TPM: Total Productive Maintenance in Regulated Manufacturing

TPM (total productive maintenance) is a cross-functional maintenance practice that involves operators, maintenance specialists and supporting functions in preventing equipment-related losses throughout the equipment lifecycle. Its scope extends beyond fixing breakdowns: it connects everyday equipment care with decisions about equipment selection, operation, maintainability and improvement.

As industry practice, TPM organizes shared equipment care and improvement. It is not itself a legal requirement, a maintenance schedule or a product-release decision. Product-specific regulatory duties belong in the regulatory mapping below; TPM terminology is not a compliance test. The distinction matters because an organizational framework cannot replace task instructions or site authorization decisions.

The organizational difference is shared attention, not interchangeable expertise. Operators commonly notice changes during use; maintenance specialists investigate technical causes; engineering, production and quality functions contribute within their responsibilities. Direct observation through a gemba walk can help these groups understand the same equipment condition rather than discuss separate versions of the problem.

A useful test of the approach is whether an observed problem becomes a shared improvement question: why does this condition recur, and what would prevent it? Completing scheduled service alone does not demonstrate that broader way of working.

The TPM pillar framework is industry practice, not legislation or a universally fixed standard. Common presentations include the pillars below, but terminology and grouping vary between organizations. Treat each pillar as a way to organize work with a recognizable output, rather than a campaign slogan.

For practical use, assign an owner to each output and identify who will use it. A pillar becomes useful when its work changes an equipment-care decision, not merely when a team adopts its name.

Conventional TPM industry practice groups production losses into the families below. Common overall equipment effectiveness (OEE) treatments associate them with availability, performance and quality; these are diagnostic conventions, not legal categories.

Use equipment-specific observations to separate symptoms. A stopped conveyor calls for checking the drive, interlocks and upstream supply; repeated short interruptions suggest examining sensor events or feed consistency. Operation below the reference rate calls for checking rate assumptions and process constraints. Rejected output calls for investigating defect types and operating conditions. Direct shop-floor observation helps distinguish these cases; a single aggregate score can conceal them.

Keep measurement boundaries explicit: define the production window, stop-classification rules, reference rate and treatment of startup output. These choices affect where losses appear. Classifying cleaning or changeover time as a loss does not establish that the activity is unnecessary. An improved OEE result does not, by itself, demonstrate cleaning effectiveness or product acceptability; assess those separately through contamination-control evidence and product assessment, including cross-contamination controls.

The product-specific provisions below are binding US regulation. TPM pillars, task-design choices and metrics are industry practice, not regulatory requirements or official guidance. The practical boundary is between choosing how to organize maintenance and meeting the applicable equipment obligations.

US finished drug products: procedures, condition and records

• For US finished drug products subject to 21 CFR 211, 211.67(a) requires cleaning and maintenance of equipment and utensils at appropriate intervals, with sanitizing and/or sterilizing as appropriate for the drug’s nature.
• For US finished drug products under 21 CFR 211, 211.67(b) requires written equipment-cleaning and maintenance procedures to be established and followed.
• For US finished drug products under 21 CFR 211, procedure content includes assigned cleaning and maintenance responsibility under 211.67(b)(1).
• For US finished drug products under 21 CFR 211, procedure content includes maintenance and cleaning schedules, including sanitizing schedules where appropriate, under 211.67(b)(2).
• For US finished drug products under 21 CFR 211, 211.67(b)(3) specifies sufficiently detailed descriptions of cleaning and maintenance methods, equipment and materials.
• For US finished drug products under 21 CFR 211, procedure content includes removal or obliteration of previous batch identification under 211.67(b)(4).
• For US finished drug products under 21 CFR 211, procedure content includes protection of clean equipment from contamination before use under 211.67(b)(5).
• For US finished drug products under 21 CFR 211, procedure content includes inspection for cleanliness immediately before use under 211.67(b)(6).
• For US finished drug products under 21 CFR 211, 211.67(c) requires records of maintenance, cleaning, sanitizing and inspection.

US human food: cleanability and contamination prevention

• For US human-food operations subject to 21 CFR 117, 117.40(a)(1) requires adequately cleanable equipment and utensils, adequately maintained to protect against allergen cross-contact and contamination.
• For US human-food operations under 21 CFR 117, 117.40(a)(2) requires appropriate equipment design, construction and use to avoid food adulteration with lubricants, fuel, metal fragments, contaminated water or other contaminants.
• For US human-food operations under 21 CFR 117, 117.40(a)(3) requires installation that facilitates cleaning and maintenance of equipment and adjacent spaces.

