By Dr Pouya Nobaha, Principal Diagnostic Lead; and Dr Ali Soofastaei, Executive Director (Technical and Science), ContinuumOps Australia

A physics-grounded framework for finding, prioritising and sustaining improvement opportunities across the mining value chain has the power to transform the industry.

Mining operations are complex systems. A decision made in drill and blast changes fragmentation; and fragmentation affects loading productivity, crusher utilisation, mill power draw, circulating load, and sometimes recovery. A haulage delay can look like a dispatch problem, but the underlying driver may be shovel digability, road condition, fuel bay congestion, crusher downtime or inconsistent stockpile management. In this environment, isolated KPIs often describe the symptom but do not reveal the causal chain.

Traditional operational reviews are usually strong at reporting what happened. They are weaker at explaining why performance moved, whether a loss was controllable, and which action should be taken first. This creates three common problems. First, teams debate different numbers because production, maintenance, geology and plant systems use different clocks, tags, and definitions. Second, improvement projects are selected based on visibility rather than value. Third, benefits are difficult to sustain because the decision process that created the loss is not redesigned.

Operational performance diagnostic (OPD) addresses these gaps by treating performance as a diagnostic question rather than a reporting question. It asks: What is the expected performance under current material and operating conditions? How large is the gap to actual performance? Which part of the gap is explained by material, asset condition, process design, control behaviour, work management or data quality? What action is technically feasible, commercially material and operationally acceptable?

HOW OPD DIFFERS FROM REPORTING

A dashboard is useful for visibility, but it generally displays KPIs after decisions have already been made. OPD adds a diagnostic layer between the KPI and the action. It tests hypotheses, uses physics constraints, checks whether event data and sensor data agree, quantifies the value of each loss, and converts the result into an action charter. The output is therefore not simply a chart; it is a ranked set of interventions with evidence, ownership and expected benefit.

Diagnostic question	Typical evidence
Where is the loss?	bottleneck map, delay tree, overall equipment effectiveness, rate and utilisation curves
Why is it happening?	event sequence, material context, process physics and shift notes
What is controllable?	root-cause tests, operating envelope and constraint analysis
What is it worth?	lost tonnes, cost per tonne, energy per tonne, downtime, dilution or recovery impact
Who must act?	role, trigger, decision right, escalation pathway and governance rhythm

OPD is a structured assessment that combines contextualised data, process physics, operating knowledge and value analysis to identify the most material causes of performance loss, and to define implementable actions for improvement.

OPD ARCHITECTURE

A robust OPD has four connected layers. The first layer is data and context: time-series sensors, fleet events, production records, shift logs, geological domains, blast designs, maintenance work orders, energy meters and cost structures must be aligned to the same operational timeline.

The second layer is the physics baseline: expected production rate, loading time, crusher power, mill-specific energy, availability or recovery are estimated under the material and operating conditions observed in the diagnostic window.

The third layer is variance decomposition: instead of treating all deviations as random noise, OPD allocates the performance gap to explainable drivers. These include ore hardness, fragmentation, equipment health, road gradient, queuing, operator practices, control limits, blocked chutes, changeover losses and maintenance execution.

The fourth layer is value and decision design: each driver is converted to economic exposure and linked to a change in the decision process.

This architecture prevents a common analytics failure in mining: producing an accurate model that cannot be used. A predictive model may forecast mill throughput, for example, but an OPD asks what the operation will do with the forecast. Will the blast design change? Will feed blending change? Will maintenance prioritise a conveyor? Will the control room use a different set point? If the decision path is unclear, the diagnosis is incomplete.

PHYSICS-GROUNDED BASELINES

The OPD baseline does not need to be a perfect digital twin; it simply must be good enough to define what performance should have been under the observed context. For haulage, this may include payload, gradient, distance, rolling resistance, speed restrictions and queue time. For crushing and grinding, it may include feed size, ore hardness, liner state, power draw, circulating load, water addition and control constraints. For maintenance, it may include asset age, utilisation, operating load, failure mode and repair history.

A useful baseline has three qualities: it is explainable to operators and engineers; it is sensitive to the drivers that can be changed; and it can be used to test whether proposed actions would close the observed gap. Machine learning models can be used, but they should be constrained by process logic, validated by site expertise and linked to decision rules.

