The same failure returns

The repair is done. Why does the same failure keep coming back?

Confirm the failure pattern, separate evidence from assumptions, and turn a repeat-failure investigation into actions you can verify.

Plant Floor · Practical maintenance management

Return to service and prevention answer different questions.

A repair can restore an asset’s function without explaining why it failed. That distinction matters when the same work keeps returning. The crew may have completed the assigned repair correctly while the condition that created the problem remains.

Before ordering another investigation meeting, define what is repeating. “This machine is unreliable” is too broad. “The same alarm appears after restart” is a symptom worth investigating, but it is not yet a confirmed failure mode. A repeated part replacement is evidence of repeated work, not proof that the replacement part caused each event.

Use the site’s risk and escalation process to decide the urgency and depth of investigation. Do not wait for a repeat count if the consequences require immediate action. This guide covers how to organize the learning, not how to inspect, test, or repair equipment.

Build a failure timeline before building a theory.

Pull the relevant work orders for the same asset and check them with the people who operated and maintained it. Look for evidence that distinguishes one event from another. Similar descriptions can hide different modes; inconsistent codes can hide the same one.

Event: when did the loss of function occur, and what was observed?
Context: what operating state or recent change was documented? Do not assume the condition was the same each time.
Work: what was actually repaired, adjusted, replaced, or left unresolved?
Evidence: which observations, records, or retained findings support the account? Have qualified personnel handle any physical evidence or testing under the site’s procedures.
Exposure afterward: how much comparable operation occurred before the next event? Calendar time alone can mislead if the asset barely ran.

Label entries as confirmed, reported, or unknown. The point is to make uncertainty visible before a persuasive explanation becomes the accepted cause. A missing close-out does not prove poor workmanship, and an operator’s recollection should not be silently treated as a measurement.

Choose an investigation that can handle the evidence.

A short causal chain may suit a focused 5 Whys discussion. Support each link with evidence, and stop when an answer becomes speculation. The number of questions is less important than whether the connection can be checked.

When several conditions may interact, map the possible branches before choosing one. A Fishbone can help a cross-functional team organize those possibilities. The diagram is an investigation plan, not proof. Each material hypothesis still needs an owner and a way to confirm or reject it.

If the question involves a new or changed design, bring engineering into a prospective risk review using the site’s method. Do not force every problem into whichever worksheet the team knows best. The RCA Method Picker is a starting point for that conversation.

A fictional repeat that is not solved by a familiar label

Original fictional scenario. This example was created for this guide. It is not a customer result or a scene from the book.

A conveyor has three recorded stoppages. Each work order is coded “sensor fault,” and each closes after a sensor replacement. The first explanation offered is that the sensors are unreliable. The timeline reveals a gap: nobody documented the condition of the removed devices, and two stoppages followed a changeover.

That pattern does not establish that changeovers caused the stoppages. It does justify checking the operating context alongside the component explanation. The team records two hypotheses, assigns the relevant evidence review to qualified people, and marks the actual failure mode unconfirmed. No one changes a setting or bypasses a protective function to make the theory fit.

The immediate management action is to improve the next record: capture the observed loss of function, operating state, approved work performed, and findings from the site’s diagnostic process. If an investigation later confirms a procedure gap, the response must identify the procedure change, its approval, who receives it, and how its use will be checked. Another “remind everyone” action would leave too much undefined.

Close the action loop, then verify the result.

Separate actions that restore function, actions that improve evidence, and actions intended to prevent recurrence. All can be useful, but they do different work. Buying a replacement part may restore service; it does not by itself show that the causal condition changed.

Link each preventive action to a supported finding.
Assign an owner, due date, and approval route.
Record where the change enters the work system: a job plan, procedure, material specification, monitoring task, or engineering action.
Define the evidence needed to confirm implementation.
Set a follow-up based on relevant operating exposure and the failure mechanism, with qualified input.

Keep “action completed” separate from “effectiveness verified.” An asset that has not run since the change cannot demonstrate improved reliability. A longer interval is encouraging, but changing duty or low exposure may also explain it. If the failure returns, reopen the assumptions with the new evidence instead of simply repeating the same action.

Make the next close-out useful to the next planner.

Use the Close-Out Quality Rubric to improve the record. The Repeat-Failure Watch List helps keep selected assets, owners, and follow-up visible. For a supported linear chain, the 5 Whys Worksheet provides a place to record the reasoning.

The RCA Close-Loop Checklist can structure the action handoff. Adapt investigation triggers and verification timing to the asset’s consequences and operating context rather than treating a template’s thresholds as universal.

Feed confirmed changes back into the job plan, then check that the revised plan reaches the next job. If recurring repairs are also displacing planned work, use the weekly-schedule guide to examine that handoff without confusing an interruption with its underlying cause.

Source notes

This original guide develops themes from Fix the System, Chapters 9 and 13, with the Chapter 16 watch-list tools as supporting artifacts. The US Department of Energy’s Maintenance Assessment Plan connects failure analysis to corrective action, maintenance history, and feedback into maintenance programs. Published in 2004 for facility assessment, it supports that management loop; it does not supply a universal recurrence threshold or an equipment repair procedure.