Yard & Bin Inventory Optimisation

Application

Every materials-handling chain — bins, stockpiles, train loadout — exists to decouple processes that run at different rhythms: haulage arrives in surges, the plant wants steady feed, trains and ships load to a schedule. Bins and stockpiles are the buffers that absorb the mismatch, but on most operations they are run at habitual levels chosen by convention rather than by what the situation demands. The result is surge capacity used by accident: bins that are too full to absorb an upstream rush, or too empty to ride through a downstream demand peak.

Minealytics Yard & Bin Inventory Optimisation is a high-level optimisation layer that continuously tracks bin levels and stockpile inventories across the chain and determines optimal bin level setpoints by reasoning over both sides of every buffer:

• Upstream constraints: truck and haulage arrivals, and the variability of ROM supply.
• Downstream constraints: plant feed demand, train loading schedules and shiploading windows.

Solution

Instead of operators holding bins at habitual levels, the optimiser positions inventory where it buys the most protection — enough headroom to absorb upstream surges, enough material to ride through downstream demand peaks. Its outputs are bin level setpoints and feeder/reclaim targets, handed to the regulatory control layer to execute; the optimiser decides where inventory should sit, and the existing control system moves it there.

The benefits compound across the chain:

• Fewer plant feed interruptions, because bins hold enough material to bridge upstream gaps.
• Fewer truck and train delays, because headroom and reclaim capacity are available when arrivals and loadouts need them.
• Surge capacity used deliberately rather than by accident — every buffer positioned for the disturbances it is most likely to face.
• Whole-of-chain visibility, with the complete inventory position — bins, stockpiles and loadout — in one place.

How It Works: Schedule to the Event, Not the Level

Classical bin control chases a nominal level with a PI loop, moving setpoints continuously and losing throughput with every move. The Minealytics scheduler works differently — it is supply-driven and event-horizon based:

1. Predict forward, not just measure now. From measured inflow, the current discharge setpoint and the time remaining to the next loading event (a train, a hatch, a plant campaign), the scheduler predicts what the bin inventory will be at the end of the event.
2. Hold while feasible. While the predicted end-of-event inventory sits inside a safe band, setpoints are held. No trajectory shaping, no continuous trimming — the bin is allowed to do its job as a dynamic buffer.
3. Move only when the prediction leaves the band. Then the scheduler recalculates the discharge setpoint explicitly from the mass balance over the remaining horizon, clamped to equipment limits.
4. Escalate upstream only when necessary. If discharge limits alone cannot restore feasibility, the scheduler issues an upstream inflow constraint request — inflow trimming is a last resort, not a habit.

This band-based hold behaviour minimises setpoint movement (each unnecessary move costs stability and throughput), keeps headroom for event changes by design, and makes constraint handling explicit: safety interlocks first, then equipment limits, then band feasibility, and only then upstream requests. Operators retain a familiar Manual / Auto / Cascade framework — in cascade the scheduler has authority; a single action hands control back.

Engineered for the Real World

Prediction-based control is only as good as its inputs, so the scheduler treats signal validity as a first-class concern:

• Validated inventory: dual level measurements are cross-checked; on discrepancy the conservative value is used and the faulty channel excluded — the scheduler stays available.
• Instrument-fault fallback: on a weigher fault, the system freezes adaptation, drops to degraded model-based control at derated throughput, and alarms — rather than acting on bad data.
• Slug rejection: physically impossible flow spikes are ignored for model updates and the control response is temporarily rate-limited.
• Deterministic guard rails: every division guarded, every prediction input validated, deterministic fallback behaviour when inputs are invalid — and every forced hold, clamp and constraint request is logged with a cause code, so behaviour is auditable after the fact.

Part of Autonomous Mine Intelligence

Yard & Bin Inventory Optimisation is a component of the Autonomous Mine Intelligence layer: the coordination tier that reasons across the whole operation and sets targets for the regulatory controls beneath it. It works alongside conveyor-level applications such as ore profile scanning and constant volume conveyor control, which deliver the steady, measured flows the inventory layer plans around.