DIAGNOSTIC DOMAINS

OPD should not be limited to one discipline. A technically correct grinding diagnosis can fail if it ignores mine planning constraints; a haulage optimisation can fail if it ignores workshop capacity; and a blast fragmentation intervention can fail if it ignores downstream blending and stockpile rules. For this reason, OPD scores and investigates six domains: material, asset, process, control, data and organisation.

Domain	OPD focus	Example loss signal
Material	ore hardness, grade, structure, fragmentation, moisture	high energy per tonne or low crusher rate by domain
Asset	equipment condition, availability, reliability, constraints	repeat stoppages, derates, speed restrictions
Process	flow, queues, handovers, operating modes	wait time, rehandle, blocked flow, mode changes
Control	set points, alarms, APC rules, operator response	unstable operation or slow response to disturbances
Data	definitions, coverage, accuracy, context labels	conflicting KPIs or unexplained variance
Organisation	decision rights, routines, ownership, capability	actions not sustained after review

LOSS ACCOUNTING

Loss accounting is the bridge between diagnosis and value. It begins with a maximum practical performance envelope rather than a theoretical nameplate target. The difference between practical potential and actual performance is then separated into categories such as planned downtime, unplanned downtime, operating rate loss, quality loss, process instability, material-driven loss and decision-driven loss. The aim is to show where value is leaking and which leaks can be reduced with realistic effort.

The diagnostic should avoid over-precision. A defensible range is often more useful than a single false-accurate number. For example, an OPD may estimate that improving crusher feed consistency could release 1.5–3.0 per cent more semi-autogenous grinding (SAG) mill throughput under selected ore domains. That range is enough to justify a field trial if the operational risk is low and the benefit can be measured.

TYPICAL OPD METRICS

• Production: Tonnes per operating hour, feed rate stability, utilisation and bottleneck duration.
• Comminution: P20/P50/P80, crusher closed side setting, mill power, specific energy, circulating load and product size compliance.
• Mobile equipment: Payload, cycle time, queue time, fuel per tonne per kilometre, speed by segment and road condition exposure.
• Maintenance: Availability, mean time between failures, mean time to repair, planned work compliance and failure recurrence.
• Value: Cost per tonne, energy per tonne, lost production cost, maintenance cost, rehandle cost and risk-adjusted opportunity value.

ROOT-CAUSE LOGIC

Each high-severity cell in the diagnostic heat map should be converted into a root-cause hypothesis. A hypothesis must include a mechanism, evidence required to prove or disprove it, expected value, operational owner and possible action. For example, if coarse blast fragmentation is increasing crusher power and reducing SAG throughput in a specific ore domain, then the OPD must check the blast design, dig block location, crusher feed size, mill load and production timeline together. Without that linked evidence, the problem may be incorrectly assigned to the plant rather than the mine.

APPLICATIONS ACROSS THE VALUE CHAIN

In drill and blast, OPD can identify whether fragmentation, dilution, digability and downstream size distribution are consistent with the intended blast outcome. The diagnosis may link burden, spacing, powder factor, timing, rock mass properties, stemming quality and initiation pattern to downstream productivity. The value question is rarely just whether the blast was compliant; it is whether the blast created the most valuable feed condition for the whole operation.

In load and haul, OPD can separate avoidable delay from structural constraints. A fleet may appear underutilised because of dispatcher behaviour, but the root cause may be crusher queue management, shovel relocation, road geometry, refuelling delays or maintenance windows. A strong diagnostic uses event sequences and route-level context to distinguish short-term inefficiency from system design limitations.

In comminution, OPD can connect feed variability to crusher, stockpile and mill behaviour. High energy intensity may be caused by ore hardness, but it may also be driven by poor feed size control, liner condition, water balance, classification inefficiency or control instability. A physics-grounded diagnostic tests these explanations before recommending capital projects or major control changes.

In maintenance and reliability, OPD helps to prioritise failures by value exposure rather than event count. A repeated minor fault may be more damaging than a rare major failure if it constrains the bottleneck every shift. The diagnostic should connect equipment condition, work management, spares, labour, shutdown planning and production consequences into one loss view.

