Back to all posts
Control Systems7 min read·October 25, 2026

How Should a Machine Restart After a Power Failure?

TB
ThynkBlox Automation Team
Automation Engineering

A PLC machine restart after power failure should follow an agreed recovery sequence based on the machine's physical state and the process step that was interrupted. Treat retained values as information to validate, rather than permission to resume motion. Define how to recover uncertain position, handle incomplete cycles and obtain operator acknowledgement; safety-related restart decisions require a qualified assessor.

What should happen when power returns?

Separate restoring power from authorising production. In the restart specification, describe what the controller must establish before the machine becomes available to run.

Start with four questions:

  • What is the machine's physical condition?
  • Which process step was interrupted?
  • Which stored values can still be trusted?
  • What action is required before recovery or production may begin?

Avoid a single rule such as “resume the previous cycle”. That leaves too much undefined. Resuming an inspection, recovering a transfer mechanism and deciding what to do with partly processed material are different decisions.

Ask a qualified assessor to determine the safety-related restart requirements, including whether automatic restart is permissible for the application. Keep that decision separate from the production preference to recover quickly.

Which retained PLC values should be trusted?

Decide retention value by value. For each item, record why it is needed after an interruption, how its validity will be checked and what happens if the check fails.

Do not assume that selecting retention establishes the complete behaviour. Verify the actual controller and device configuration for the interruption conditions included in the project scope.

ValueQuestion to resolveProposed recovery rule
Recipe selectionIs it still the approved recipe for the material present?Confirm the selection and validate parameters before use
Process stepDoes the stored step agree with the observed machine state?Treat it as a recovery clue until checks pass
Part countWas the interrupted part already counted?Reconcile the count with the defined completion event
Timer valueShould interrupted time still count towards the process?Define whether to preserve, restart or invalidate it
Position valueDoes it still represent the mechanism's actual position?Validate position before permitting dependent movement
Start requestWas this a command issued before the interruption?Define how stale requests are cleared or rejected

Keep configuration and production records distinct from commands. A recipe may need to remain available while an earlier start request should not become a fresh command merely because communications return.

Specify what the operator sees when a retained value is missing, invalid or inconsistent. A default value should not silently stand in for an unknown process condition.

How should lost position be recovered?

First define what counts as known position. For each mechanism, identify the evidence required before movement is permitted: device status, position feedback, reference sensors or another agreed check.

Then describe the recovery route when that evidence is unavailable.

Do not write “home the axis” without describing the starting conditions. Consider the intended path, material in the machine, connected mechanisms and any obstruction that must be resolved. Have safety-related recovery movements assessed by a qualified assessor.

The specification should answer:

  • What makes position valid or invalid?
  • Which movements depend on valid position?
  • What checks precede a reference movement?
  • What happens if the expected reference is not reached?
  • How does the operator leave an unsuccessful recovery attempt?

Where position cannot be established automatically, define an authorised manual recovery procedure. Make its limits explicit, including when maintenance involvement is required.

What should happen to an incomplete cycle?

Define completion at the process level, rather than relying only on the last stored step number.

For each interrupted step, choose a disposition: resume, repeat, reverse through an approved recovery sequence, remove the material, or hold it for a decision. Record the conditions for that choice.

Ask whether repeating the action would be acceptable. Would it add material twice, repeat a treatment or issue a second production record? Also ask what evidence would establish that the action finished before power was lost.

Where completion cannot be established, specify an “outcome unknown” state. Give that state a clear resolution procedure instead of forcing it into either complete or incomplete.

Include product quality in the decision. Define whether an interrupted process needs inspection, rework or rejection, and identify who may authorise that disposition. A machine ready to move and material ready for further processing are separate acceptance conditions.

What would a restart decision look like in practice?

Illustrative example: consider a machine that clamps a component, performs an operation and transfers the component out.

The proposed restart rules might be:

Observed condition after restorationRequired decision before production
No component present; mechanisms in verified starting positionsComplete readiness checks and accept a fresh start request
Component present; operation confirmed completeValidate clamp and transfer conditions before an approved transfer
Component present; operation outcome uncertainHold for inspection or an authorised disposition
Transfer stopped between verified positionsRecover position through the defined procedure
Stored step conflicts with observed conditionBlock normal sequencing and explain the mismatch

These are illustrative specification choices, not a universal control sequence.

For the uncertain-operation case, the operator message should describe the problem and the next action: “Operation completion is unknown. Inspect the component and select its disposition before recovery.”

That is more useful than “Restart fault”, because it identifies the unresolved decision.

What should operator acknowledgement mean?

Give acknowledgement a specific purpose. It might confirm that the operator has inspected the work area, identified the material present or selected an approved recovery route.

Avoid one button that acknowledges an alarm, resets a fault and starts movement without making those consequences clear. Define each action and its acceptance conditions.

For each acknowledgement, state:

  • What the operator is confirming.
  • Which checks the controller must complete independently.
  • Whether the action enables recovery, production or neither.
  • What further command is required.
  • What invalidates the acknowledgement.

Do not use acknowledgement to replace a missing position check or an unresolved process result. The interface should show what remains unknown and how to resolve it.

The restart assessment should also address remote commands and queued requests. Define which requests remain valid after restoration and how a fresh command is distinguished from an old one.

For related interface decisions, see why operators ignore alarms. Apply the same discipline here: explain the condition, its consequence and the required response.

What common mistakes should the specification prevent?

Retaining the step and assuming the machine agrees. Require agreement between the stored sequence state and the physical evidence needed for that step.

Resetting everything to the first step. Define what happens to material and mechanisms already partway through a cycle before allowing the first step to run.

Treating every stop as equivalent. Review power interruption, controller restart, device restart and communications recovery separately. Record where they share a recovery route and where they differ.

Using an alarm reset as restart permission. State what reset clears, what remains blocked and what authorises the next action.

Designing only for an empty machine. Include recovery with material present, an incomplete operation and uncertain position.

Making safety decisions through production logic alone. Have a qualified assessor determine the safety-related requirements and the evidence needed to verify them.

How should restart behaviour be specified and tested?

Write a recovery table for each relevant process step. Include the interruption point, expected physical condition, retained information, required checks, permitted actions and operator message.

Put these decisions in the functional design specification. Agree them before commissioning.

Use this acceptance checklist:

  • Interrupt each relevant step under an approved test procedure.
  • Check behaviour with material present and position uncertain.
  • Verify retained values against the intended recovery rules.
  • Check that old start requests are handled as specified.
  • Confirm that unresolved conditions block the relevant actions.
  • Check operator messages and unsuccessful recovery routes.
  • Record the observed result against the acceptance condition.

Coordinate interruption testing with the responsible engineering team and qualified assessor. Define the test conditions and controls before deliberately removing power.

Frequently asked questions

Should a PLC automatically restart after power returns?

Only where the application's assessed requirements permit it. Define the required physical state, process conditions and authorisation explicitly, with safety-related decisions made by a qualified assessor.

Is retaining the cycle step enough?

No. In the proposed recovery design, validate that step against position, material condition and evidence of process completion before using it to resume sequencing.

Must every machine return to its starting position?

Determine that for the application. Specify when reference recovery is required and assess the recovery movement itself.

Does operator acknowledgement make restart acceptable?

Acknowledgement resolves only the confirmation assigned to it. Keep position checks, process disposition and safety-related restart requirements explicit and separately verifiable.

Next step

Want to apply this to your plant?

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