← Back to product Smart Instrument · Neural Network

Rock Analyzer.

Feed sizing · conveyor-mounted · edge inference  ·  also known as RockSense

The Rock Analyzer is a non-contact, camera-based particle-sizing system. Mounted above the SAG feed conveyor, it sees every parcel of ore as it passes and reports the full particle size distribution — fines through top size — every few seconds, with no sampling, no lab delay, and no contact with the ore.

What it measures

A purpose-trained neural network segments each frame, identifies individual rocks, and returns the live PSD. Lab-calibrated against wet-sieve data with a Rosin-Rammler fit overlay, the instrument is retrainable on any ore type without a hardware swap.

  • F80, F50, top size — the percentile diameters the metallurgist actually uses
  • Fines fraction — per-frame counts and area share for the <25 mm class
  • Coarse / oversize alerts — surges and primary-crusher gap drift flagged before they reach the SAG
  • Frame-rate scaling — CPU or GPU, INT8 quantisation for edge boxes

How it changes the circuit

The full feed PSD becomes a real-time signal the controller can act on. The MPC pre-positions feed rate, mill speed, and water for the incoming size, instead of reacting after the mill has already responded. The digital twin uses the same measurement to keep its breakage model anchored to the real ore, so soft-sensor predictions stay honest as the geology drifts.

Deployment

On-edge inference (CPU or GPU), pixel-to-mm calibration, configurable region of interest, and an annotation studio workflow for retraining on site. No cloud dependency, no per-prediction licensing.

See how this instrument performs on your circuit.

Request a demo