The full power of TDengine, now free forever for up to 5,000 tags.

Explore More

How to Build a Reliable Machine Downtime Record

Juno Qiu

November 9, 2026 /

Machine downtime analysis starts with a question that sounds simple: how long was the machine unavailable when it was supposed to be producing? Answering it reliably takes more than adding up alarm or event durations. The records may overlap, a machine state may be unclear, or an operator’s reported reason may still need investigation.

A reliable downtime record connects the equipment, the relevant production schedule, the time interval and the available evidence. It also separates the duration you can count from the reason you think the downtime occurred. This gives a production engineer a defensible starting point for later analysis.

Define downtime against the production schedule

Start with one machine and one reporting period. Agree on which equipment states count as downtime under your site’s definitions and when the machine was scheduled to produce. A stopped machine during a planned shutdown may need a different classification from an unplanned stop during a scheduled run. The distinction depends on the site’s rules and the question the report is meant to answer.

Document the schedule and the version or source used for it. If an event begins before the scheduled period or ends after it, only the portion that falls within the chosen reporting period belongs in that period’s scheduled-time calculation, subject to the site’s agreed rules. If the schedule changes, keep enough context to explain which schedule was used.

A state signal is evidence, not a complete downtime definition. A value labeled idle might indicate a normal pause, an unavailable machine or a signal that has not updated. Check the meaning with the people who operate and maintain the equipment. Missing or stale data should be marked as uncertain rather than treated automatically as either running or stopped.

Separate state, events, duration and cause

These terms answer different questions:

  • Equipment state: What did the available signal indicate, such as running, idle or faulted? The state depends on its source and interpretation.
  • Downtime event: What time-bounded occurrence has been recorded for review? A recorded event may be based on a state or another condition, but it still needs to meet the site’s downtime definition before it is counted.
  • Counted downtime duration: How much unique time qualifies for the chosen machine and reporting period after applying the agreed boundaries and handling overlaps?
  • Reported reason or reason code: What explanation was entered or assigned to classify the event? It may be operator-supplied, provisional or unknown.
  • Confirmed physical root cause: What underlying physical cause has been established through an appropriate engineering investigation? A reason code or time correlation alone does not establish it.

A trustworthy total can exist while the cause remains unresolved. Conversely, an apparent reason does not make the duration accurate if the timestamps or schedule are wrong.

Build a record that can be checked

For each event, retain the machine identity, source, start and end timestamps, time zone and relevant schedule. Record how the machine state maps to the downtime definition. If the event is still open, make that clear and state the cutoff used for any provisional calculation. Use a consistent convention for time boundaries, especially when adjacent events meet at the same timestamp.

Keep the original event records and references alongside any reconciled result. For a corrected timestamp or changed reason, preserve the original value and the source of the correction according to your site’s recordkeeping practices. This lets another engineer trace a reported total back to the evidence.

Record unknown reasons as unknown. If an operator reports a suspected cause, retain it as a reported observation, together with its source and review status. Do not replace uncertainty with a more specific explanation simply to complete a report. A later investigation can confirm, revise or leave the reason unresolved.

Avoid counting overlapping downtime twice

Consider a hypothetical production line scheduled from 13:00 to 14:00. Two records for the same machine show these intervals:

RecordStartEndRecorded duration
A13:0013:2020 minutes
B13:1513:4530 minutes

First, the production engineer checks that both records represent downtime under the site’s definition and belong to the same machine and reporting period. The records overlap from 13:15 to 13:20. Adding their durations gives 50 minutes, but that counts the five-minute overlap twice.

Together, the qualifying intervals cover 13:00 to 13:45, or 45 minutes of unique downtime. Preserve both original records and their sources; the reconciled duration is a calculation derived from them, not a replacement for them. This example assumes the timestamps are reliable and the whole interval is within scheduled production. If either assumption is uncertain, resolve or flag it before presenting a final total.

Keep the total separate from reason attribution

Suppose A and B carry different reported reasons. The 45-minute total does not tell you which reason caused each minute. Both conditions may have been present during the overlap, and neither code by itself proves the physical cause.

Use a documented site policy for any reason-based reporting, including how to treat concurrent reasons and unresolved time. Explain that policy beside the result, retain both source records, and do not present an attribution rule as proof of causation. If the site has no agreed rule, report the unique duration and leave the overlapping attribution unresolved until the right people review it.

Review the record before using it

A practical review with operators and the appropriate data owner should cover the following:

  1. Scope: Is the machine identity correct, and does the schedule match the production period being analyzed?
  2. Boundaries: Are timestamps consistent in time zone and precision? Were open events, late data and intervals crossing a shift boundary handled consistently?
  3. Evidence: Do state changes and source records support the interval? Are gaps, stale readings or conflicting observations visible?
  4. Overlap: Have qualifying intervals for the same machine been reconciled without deleting the original records?
  5. Reasons: Are unknown, reported and confirmed explanations clearly distinguished, with any attribution policy documented?

When an interval cannot be established confidently, flag the affected period and explain the limitation. An estimate may be useful for discussion, but it should not look like a fully verified record.

Use event data as a starting point for analysis

An event-capable industrial data platform can help organize the evidence. For example, TDengine IDMP Events documents event records with start and end times and other context. Its real-time analysis trigger documentation describes state windows that respond to changes in a configured state. Those capabilities can contribute source information for a downtime record when the data and rules are appropriate.

The engineering team still needs to define qualifying downtime, check data and reconcile intervals against the schedule. Event and state-window support alone does not establish that overlapping intervals are automatically merged, that a production calendar is applied, or that a reported reason is the correct physical cause.

Once the record is reviewed, it can support comparisons and later investigations, including AI-assisted work. Those investigations remain dependent on the quality and context of the underlying records. For a related look at moving from a performance measure into investigation, see how engineers move from OEE monitoring to root-cause analysis.

Start with one machine and one scheduled period. Agree on the definition, review the source records with the people closest to the equipment, and make uncertainty visible before using the total to guide a decision.

Related Articles