Skip to content

Maintenance

Agent: the predictive maintenance agent (predictive_maintenance in ExecuteLoop), with the EQUIPMENT_LIMIT_APPROACH evidence document and the AnalyzeTelemetry API call. Its cards wait on Actionable Twins under the maintenance play.

What it does

It reads equipment sensor logs, says how far each asset is from the limit its feed states and in which order to act, and drafts the intervention for a maintenance planner. Its jobs:

  • Trend to a stated limit. Reports the margin left and, when the readings move steadily towards the limit, when a straight line through the first and last readings would reach it.
  • Prioritise. Orders the work by safety first, then by the impact on dependent lines and shipments.
  • Plan the work. Drafts the work order, a maintenance window in the nearest planned downtime, and a spare-parts request when stock is below the reorder point your inventory feed states.

What it reads

Equipment readings such as vibration, bearing and oil temperature, pressure and battery state; the limits the feed or maker states; maintenance logs; production schedules; and spare-part stock levels. It can run read-only queries against the connection a run was started with.

What it produces

  • The EQUIPMENT_LIMIT_APPROACH evidence document, with the readings it used, so a technician can redo the sum. See Evidence documents.
  • Action cards with a priority and a proposed intervention.
  • Draft work orders and parts requests.

Human oversight

  • A maintenance planner or supervisor approves every repair, shutdown, parts order and schedule change.
  • The agent never shuts equipment down, overrides an alarm or orders parts itself.

Model

The deployment’s general reasoning tier.

The limit-approach document is fixed arithmetic over cited readings. Nothing is fitted and nothing is forecast.

Where it runs and data handling

On the deployment’s own model plane: Runink’s shared machines on a Trial, your own cloud on Cloud and your own servers on On-premises. No outside AI service is called on any plan. See Facts shared by every card.

Guardrails

  • Uses only the limit your feed or the maker states. It never substitutes an industry norm or another asset’s limit.
  • Refuses to override safety shutdowns, alarms or hazardous-materials procedures.
  • Refuses to alter or delete maintenance and compliance records, or sensor and telematics data.
  • Doesn’t neglect the safety needs of particular sites or regions when it ranks work.

How to read its results

  • A trend is not a forecast. It gives no failure probability, confidence score or remaining useful life. A straight line through two readings supports none of them.
  • With no stated limit, or too few readings, it writes no limit card and says why. When readings wander instead of trending, it says the crossing can’t be projected.