Skip to main content

Get Started with Predictive Maintenance

Use this guide when a site is preparing its first Predictive Maintenance workflow in FactVerse. The goal is to make equipment health, anomalies, maintenance recommendations, and work-order feedback usable by operations teams without hiding the data and review assumptions behind the model.

Predictive Maintenance works best as a governed operations loop: equipment and signals are connected first, health is reviewed in context, anomalies are triaged by a responsible owner, accepted recommendations become maintenance work, and field outcomes feed the next tuning cycle.

Setup Flow

Prerequisites

Before the first rollout, confirm:

RequirementWhy it matters
Equipment ownerSomeone must approve the equipment scope and review health or anomaly output.
Data ownerSource-system access, signal meaning, and data quality issues need a named owner.
Maintenance ownerAccepted advisories need a team that can create, assign, and close work.
Initial equipment scopeA small group of similar assets makes baseline review and operator feedback easier.
Work-order feedback pathOutcomes need to return to the workflow after field action.

Step 1: Select the Equipment Scope

Start with equipment where maintenance ownership, signal data, and failure history are clear enough for review.

DecisionGuidance
Equipment classGroup similar assets such as chillers, pumps, motors, compressors, AHUs, UPS systems, fans, filters, or production utilities equipment.
Operating boundaryDefine the site, building, production line, room, equipment group, or service system covered by the first rollout.
Maintenance ownerAssign a team that will review anomalies, accept or reject advisories, and close work orders with outcomes.
Decision horizonDecide whether the first use case is daily triage, weekly maintenance planning, reliability review, or a pilot dataset validation.

Avoid starting with every asset in a site. A smaller set with stable identity and usable history creates better early evidence.

Step 2: Prepare Identity and Signals

Predictive Maintenance needs a stable link between each equipment record and its source signals.

InputRequired preparation
Equipment registryEquipment ID, name, class, location, manufacturer, model, owner, and optional bridge to the main equipment directory.
Sensor or telemetry historyTimestamp, equipment reference, signal name, unit, value, and quality status.
Work-order historyEquipment reference, failure type when available, open/close timestamps, priority, root cause, action, and outcome.
Inspection and alert evidenceInspection findings, alarms, comments, photos, and event timestamps that can explain the operating context.

For connector setup, mappings, and source data quality, use Prepare Signal History for Predictive Maintenance.

Step 3: Enable the Operating Surfaces

The current application exposes these primary views:

ViewRouteUse
Dashboard/pdm/dashboardReview health distribution, active anomalies, maintenance status, and high-level operating signals.
Equipment list/pdm/equipmentFind monitored assets and open profile detail.
Equipment detail/pdm/equipment/:idReview profile, health snapshot, health history, and equipment context.
Fleet/pdm/fleetCompare equipment groups and component attention signals.
Anomalies/pdm/anomaliesTriage unresolved anomalies and generate maintenance action where appropriate.
Maintenance/pdm/maintenanceReview maintenance records and schedules connected to monitored equipment.
Criticality/pdm/criticalityPrioritize work using risk and criticality signals.
Energy/pdm/energyReview energy baselines where energy data is connected.
Health/pdm/healthReview health-centered operating views where enabled.

Use pdm in route and API references. In customer-facing prose, call the capability Predictive Maintenance.

Step 4: Validate the First Equipment

Before adding more assets, validate one representative equipment record end to end:

  1. Open the equipment profile and confirm identity fields.
  2. Confirm recent signal history is mapped to the right equipment.
  3. Review the latest health snapshot and health history.
  4. Check whether the dashboard includes the equipment in summary counts.
  5. Open the anomaly list and confirm unresolved events appear with useful context.
  6. Create or link a maintenance action only after the advisory has been reviewed.
  7. Close the linked work order with an outcome label.

This validation prevents a common failure mode: building a model workflow before the team trusts the underlying asset and signal identity.

Step 5: Define Review Cadence

Predictive Maintenance should become part of an operating rhythm.

CadenceReview focus
Daily shift reviewNew high-severity anomalies, blocked advisories, and urgent equipment status changes.
Weekly maintenance planningAccepted advisories, work-order backlog, criticality ranking, maintenance windows, and parts readiness.
Monthly reliability reviewHealth trend, repeated failure modes, advisory precision, false positives, and rule or template changes.
Pilot acceptanceData completeness, reviewer confidence, work-order outcome capture, and expected next rollout scope.

Expected Output

At the end of the first setup pass, the team should have:

  • a defined equipment scope;
  • mapped signal history for the selected assets;
  • at least one reviewed equipment profile;
  • dashboard and anomaly views that reflect real source data;
  • a named owner for advisory decisions;
  • a work-order feedback path for accepted recommendations;
  • a short list of data or readiness gaps before broader rollout.

Validation Checklist

  • Equipment identity is stable across Predictive Maintenance, DFS, CMMS, and work-order records.
  • Source signals have expected units, timestamps, and equipment mappings.
  • The first equipment profile has health history or a clear no-data state.
  • Anomaly review has a named owner and a response path.
  • Accepted advisories can create or link to work orders.
  • Work-order closure captures outcome evidence.
  • Dashboard and Agent summaries match the underlying records.

Troubleshooting

SymptomCheck
Equipment is missing from the dashboardTenant scope, equipment profile, source mapping, and route permissions.
Health view shows no dataRecent readings, timestamp format, unit mapping, and equipment ID bridge.
Anomaly list is emptySignal coverage, threshold or baseline readiness, and selected time window.
Advisory cannot become workWork-order permission, CMMS configuration, and maintenance ownership.