🔧Preventive & Predictive Maintenance

Preventive maintenance schedules

A preventive-maintenance (PM) schedule automatically generates a work order at a defined interval. Intervals can be calendar-based (every 30 days), runtime-based (every 500 hours), cycle-based (every 10,000 cycles), distance-based (for vehicles), or condition-based (when a sensor crosses a threshold).

Most facilities run on a mix: calendar-based for time-driven jobs (annual inspections, quarterly safety audits) and runtime-based for usage-driven jobs (oil changes, filter swaps, bearing greasing). The key benefit of runtime-based is that you stop doing PMs on machines that have not actually been running, and you start doing them on machines that have run far more than expected.

1. Go to Schedules in the sidebar.
2. Click "+ New Schedule".
3. Pick the machine (or a group of machines — a single schedule can cover an entire fleet).
4. Choose the trigger: Calendar, Runtime, Cycles, Distance, or Condition.
5. Define the interval (e.g. "every 30 days at 06:00", or "every 500 spindle-on hours").
6. Set lead-time (how many days before the trigger should the work order be created? Default 7 days).
7. Author the task checklist — what should the technician actually do? Each step can require a photo, a measurement, a pass/fail, or a signature.
8. Optionally attach a parts list and required tools (the parts will pre-allocate from stock when the WO fires).
9. Save. The next-due date appears immediately.

Predictive maintenance with AI

When sensors are connected to a machine, Myncel's AI engine learns the machine's normal behavior over a 7–14 day baseline period. After that it watches in real time for deviations: a vibration signature creeping up, a motor running hotter than usual, current draw drifting outside the normal envelope, a pump's discharge pressure drifting downward.

When the AI sees a meaningful change it raises a predictive alert and (optionally) auto-creates a work order. Predictive alerts include the machine, the sensor, what changed, the confidence level, and an estimated time-to-failure window when one can be calculated. The alert links to a chart so the technician can see exactly which signature triggered it.

The AI is not a black box. Every prediction is explainable — clicking "Why this alert?" shows the contributing signals, the historical baseline, and the deviation magnitude. Technicians and reliability engineers learn the machine alongside the model.

For the full configuration walkthrough — choosing between the Statistical, Hybrid, and LLM-Assisted models; tuning sensitivity (the slider maps linearly onto a sigma threshold, with 50 = 3σ as the default SPC standard); setting per-machine overrides; and reviewing detections in the confirm/reject feedback loop — see the dedicated AI & Predictive Maintenance chapter.

• No data-science expertise required — baselines are automatic; sensitivity is a single slider per organization with optional per-machine override.
• Three model kinds: Statistical (rolling z-score + EWMA, default), Hybrid (statistical + rule-based context), LLM-Assisted (statistical + language-model annotated recommendations).
• Configurable in two places: workspace defaults at /settings/ai, per-machine override on the equipment detail page → 🤖 AI tab.
• Available on every plan from Starter upward — what changes by plan is the LLM-call quota for the LLM-Assisted model (Statistical and Hybrid have no per-call cost).
• Models supported include vibration ISO 10816 zoning, electrical-signature analysis, thermal drift, pump cavitation, compressor surge, bearing fault frequencies, and gearbox mesh-frequency analysis.

PM checklists and digital forms

Each schedule can include a structured checklist. Technicians complete it on mobile or web — text fields, dropdowns, photos, signatures, measurements with units, and pass/fail items are all supported. Completed forms are saved permanently against the work order and can be exported as PDF for audits.

Conditional logic is supported: "If Q3 is Fail, then show Q3a-Q3c". This keeps the form short for the common case but captures detail when it is needed (e.g. the failure path on a 100-step electrical inspection).

Reliability metrics — MTBF, MTTR, OEE, PM compliance

As soon as you have a few weeks of work-order history, the standard reliability metrics start showing real numbers. They are computed automatically per machine, per group, per facility, and across the whole organization, and update in near real-time.

• MTBF — Mean Time Between Failures, computed from corrective work orders only.
• MTTR — Mean Time To Repair, computed from In-Progress timer minus On-Hold pauses.
• PM Compliance — completed-on-time PMs / total PMs in the period.
• OEE — Overall Equipment Effectiveness (Availability × Performance × Quality), where Availability comes from runtime data, Performance from cycle-time vs design, and Quality from rejected-part counts (manual entry or MES integration).
• Backlog age — open work orders older than X days, useful as a leading indicator of pain.