrated significant gains in throughput and energy efficiency by applying integrated M2M programs.

Implementation road map: from weeks to months

The rollout of M2M can follow a staged approach:

• Phase 0 (0–4 weeks): Foundations – appoint governance roles, conduct sensor/data audits, test latency budgets.
• Phase 1 (5–12 weeks): Sensing and soft sensors – deploy fragmentation and mill fill monitoring, validate cyclone tomography, build soft sensors.
• Phase 2 (4–9 months): Model-in-the-loop APC – configure advanced process control systems, integrate digital twins for scenario testing.
• Phase 3 (10–18 months): Closed-loop optimisation – connect mine dispatch with APC requests, establish variance reviews, and conduct cybersecurity audits.

This structured approach reduces risk, builds trust in data and ensures that early wins sustain momentum.

Data, integration and cybersecurity

The transition to near-real-time integration requires strong data governance. Clear standards define roles and data exchanges. Secure, scalable telemetry and disciplined certificate management ensure resilience. Cybersecurity must be embedded from day one to avoid critical vulnerabilities, including zone segmentation, multifactor administration, and rigorous patch and backup regimes.

A historian-first strategy – treating the historian and asset models as the event backbone – ensures that data quality and lineage underpin every optimisation step.

Economics and funding the business case

Unlike capital-heavy expansions, M2M is often capex-light and opex-heavy. Value comes from instrumentation, control and software rather than extensive mechanical upgrades. This makes it attractive in today’s capital-disciplined environment. The recommended approach is incremental validation: prove value in grind size stability, then flotation recovery, then mine–plant coordination. Each wave builds a stronger case for broader rollout.

Common pitfalls to avoid

Experience shows that M2M projects can falter if not managed carefully. Common pitfalls include:

• Data swamps without ownership: solved by governance boards and asset models.
• Over-reliance on unverified ML: avoided by tying outputs to physics checks.
• Latency surprises: managed by budgeting end-to-end cycle times and failover protocols.
• Security bolted on late: best to integrate cybersecurity from the outset.
• Over-focus on blasting: ore blending and routing can deliver equal or greater value, and must not be overlooked.

Governance, talent and culture

Technology alone does not deliver transformation. Successful M2M requires cross-functional squads that span mine planning, blasting, processing, control, IT/OT and finance. Shared objectives – such as reducing grind variability, improving recovery, or cutting misrouted tonnes – align incentives and overcome silo thinking. Embedding M2M into daily production meetings and dashboards ensures that optimisation becomes part of the organisation’s DNA rather than a pilot project.

Looking ahead

As sensors multiply and digital standards mature, the bottleneck to M2M integration is no longer technology, but organisation. Companies that learn to treat variability as a controllable, tradeable asset – rather than a nuisance – will build a durable competitive advantage. In this future, near-real-time modelling is not simply a technical capability; it is a strategic differentiator that turns mining from a reactive industry into a proactive, optimised enterprise.

M2M integration with near-real-time modelling is reshaping the future of mining. By combining advanced sensors, physics-informed data models, digital twins, and secure interoperable platforms, mining companies can achieve sustainable improvements in throughput, recovery, energy efficiency and compliance. For executives, the appeal lies in the alignment of technical innovation with business value: higher EBITDA, reduced risk, and improved resilience in a volatile commodity environment. The message is clear – those who master M2M integration today will define the benchmarks for mining performance tomorrow.