Glossary
Predictive maintenance
Using data and machine learning to forecast when equipment is likely to fail, so maintenance happens before the failure instead of after it or on a fixed schedule.
Traditional maintenance is either reactive (fix it when it breaks) or preventive (service it every N kilometres or months, whether it needs it or not). Predictive maintenance uses the equipment's own history — work orders, sensor and telematics data, usage patterns — to estimate the likelihood of a specific failure in a given window, and schedules work accordingly.
The benefits are fewer unexpected breakdowns, less downtime, longer asset life and lower cost, because healthy assets are not over-serviced and failing ones are caught early. The hard parts are data quality (work orders are messy), choosing a prediction target that crews can act on, and earning trust by being right in a way operators can verify.
Why it matters to our work
This is the problem our forecasting engine was first proven on, in the City of Kitchener fleet pilot, before the same discipline was applied to brain health.