US dietary supplements: intended use and sanitation conditions

• For US dietary supplement operations subject to 21 CFR 111, 111.27(a) requires equipment and utensils of appropriate design, construction and workmanship for their intended use.
• For US dietary supplement operations under 21 CFR 111, 111.27(d) requires maintaining, cleaning and sanitizing, as necessary, equipment, utensils and contact surfaces used to manufacture, package, label or hold components or supplements.
• For US dietary supplement operations under 21 CFR 111, 111.27(d)(1) requires taking equipment and utensils apart as necessary for thorough maintenance, cleaning and sanitizing.
• For US dietary supplement wet-processing operations under 21 CFR 111, 111.27(d)(3) requires necessary contact-surface cleaning and sanitizing before use and after interruptions during which contamination may have occurred.
• For US dietary supplement operations under 21 CFR 111, 111.27(d)(7) requires storing cleaned and sanitized portable contact equipment and utensils in a location and manner that protects them from contamination.

As industry practice, translate these boundaries into task instructions: identify who performs the work, what condition is checked and what evidence supports reuse. A TPM board or software task list supports coordination; it does not replace human review and approval.

For US human-food sanitation, check the current text of 21 CFR 117.35 for scope. Review device, chemical and non-US applications separately rather than extending this mapping. Related product frameworks: finished-drug cGMP, human-food cGMP and dietary supplement cGMP.

As industry practice, TPM can organize maintenance around the following risk-based workflow. Start by defining the asset boundary, including shared services and product-contact interfaces, and documenting its baseline condition. The equipment owner and maintenance lead select tasks against credible failure modes, then assign competent people and escalation responsibilities.

Asset boundary and baseline → task selection and authorization → observations → abnormality record → coordinated intervention → readiness review → controlled reuse. Recurrence evidence loops back to task selection; an unresolved finding returns to assessment rather than advancing to reuse.

This is a decision map, not an identical sequence for every event. A normal observation may simply close the check; a hazardous condition may call for immediate isolation under site procedures before further diagnosis.

Define the operator–maintenance boundary by task, not job title alone. Observation, external cleaning, adjustment, lubrication, disassembly and troubleshooting can involve different hazards and authorizations. Specify permitted actions, competence, isolation conditions and stopping points for each task. Autonomous maintenance does not mean unrestricted intervention.

During a routine check or gemba observation, an operator might notice abnormal sound, leakage or unexpected wear. Preserve the observation without prematurely naming a cause: record the location, operating state, product or batch context and when the condition appeared. Route it through the site's andon escalation or other agreed channel to an accountable supervisor or maintenance role. That person coordinates access and identifies the evidence needed before inspection, intervention or continued operation.

Keep human decisions distinct. Maintenance personnel document work completion; the site's authorized people assess equipment readiness and, separately, affected-product disposition. A completed repair settles neither decision by itself. Relevant evidence may overlap, but permission to use equipment is not a decision about product acceptability.

After reuse, the equipment owner and maintenance lead review recurrence evidence and revise task content, frequency or authorization where the routine failed to detect or prevent the condition.

An illustrative industry-practice design links each asset to its tasks and observations, each intervention to the task or observation that initiated it, and each equipment-readiness review to the intervention assessed. These relationships can structure a CMMS record without prescribing a particular database.

• Observed condition: what someone directly saw or measured, with units where applicable.
• Suspected cause: an interpretation awaiting confirmation, not a finding.
• Action performed: the work actually completed, rather than merely planned.
• Review disposition: the decision about equipment readiness, its basis and any unresolved conditions.

Keeping these separate helps prevent an unconfirmed diagnosis from becoming an apparent fact in later recurrence analysis.

Stable asset identifiers preserve equipment identity; procedure revisions identify the instructions used; event times establish sequence; responsible roles locate accountability. Potentially affected lot references connect the event to production through production context tagging. Treat a missing relationship as an unresolved information issue for review—not as evidence that no lot was affected.

As covered above, for US drug products under 21 CFR 211, 211.67(c) requires records of maintenance, cleaning, sanitizing and inspection as specified in §§ 211.180 and 211.182. This model can help organize those records; it is not a universal legal schema, and its proposed fields are not individually mandated by that provision.

The following are suggested industry-practice indicators, not regulatory targets. Define event-counting rules, asset scope and observation windows before comparing results.

Keep active repair duration separate from total out-of-service duration. For example, sum hands-on intervals from repair activity starting to pausing or finishing; measure out-of-service time from equipment withdrawal to authorized return to use. Use hours consistently. Record waiting, repair and readiness-review intervals separately when the distinction changes the next action; identify overlapping intervals rather than adding them twice.

Pair task completion with recurrence or defect outcomes. High completion can coexist with repeated failures when tasks miss the failure mechanism. Compare similar duties, products, asset conditions and measurement windows—not unlike assets merely because they share a dashboard.

Investigate direction, not just target attainment: did the numerator improve, did operating exposure change, or was loss reclassified? Use direct observation at the equipment to test explanations before changing the maintenance plan.

These hypothetical industry-practice cases illustrate decisions, not acceptance criteria or production outcomes.

Drug blender: seal repair with an uncertain impact interval

Assume replacing a blender seal opens the product-contact zone and the affected production interval is initially uncertain. Record downtime in minutes and potentially affected lots as counts. Organize condition observations into seal integrity, residue, leakage and foreign-material evidence.

