Predictive maintenance is useful when a detectable change gives the plant enough time to make a worthwhile intervention. Scheduled preventive maintenance fits tasks with a justified time or usage basis. Corrective maintenance can be a deliberate choice when the consequences of waiting for failure are acceptable. NASA’s reliability-centered maintenance guidance includes all three approaches and also recognizes that some problems require design changes. NASA RCM guide.
The choice belongs at the failure-mode level. A pump’s bearing, seal and standby function need separate questions, even when they share an asset number. This guide proposes a practical review for maintenance, reliability and production teams: identify what can fail, establish the consequence, test whether there is useful warning, and document the task that follows. The examples are hypothetical, not reported plant results. The cited US guidance and Great Britain safety guidance provide technical context; local equipment obligations and approved procedures still govern each installation.
Define the failure before choosing the maintenance label
For this comparison, preventive means scheduled work triggered by elapsed time or use. Predictive means work informed by measured condition and its interpretation. Corrective means restoring a function after a fault has occurred. Terminology varies between organizations, so record the trigger on the work order as well as the label. The DOE guide distinguishes reactive, preventive, predictive and reliability-centered approaches in its maintenance chapter. DOE O&M guide, chapter 5.
Write a functional requirement before listing tasks. In a hypothetical packaging plant, “conveyor runs” is too broad. The required function might include moving packs at the agreed rate without damaging them. A conveyor that still moves but repeatedly damages packs has a functional problem worth examining. Ask production to define the unacceptable outcome in its own terms; maintenance can then connect that outcome to a component and a credible mechanism.
A useful review entry might read: “Drive bearing degrades; conveyor cannot maintain the required duty; upstream production must stop.” A separate entry could cover a failed local indicator that does not control the conveyor or communicate a safety status. The same assembly now has two decisions. Do not transfer the consequence assigned to the bearing into the indicator row without checking what that indicator actually does.
Keep task purpose distinct from scheduling convenience. Replacing a component during a shutdown does not, by itself, explain whether the replacement was triggered by age, observed degradation or an existing fault. The shutdown is the access opportunity. Record the reason for the intervention separately; otherwise, later analysis will compare unlike jobs under one preventive code.
NASA frames maintenance around system function and applicable tasks. SAE’s current JA1011 publication is explicitly scoped to organizations managing physical assets or systems, rather than a particular industry. Those are useful boundaries for a review across industries, although this worksheet is not a claim of full RCM compliance. NASA RCM principles, SAE JA1011 scope.
Before accepting a row, have an operator and a technician read it. They should recognize the failed function and the fault being discussed. If the wording could describe several unrelated events, narrow it before debating sensors or intervals. The resulting record should support the next shift’s decision without requiring the original author to explain it or reconstruct the circumstances from memory.
Use predictive maintenance when the warning supports action
A predictive proposal needs a chain of evidence: a specific degradation mechanism, a measurement that can reveal it, an interpretation the plant can trust, and time to complete the response. NASA’s discussion of degradation detection includes uncertainty in prediction and the interval between condition checks. A useful signal therefore needs more justification than its presence on a trend display. NASA RCM guide, section 4.1.
Consider a hypothetical process pump with suspected bearing deterioration. Ask which operating states make measurements comparable. A trend collected during changing speed or load should prompt a context check before anyone declares worsening mechanical condition. For the trial, record the measurement method, operating state and technician’s interpretation alongside each observation. The proposal should say what additional check would distinguish the suspected mechanism from another explanation.
Then work backward from the repair. Identify who reviews a finding, what confirmation is necessary, whether a spare is available and when production can release the equipment. Include the preparation needed to perform the work safely. HSE’s maintenance guidance calls for planning, competent people and safe working arrangements; the warning window has to accommodate that work. HSE maintenance guidance.
Make a trial acceptance record before collecting more data. Suggested fields are confirmed actionable findings, findings that did not justify action, failures the method missed, time available after detection, and the actual response completed. Treat these as questions to investigate, not targets borrowed from another plant. A correct diagnosis received after the repair opportunity has closed does not satisfy the proposed operational purpose.