Use case	Diagnostic objective	Potential action
Blast-to-mill	link fragmentation to plant rate and energy	Adjust blast design by ore domain and target partical size distribution
Haulage fuel	explain fuel per tonne variation by route, payload and delay	road maintenance, speed rules and payload control
Crusher constraint	separate mechanical, feed and operational losses	feed control, liner plan and choke-feed practice
SAG instability	identify causes of rate and product-size swings	blend rules, control limits and operator advisory
Maintenance backlog	rank work by production consequence	risk-based planning and shutdown bundling

PRIORITISATION CRITERIA

• Value: Expected contribution to tonnes, cost, energy, recovery, safety or risk reduction.
• Confidence: Strength of evidence and repeatability of the observed mechanism.
• Control: Degree to which site teams can influence the root cause.
• Speed: Time required to implement and verify the action.
• Complexity: Integration, change management, capital, data and operating discipline required.
• Sustainability: Likelihood that the benefit will remain after the diagnostic team leaves.

IMPLEMENTATION ROAD MAP

A practical OPD engagement should begin with a narrow but materially valuable question. Examples include: Why is the SAG mill below the expected rate in domain A? Why is fuel intensity rising on a specific waste haul? Why does crusher availability look acceptable while plant feed is still unstable? A focused question creates speed and prevents the diagnostic from becoming a broad data audit.

The first stage is preparation. The team defines the boundary, stakeholders, data sources, KPI definitions and decision cadence. The second stage is baseline building: data is cleaned, aligned and contextualised, then compared with a practical performance envelope. The third stage is diagnostic testing: hypotheses are developed, evidence is gathered, root causes are ranked, and field validation is completed. The fourth stage is action chartering: each recommendation is converted to a short implementation plan with owner, trigger, expected benefit and risk. The final stage is measurement and verification.

GOVERNANCE MODEL

Governance should be light but explicit. The executive sponsor protects the value question and removes barriers. The site owner confirms operational feasibility. The technical lead manages the diagnostic logic. The data lead controls definitions and quality checks. The improvement owner implements the action. The measurement owner tracks the benefit after the change. Without these roles, OPD can produce high-quality analysis that does not change performance.

LIMITATIONS AND SAFEGUARDS

OPD is not a substitute for engineering judgement, safe operating practice or detailed design; it is a decision-support method. Recommendations should be tested within approved operating envelopes and reviewed by responsible site personnel. The diagnostic should also state confidence levels clearly, because poor data quality, short time windows and unmeasured context can all create misleading conclusions.

LOOKING FORWARD

An OPD gives mine sites a disciplined way to move from KPI visibility to causal improvement. By combining contextualised operational data with process physics, value analysis and decision governance, OPD can identify hidden losses, prioritise practical actions and sustain benefits. Its real strength is integration: it connects mine, plant, maintenance, data and leadership teams around one evidence-based view of where value is lost and how it can be recovered.

---

References

McKee, D. J. (2013). ‘Understanding Mine to Mill.’ Cooperative Research Centre for Optimising Resource Extraction, Brisbane.

Morrell, S. (2004). ‘Predicting the specific energy of autogenous and semi-autogenous mills from small diameter drill core samples.’ Minerals Engineering, 17(3), 447-451. https://doi.org/10.1016/j.mineng.2003.10.019.

Napier-Munn, T. J., Morrell, S., Morrison, R. D., and Kojovic, T. (1996). ‘Mineral Comminution Circuits: Their Operation and Optimisation.’ Julius Kruttschnitt Mineral Research Centre.

van der Aalst, W. M. P. (2016). ‘Process Mining: Data Science in Action.’ Springer. https://doi.org/10.1007/978-3-662-49851-4.

Nakajima, S. (1988). Introduction to TPM: Total Productive Maintenance. Productivity Press.

ISO. (2024). ‘ISO 55000:2024 Asset management – Vocabulary, overview and principles’. International Organization for Standardization.

ISA. (2025). ‘ANSI/ISA-95.00.01 Enterprise-Control System Integration – Part 1: Models and Terminology’. International Society of Automation.

Grieves, M. (2014). ‘Digital Twin: Manufacturing Excellence through Virtual Factory Replication’. Florida Institute of Technology.

Davenport, T. H., and Harris, J. G. (2007). Competing on Analytics: The New Science of Winning. Harvard Business School Press.