We’re putting $5.5M behind research and development of smaller, specialized AI models for real-world deployment — Read our manifesto →
    All posts

    [Case Study] Fuel Is a Dispatch Problem: 1,242 Hours Recovered Across 30 Haul Trucks

    How a surface gold mine changed refuel assignments using fuel level, distance, and queue data to recover truck hours and increase moved tonnage.

    Conscious Engines

    Industry: Surface mining
    Organization: An unnamed gold mine in Suriname
    Use case: Haul-truck refuel assignment optimization
    Evidence basis: Mine Tech Services case study
    Disclosure: This is an independent analysis by Conscious Engines. The mine is unnamed and results are consultant and customer reported, limiting independent verification.

    1. Outcome at a Glance

    The operation reconfigured its fleet-management system to assign refueling based on fuel level, distance, and expected queue rather than simple routines. Across 30 haul trucks, it reported recovering 1,242 operating hours per year, supporting an additional 348,000 tonnes moved and approximately $7 million in revenue.

    Key Outcomes

    30

    Haul trucks

    Fleet in scope reported in the cited case.

    About $7 million

    Revenue impact

    Company estimate reported in the cited case.

    MeasureReported resultEvidence note
    Haul trucks30Fleet in scope
    Truck hours recovered1,242 per yearModelled operating result
    Additional material moved348,000 tonnesReported throughput impact
    Revenue impactAbout $7 millionCompany estimate

    The case reframes fuel efficiency. The largest problem was not necessarily fuel burned per engine hour. It was production time lost through premature refueling, queues, and trucks running short. Optimizing the decision of when and where to refuel increased equipment availability.

    The public source does not name the customer or publish detailed calculation assumptions. The revenue number should therefore be treated as a case estimate, not an independently audited return.

    2. The Operational Problem

    Haul trucks operate in a synchronized production system. A refuel stop removes a truck from the haul cycle. If many trucks arrive together, queues form. If a truck refuels too early, it sacrifices productive time. If it runs out or must make an emergency trip, disruption is greater.

    The optimal choice depends on live fuel level, burn rate, route, assignment, distance to fuel, queue length, shift plan, and production priority. Manual rules struggle to update these variables continuously.

    Fuel data can also be unreliable. Gauge calibration, telemetry gaps, inconsistent delivery records, and different fuel sources can distort decisions. An enterprise must first reconcile receipt, storage, dispensing, and equipment consumption if it wants trustworthy optimization.

    3. What Was Built

    The intervention used the existing fleet-management system and changed its refuel-assignment logic.

    System at a Glance

    Current fuel level

    Estimate urgency and remaining operating range.

    Distance to fuel point

    Calculate travel and diversion cost.

    Queue condition

    Avoid nonproductive waiting.

    Truck assignment

    Protect current production cycle.

    Burn-rate history

    Predict when fuel becomes critical.

    Dispatch state

    Coordinate the decision across the fleet.

    InputDecision role
    Current fuel levelEstimate urgency and remaining operating range
    Distance to fuel pointCalculate travel and diversion cost
    Queue conditionAvoid nonproductive waiting
    Truck assignmentProtect current production cycle
    Burn-rate historyPredict when fuel becomes critical
    Dispatch stateCoordinate the decision across the fleet

    This is a constrained optimization use case, not a generative AI task. A language model can support exception explanations, shift summaries, and operator queries, but the dispatch decision should be calculated from live data and explicit constraints.

    A bespoke system can add anomaly detection for suspected fuel leakage, reconciliation across suppliers, driver voice capture, and a model that forecasts consumption by route, payload, grade, weather, and equipment condition.

    4. How It Reached Production

    The public account emphasizes configuration and training, which is important. Optimization creates value only when dispatchers and operators trust and follow the assignments.

    Audit fuel telemetry. Validate gauges, timestamps, dispensers, and missing data before changing assignments.

    Model the whole delay. Include travel, queue, fueling, return, and interruption to the haul cycle.

    Set operational guardrails. Minimum safe fuel, emergency routing, equipment restrictions, and supervisor override should remain explicit.

    Train dispatch and operators. Users need to know why assignments changed and how to report exceptions.

    Verify production impact. Compare recovered hours, tonnes, fuel per tonne, queue time, runouts, and maintenance effects against normalized baselines.

    The revenue estimate depends on additional tonnes being processed and sold. If the plant, market, or downstream operation is constrained, recovered truck hours may not realize the same financial value.

    5. What Mining Leaders Should Take Away

    This mine shows why fuel programs should connect procurement, storage, dispensing, equipment, dispatch, and production. A narrow change in dispatch logic reportedly recovered more than a thousand truck hours.

    A production fuel-intelligence stack reconciles fuel receipts from different sources, detects loss and anomalies, forecasts consumption, optimizes refuel assignments, and explains exceptions through a private operations assistant. Speech-to-text can capture field events where manual entry is difficult.

    The core metric is total fuel and delay cost per tonne moved, not liters alone. That keeps the system aligned with production while preserving safety and equipment constraints.

    Sources