The economic case needs the same discipline. Douglas Thomas examines the costs and losses associated with different maintenance techniques and the data needed to estimate them. His NIST report supports asking what evidence the plant needs before assigning a monetary benefit. It does not supply a guaranteed saving for this pump. NIST maintenance economics report.
For a local comparison, list sensor installation, analysis effort, inspections prompted by findings, repair preparation and ongoing support. Compare those with the consequence the proposed action could actually avoid. Keep speculative avoided losses visibly separate from observed costs. If the proposal depends on detecting every failure, ask what happens when the signal is absent or the system is unavailable. That unanswered case belongs in the decision record before expanding the program.
Keep preventive tasks that have a defensible purpose and interval
Scheduled preventive maintenance earns its place when the task addresses the mechanism and its timing is justified. The DOE guide describes time-based preventive work and the role of other maintenance approaches. Use that distinction to ask why this particular action occurs at this particular interval. “It has always been in the system” is a history of the schedule, not evidence for its effectiveness. DOE O&M guide, chapter 5.
In a hypothetical food packaging operation, a consumable replacement might fit a planned production break. The review still needs to establish what condition the replacement controls, which manufacturer instructions apply and whether the current duty differs from the assumptions behind the schedule. Do not infer a replacement interval from this example. The site’s actual equipment specification and operating evidence must supply it.
Separate routine servicing from scheduled restoration or replacement. A lubrication task should specify the correct material and method; a restoration task needs an explanation of what it restores. For either, ask the technician what findings are recorded. A completion tick proves that someone closed the job. It does not tell the next reviewer whether deterioration was present or the task changed the condition.
Inspection has a different purpose again. HSE says inspection needs and frequencies should follow risk assessment, and identifies manufacturer recommendations, operating conditions and experience as relevant inputs. In Great Britain, this guidance concerns work equipment safety; it is not a universal replacement schedule for every industry. HSE equipment inspection guidance.
An inexpensive part is not sufficient reason to replace it repeatedly. Ask whether replacement addresses the suspected failure, what access requires and what errors the intervention could introduce. HSE’s major-hazard maintenance guidance explicitly identifies human error during maintenance and incorrect recommissioning as concerns. Cost the complete job and specify the return-to-service checks, rather than comparing the spare’s price with the price of a sensor. HSE maintenance procedures.
For the next schedule review, select a small group of recurring tasks and retrieve their last completed records. Look for a task objective, interval basis, useful findings and acceptance criteria. Where the evidence is missing, assign an investigation; do not silently delete the job. Retain applicable mandatory work while the responsible role evaluates a change. This gives the reviewer a concrete outcome: confirm the task, improve its instructions, investigate its interval, or route a proposed change through the site’s approval process.
Accept planned corrective work only within clear limits
Run-to-failure can be an intentional maintenance decision. NASA explicitly includes repair after failure among the possible outcomes of its analysis. The permission is conditional, however: it comes from understanding the consequences of that failure. It does not follow from a low equipment price or an absence of maintenance history. NASA RCM guide, section 3.3.
Take a hypothetical convenience indicator in a noncritical utility area. Assume the review confirms that it has no safety, control or product-release role, that failure is obvious, and that its loss does not hide another fault. If an approved replacement can be fitted safely when needed, planned corrective work may be a candidate. Change any of those assumptions and reopen the decision. An apparently similar lamp indicating a protective function would require a different assessment.
Prepare recovery before accepting the strategy. Record the spare specification, where it is held, who can authorize the work and how the equipment returns to service. Distinguish deliberate run-to-failure from deferring a fault that has already appeared. The latter needs its own assessment of the present condition, permitted operation, responsibility and review point. Calling a backlog item “planned corrective” does not establish that continued operation is acceptable.
If the argument depends on a standby asset, ask for evidence that it will perform the required duty. Proposed review questions include its latest functional test, available capacity, changeover arrangement and shared dependencies. Two units on the same vulnerable supply need a different discussion from independent alternatives. HSE’s testing guidance emphasizes clear pass/fail criteria and the management of overdue work; apply that discipline to the evidence supporting the standby assumption. HSE maintenance, inspection and testing.