Start with the last documented satisfactory observation; trace subsequent operations through shutdown, opening, repair and cleaning. That observation anchors the review but does not prove when deterioration began. Progressive wear could implicate earlier operations; damage attributable to disassembly could narrow the concern to the intervention. Compare inspection records, removed-part condition and batch-equipment history before narrowing the potentially affected interval.

Apply the lifecycle distinction above: repair completion supports a mechanical decision, not equipment reuse or lot acceptance. Route unresolved cross-contamination concerns and identified lots for quality assessment; do not infer disposition from a successful test run.

Food conveyor: completed tasks, recurring stops

Assume scheduled tasks are complete, but an inaccessible roller repeatedly binds. Record stop counts alongside operating hours for comparable operating conditions. Repeated adjustment treats the symptom; it does not distinguish misalignment from wear, washdown exposure or incomplete task coverage.

Alignment measurements that drift before stops would favor an adjustment-related explanation. Rough rotation or visible damage would favor wear. Recurrence after washdown, accompanied by moisture ingress or corrosion, would favor exposure. A task that omits the roller—or cannot physically reach it—would favor a coverage gap despite recorded completion.

Use a gemba observation to compare actual access and execution with the task instructions. Choose targeted inspection, repair or task redesign according to the findings, then monitor recurrence against operating hours rather than counting completed work orders.

Dietary supplement wet process: exposed contact surface

Assume an interruption exposes a contact surface and initial observation cannot exclude contamination. Record interruption duration in minutes and potentially affected batches as counts. Apply the interruption boundary covered above rather than treating elapsed time alone as evidence of cleanliness.

Intact protection and documented handling could support limited exposure; splash, condensate or uncontrolled contact would support a broader concern. Reconstruct exposure conditions and batch sequence, then document the cleaning and sanitizing response and subsequent surface condition. An interruption log establishes that an event occurred. It does not resolve the cleaning questions, equipment-use decision or assessment of potentially affected batches. Keep unresolved questions with designated reviewers before resumption.

Software-support design practice: define record ownership before connecting systems. One workable allocation gives the asset master ownership of equipment identity, the manufacturing batch record ownership of production context, the maintenance work order ownership of execution details, and the quality-event record ownership of impact assessment. These are implementation choices, not additional regulatory requirements.

For maintenance–manufacturing exchanges, use stable event identifiers, matching asset references and preserved event times. Define how duplicates are recognized without repeating an action, and how out-of-order messages are reviewed rather than allowed to replace newer states. Distinguish message receipt from acceptance of the business action: a delivered completion message can still fail validation.

Make recovery visible, assign an accountable reviewer and reconcile against the originating record:

• Offline entry: flag pending synchronization; the maintenance supervisor checks original event times and source entries before accepting the uploaded work history.
• Failed transfer: retain a failed-delivery exception; the interface owner reconciles the source payload and target receipt before retrying with the same event identifier.
• Conflicting equipment states: expose both states; the designated equipment-status reviewer compares their originating records and records the resolution rather than silently overwriting either.
• Unmatched lot reference: hold the association as unresolved; the production-record owner checks the originating batch record and documents any corrected link.

Apply the same ownership discipline to manufacturing–quality exchanges. Reconciliation repairs record relationships; equipment-use and product-release decisions remain at the human decision boundary described in the lifecycle above.

These are maintenance-practice controls, not additional regulatory requirements. Use them to correct a specific execution weakness rather than adding the same checks to every task.

An unloaded run is not universally sufficient evidence that an intervention worked. For example, if leakage appeared only after a clean-in-place cycle, a cold idle check does not address that reported condition. Plan a controlled check around the relevant thermal or cleaning exposure, with an observable result that can support or challenge the repair diagnosis.

Keep remedies proportionate to the mechanism. More checklist items, data fields or frequent work can increase burden without resolving unclear task purpose or poor diagnostic information. Before expanding a task, identify what decision the added observation would change; if that is unclear, improve the task definition first.

Use these industry-practice questions to test readiness, not to determine legal compliance. Record unresolved answers as actions with an owner and a review point.

• Is the equipment boundary clear? Identify included components, utilities and interfaces with adjacent equipment.
• Is the baseline condition understood? Distinguish known defects from acceptable operating condition before judging improvement.
• Does each task address an identifiable condition or failure mechanism? Explain what finding would prompt action.
• Are responsibilities workable across shifts? Check access, competence, handover and escalation coverage with the people doing the work.
• Can an observation be traced through intervention and review? Follow a real example, including unresolved findings.
• Can the chosen metric change a maintenance decision? Identify which result would justify revising a task.

Start with a bounded equipment application. Evaluate whether routines produce useful observations and resolvable work. Expand based on demonstrated task usefulness, record quality and manageable exception handling, rather than the number of assets enrolled.

Review the program after meaningful changes in equipment, materials, operating conditions or recurring failure patterns, alongside the site's chosen review rhythm. Connect relevant changes to change control. Use the earlier regulatory mapping and failure-control table as references when deciding which routines to retain, revise or retire.