Turning sensor data into predictive maintenance intelligence at mine-site scale.
STRATUM is an industrial intelligence platform built for Goldfields' Ghana operations. It aggregates real-time sensor data from across the mining equipment fleet, models predictive maintenance windows using time-series ML, and delivers shift supervisors a live operational dashboard — converting reactive maintenance into a precision engineering discipline.
Unplanned equipment failures were generating unpredictable downtime events costing millions of dollars per incident in lost production. Maintenance schedules were calendar-based — following manufacturer service intervals that had no relationship to actual equipment wear, load conditions, or environmental stress at the mine site.
We deployed a network of edge computing nodes at each active site to ingest high-frequency sensor streams from 1,200+ equipment data points. A time-series ML model trained on historical failure data correlates sensor patterns with upcoming failure events, generating maintenance predictions with a 72-hour advance window — enough time to schedule work during planned production pauses.
Pilot phase (3 sites): 40% reduction in unplanned downtime across the monitored equipment fleet. Predictive windows accurate to within 72 hours across all sensor-equipped machinery. Full fleet deployment now underway.
"In the mining industry, every hour of unplanned downtime is a direct, measurable loss. STRATUM has given us something we genuinely did not believe was possible — the ability to see a failure coming three days in advance."