Hidden protective failures need particular care. A device that is only demanded in an abnormal event cannot be treated as an obvious nuisance failure merely because normal production continues. HSE’s major-hazard guidance specifically asks whether proof-check periods for safety-critical items are adequate and observed. The actual test and interval must come from the applicable safety arrangements. HSE maintenance procedures.
Where no effective maintenance task controls an unacceptable safety consequence, refer the problem for engineering and protection review. NASA includes redesign or redundancy as an outcome when maintenance cannot reduce failure and failure is unacceptable. Do not convert that unresolved case into run-to-failure simply to finish the worksheet. NASA RCM decision outcomes.
Apply a decision matrix to individual failure modes
Use the following matrix as a review aid. Every row is hypothetical and conditional; none authorizes a maintenance change. Replace the assumptions with site evidence before choosing a strategy. The structure draws on NASA’s function-and-consequence approach and HSE’s requirement to consider deterioration when deciding inspection needs. NASA RCM analysis, HSE inspection guidance.
| Failure mode | Consequence to establish | Detectable signal and action window | Constraints to verify | Justified strategy if assumptions hold |
|---|---|---|---|---|
| Process-pump bearing deterioration | Loss of required production duty | Repeatable condition evidence early enough to confirm and repair | Comparable duty, competent analysis, spare and safe access | Predictive monitoring plus the applicable routine servicing |
| Packaging-machine consumable degradation | Loss of the specified operating function | A supported usage relationship; monitoring adds little useful warning | Manufacturer basis, actual duty, replacement and restart controls | Scheduled preventive task with a reviewed interval |
| Noncritical convenience indicator fails | Tolerable loss with no hidden safety or quality role | Failure is obvious; recovery can occur within the accepted service limit | Correct spare, safe replacement and named owner | Planned corrective work |
| Standby utility unit fails to start | Required backup capacity is unavailable | Functional testing establishes readiness; production alone does not | Test coverage, changeover, shared dependencies and overdue tests | Failure-finding tests and an approved recovery arrangement |
| Protective function has an uncontrolled failure mode | Unacceptable safety consequence | No demonstrated effective maintenance warning or task | Safety assessment and adequacy of engineered protection | Engineering review; no default run-to-failure decision |
| Work equipment subject to a required inspection or test | Safety consequence addressed by the applicable requirement | Required inspection or test and any complementary condition evidence | Applicable jurisdiction, equipment rules and approved inspection scheme | Required work plus condition monitoring where justified; no automatic substitution |
Read across a row before reading down the strategy column. The bearing example supports a predictive candidate only after the measurement and response assumptions hold. The consumable example supports preventive work only when the proposed action and interval have evidence. Neither the cost of the component nor its position in a criticality register completes those arguments.
For each row, ask production to describe the consequence in a particular operating state. Would the failure stop the current campaign, damage material, or merely remove a convenience? Ask maintenance to show the available recovery arrangement. Where the answers disagree, retain both explanations as open items and name the person who will resolve them. A neat table should not hide disagreement that matters to the decision.
Use a separate economic note after the risk constraints are established. Record which loss estimate is measured and which is assumed, with a source for each. NIST’s report examines the data needed to estimate maintenance-related costs and losses; this matrix supplies no financial threshold. NIST maintenance economics report.
Finally, allow mixed strategies. A condition route, routine servicing and a proof test can serve different purposes on the same assembly. Write those purposes beside the tasks so future reviewers do not remove one merely because another carries a more advanced label. This worksheet is an initial decision aid, while formal RCM evaluation has its own criteria under SAE JA1011.
