Back to all posts
Plant Data7 min read·October 25, 2026

How to Capture Machine Downtime Automatically

TB
ThynkBlox Automation Team
Automation Engineering

To capture machine downtime automatically, use machine-state signals to record when production stops and resumes, then ask operators to confirm the reason. Agree what counts as a stop, retain short events, and keep automatic suggestions separate from confirmed causes. The record should distinguish what the machine reported, what someone observed, and what remains unknown.

What should the machine record, and what should operators enter?

Give the machine the timing job and give people the explanation job.

Specify an event record with the equipment identity, start and end times, observed state, signal quality, and relevant production context. Keep reason entry separate so an unanswered prompt does not prevent a stop from being recorded.

Ask operators to select a reason and add a brief note where needed. Allow “unknown” and “awaiting investigation” as honest answers. Do not force someone to choose a plausible category just to close a screen.

Record elementProposed ownerRecording rule
Stop start and endAutomatic captureUse the agreed machine-state signals
Production contextConnected system or authorised userRetain the job or product active during the event
Suggested reasonDefined automatic ruleLabel it as a suggestion
Confirmed reasonOperator or designated reviewerRecord who confirmed it
Root causeAssigned investigatorKeep it separate from the initial reason
CorrectionsAuthorised userPreserve the previous value and explanation

A useful initial reason is “waiting for material”. It need not pretend to explain why the material was unavailable.

Which signals should define a stop?

Start with a written definition of productive operation for each machine type. Then identify which available signals support that definition.

Candidate inputs include a running state, cycle completion, production count, operating mode, fault state, and upstream or downstream availability. Treat these as inputs to an agreed rule rather than interchangeable proof of production.

For example, decide how to classify a machine that reports “running” but completes no cycles. Also decide whether a machine awaiting its next scheduled job belongs in downtime reporting at all.

Write a state table covering:

  • Producing.
  • Stopped during scheduled production.
  • Setup or changeover.
  • Planned non-production.
  • State unknown because valid data is unavailable.

Keep “unknown” visible. Specify that a lost connection must not silently become either running time or stopped time.

Select the stop rule separately from the connection method.

How should you set the stop threshold?

Separate the rule that detects an interruption from the rule that requests an explanation.

Use a detection rule that rejects invalid signal changes. Then apply an agreed duration threshold to decide whether the event needs an operator reason. Choose that threshold against the machine's normal cycle and operating sequence.

Define what happens when the threshold is crossed: retain the original stop time, rather than starting the clock when the reason prompt appears.

Illustrative example: a plant chooses a 60-second threshold for requesting a reason. A machine stops for 90 seconds. The record contains the full 90-second event; the prompt becomes due after 60 seconds.

Also define the restart rule. If a signal briefly changes to running during a continuing interruption, decide what evidence is required before closing the event. Do not let incidental signal changes rewrite the history without a documented rule.

What should happen to short stops?

Retain short stops even when they do not require individual explanations.

Give them a separate review view showing event count, total duration, and distribution by machine or production context. Mark them as short stops without confirmed reasons unless someone reviews them.

Avoid prompting operators after every brief interruption. Instead, agree when a repeated pattern deserves attention and provide the underlying event list for review.

Keep the reporting treatment explicit. If short stops sit outside the plant's downtime category, show where their time goes and how the chosen measure handles it. Our article on what OEE hides is a useful companion when agreeing those reporting definitions.

How should overlapping causes be recorded?

Use one elapsed-time record for the equipment stop and attach reasons to the relevant parts of it. Do not add overlapping reason durations together as though they were separate machine losses.

Illustrative example: a machine is stopped for ten minutes. A material shortage applies throughout the event. During the final four minutes, maintenance also investigates a sensor issue.

Record ten minutes of machine downtime. Attach both observations, including their time ranges. If the reporting system requires one primary reason, apply an agreed allocation rule and retain the secondary observation.

Distinguish these situations:

  • Consecutive reasons: one condition ends and another keeps the machine stopped.
  • Concurrent conditions: more than one condition exists during the same interval.
  • Uncertain sequence: the available evidence does not establish which condition came first.

For consecutive reasons, split the stop into linked segments if the transition is known. For concurrent conditions, retain the overlap. For uncertain sequences, preserve the uncertainty rather than inventing precise boundaries.

At line level, define the equipment boundary too. Keep individual machine events, but calculate the line's stopped time from the agreed line state instead of summing every machine's downtime.

When can the system suggest a reason?

Allow suggestions when there is a documented rule and the supporting signal is available.

A fault signal might produce “drive fault reported” as a suggested reason. That wording records the observation without claiming that the drive caused the entire interruption.

Keep three fields distinct:

  1. Observed condition: the signal or state captured.
  2. Suggested reason: the interpretation proposed by a rule.
  3. Confirmed reason: the explanation accepted after review.

Show pending suggestions as pending in reports. Do not convert them into confirmed reasons when a shift ends or an operator leaves a prompt unanswered.

Use alarm history as supporting evidence. Keep the first alarm, the operator's reason, and the eventual root cause as separate records unless a review establishes their relationship.

How should corrections work?

Provide a controlled way to correct reasons, event boundaries, and production context.

For each correction, retain the original value, replacement, person making the change, and explanation. Preserve the original signal history so a reviewer can distinguish a changed interpretation from changed evidence.

If someone splits or combines events, retain the relationship to the original records. Require an explanation for changing captured start or end times.

Define how reports handle later corrections. A previously issued report should either remain an identifiable snapshot or clearly indicate that its underlying records have changed.

What common mistakes should you avoid?

The main mistakes come from asking the record to look more certain than the evidence allows.

  • Treating every non-running state as production downtime. Define schedule and operating-mode rules first.
  • Discarding events below the reason threshold. Retain them for short-stop review.
  • Using the first alarm as the confirmed cause. Preserve it as an observation pending review.
  • Counting overlapping reasons twice. Reconcile reasons against the equipment's elapsed stop time.
  • Forcing a reason before facts are available. Allow an explicit unknown state.
  • Treating missing data as a machine state. Show the gap and resolve it separately.

Keep downtime capture within its monitoring purpose. Any proposed change to safety-related controls should be designed and reviewed with a qualified assessor.

What should you verify before rollout?

Agree acceptance criteria before choosing dashboard layouts.

Test a normal stop, a short stop, a restart that does not hold, consecutive reasons, overlapping conditions, missing data, and a corrected event. Include an unanswered reason prompt.

For each case, check the captured boundaries, threshold behaviour, reason status, elapsed-time totals, and correction history. Test communication loss and recovery, including duplicate or delayed event delivery.

Put these decisions into the functional design specification. Acceptance should depend on explainable records that match the agreed scenarios.

Related reading: reconciling machine and production report counts and structuring downtime reason codes.

Frequently asked questions

Can downtime capture be fully automatic?

Specify automatic capture for event timing and machine observations. Automate reason confirmation only where the evidence and acceptance rules justify it. Leave other explanations pending for a person to review.

Should operators enter start and end times?

Use captured times as the default. Provide manual entry for missing records and authorised corrections, with an explanation and a visible history.

What if operators disagree about the reason?

Retain both observations and assign a reviewer. Keep the reason unresolved until the evidence supports a decision.

Should every machine use the same threshold?

Choose thresholds for the relevant operating sequences, then document them. Use consistent definitions across reports so readers can understand how events were classified.


Planning downtime capture for your plant? Explore our automation services or contact us to discuss machine signals, reason entry, and acceptance criteria.

Next step

Want to apply this to your plant?

Every project starts with a fixed-fee assessment, and you can stop after the report.