Review the plant plan with an owner and a dated record
Start with one bounded system and the people who operate, maintain and approve changes to it. Bring the equipment instructions, recent work orders, relevant condition records and applicable inspection requirements. For each selected failure mode, complete the worksheet below. It is a proposed working record, not a substitute for the site’s safety assessment or maintenance procedure.
| Worksheet field | Entry to copy into the review record |
|---|---|
| System and required function | Asset identifier, operating state and required performance |
| Failure mode and consequence | Specific mechanism and affected safety, environment, quality or service function |
| Current task | Actual trigger, interval, procedure and recorded findings |
| Evidence | Document or work-order references, dates, measurements and known gaps |
| Warning and response | Detection method, uncertainty, confirmation and time needed to act |
| Constraints | Applicable requirements, safe access, spares, competence and standby evidence |
| Proposed strategy and reason | Task or combination of tasks; assumptions supporting the choice |
| Owner | Person accountable for resolving gaps and obtaining the required approval |
| Review date | Calendar date for the next review and events that trigger an earlier review |
| Acceptance and follow-up | Evidence required before changing the task and after completing the work |
Agree what would make the proposed strategy fail its review. For a predictive trial, that could be inadequate warning for the planned response. For scheduled replacement, it could be a missing basis linking the task to the failure mechanism. For a corrective arrangement, it could be an unverified recovery assumption. These are proposed acceptance questions; the responsible plant team must decide the applicable criteria and what interim arrangements are necessary.
HSE’s maintenance, inspection and testing guidance calls for records, tracking, communication and management of overdue work. Assigning an owner and review date gives those activities a place in this record. An unresolved question should remain visible rather than turning into an implied approval when the meeting ends. HSE MIT overview.
Keep required work distinct from proposed optimization. HSE’s major-hazard guidance considers several maintenance regimes for plant with safety functions and also examines proof-check adequacy. Regulation therefore should not be reduced to an instruction to make every task exclusively preventive. Establish the actual obligation and obtain the required authorization before altering the arrangement. HSE maintenance procedures.
Choose a review date that the team can own, and specify earlier triggers such as a changed duty or a failed functional test. Attach the actual evidence to the record; maintenance troubleshooting with evidence offers a related way to organize that material. For help defining a bounded plant information use case, contact WizeeMind. Bring one system, its current tasks and the questions your team cannot yet answer.
Frequently asked questions
What is the difference between predictive, preventive and corrective maintenance?
Predictive maintenance uses condition evidence to anticipate a need for intervention. Scheduled preventive maintenance acts at a defined time or usage interval. Corrective maintenance restores a faulted function. A plant can combine these approaches on one asset because its failure modes differ. Record the trigger and purpose of each task so the categories remain useful when reviewing work history. DOE maintenance approaches.
Does every critical asset need predictive maintenance?
No. Criticality establishes the importance of managing failure consequences. Predictive maintenance also needs a detectable failure mechanism, sufficient warning to act and a justified monitoring process. Where effective maintenance cannot control an unacceptable consequence, engineering or protection changes need review. A high priority should lead to a more careful assessment of the options, rather than an automatic sensor purchase. NASA RCM principles.
When is preventive maintenance the better choice?
Scheduled preventive maintenance is a candidate when an applicable task and defensible interval control the relevant deterioration or preserve function. Use operating history, manufacturer instructions and applicable obligations to justify the task. A low replacement price alone does not establish that replacement prevents the failure. Include safe access and return-to-service checks in the proposed job. HSE maintenance guidance.
When is planned corrective maintenance acceptable?
It can be acceptable when the assessed consequences are tolerable, failure is evident, recovery is practical and the responsible owner accepts the arrangement. Check safety, quality, environmental and service effects before making that decision. Redundancy must work under the actual conditions in which it is needed. Ask for the test record and recovery plan before relying on a standby unit. NASA RCM decision outcomes.
Can predictive maintenance replace mandatory inspections?
Do not remove required inspections or tests merely because condition monitoring is available. Review the applicable jurisdiction, equipment rules and approved arrangements. Monitoring may complement the required work; a change needs a documented basis and the appropriate authorization. For Great Britain work equipment, HSE connects inspection needs and intervals to risk assessment, including relevant deterioration. HSE inspection guidance.