Autonomous Maintenance: The TPM Pillar in Practice

Autonomous maintenance is the pillar of Total Productive Maintenance (TPM) in which machine operators take ownership of basic equipment care: cleaning, inspecting, lubricating and tightening, so abnormalities are caught before they become breakdowns. It works because the operator is at the machine all shift and notices the new rattle or fresh oil spot long before anyone else. This guide covers the classic seven steps in practical terms, why cleaning is really inspection, how the handover with the maintenance team should work, and the failure modes that sink most rollouts.

What autonomous maintenance actually is (and is not)

Autonomous maintenance, known in the original TPM literature as jishu hozen, hands routine equipment care to the people who run the machines. In practice that means operators clean, inspect, lubricate and tighten (often abbreviated CILT) on a defined schedule, and raise abnormalities the moment they find them. The logic is simple: the operator stands at the machine all shift and hears every new noise and sees every fresh leak, while a maintenance technician might only see that asset for a few minutes on a planned round.

It is worth being equally clear about what autonomous maintenance is not. It is not a way to shrink the maintenance department, and it is not operators doing repairs. Operators take on the frequent, simple, low-risk tasks; skilled trades keep the diagnostics, overhauls and anything needing isolation, permits or specialist tools. Plants that treat it as free labour rather than a capability-building programme reliably get resentment and pencil-whipped checklists instead of healthier machines.

The seven steps, condensed for a working plant

The classic rollout has seven steps. You do not need to recite them in an audit, but the sequence matters, because each step builds the capability the next one depends on. Condensed for the real world, they look like this:

Most plants live in steps one to four for a long time, and that is fine. A line that has genuinely completed the initial clean and killed its contamination sources is already far more reliable than one that skipped straight to laminated checklists.

• 1. Initial clean: take the machine back to base condition. Strip off the grime, find every leak, loose fastener and worn part, and tag each abnormality for action.
• 2. Eliminate contamination sources: fix the root causes of the mess you just cleaned. Stop the leaks, guard the swarf, redirect the coolant spray, and open up hard-to-reach points so the next clean takes minutes, not hours.
• 3. Provisional standards: write the first cleaning, lubrication and tightening standards: what gets done, by whom, how often, with photos of what good looks like.
• 4. General inspection skills: train operators in how the equipment actually works (drives, pneumatics, hydraulics, fasteners) so their inspections catch real defects, not just dirt.
• 5. Autonomous inspection: merge operator checks with the maintenance department's PM routes so nothing is duplicated and nothing falls into the gap between the two.
• 6. Organisation and standardisation: extend the discipline beyond the machine to the whole workplace: tools, gauges, materials, data recording and visual controls.
• 7. Self-management: teams maintain and improve their own standards, track their own losses and run their own small improvements without being pushed.

Cleaning is inspection

The founding insight of autonomous maintenance is that cleaning is inspection. You cannot wipe down a gearbox without your hand passing over the housing, and a hot housing, a weeping seal or a loose mounting bolt is information. A machine kept at base condition makes new problems visible immediately: fresh oil under a clean machine is a signal, while fresh oil under a filthy one is invisible.

Make each finding count. Every abnormality found during cleaning gets a tag (the classic system uses one colour for items operators can fix and another for items maintenance must take), the tag gets logged, and someone visibly closes it out. A wall of faded tags nobody has actioned kills the programme faster than never starting, because it teaches operators that looking closely is pointless.

The handshake with the maintenance team

Autonomous maintenance only works as a handshake between production and maintenance, with an explicit line drawn through the task list. A useful rule of thumb: a task transfers to operators when it is frequent, quick, low-risk, needs no isolation beyond normal operating procedure, and relies on senses or simple tools rather than diagnostics.

The handshake runs both ways. Maintenance trains operators, writes the standards with them and responds quickly to tags; operators absorb the routine care that used to interrupt technicians, freeing them for planned maintenance and root cause work. If either side treats the other as a threat (operators fearing dumped work, technicians fearing lost jobs), renegotiate the split in the open rather than imposing it.

• Transfers to operators: routine cleaning, visual and sensory inspection, topping up lubricant at marked external points, tightening accessible fasteners, simple condition checks (gauges in the green zone, filters, belt condition) and first-line response such as clearing jams to a standard.
• Stays with maintenance: fault diagnosis, repairs and part replacement, precision alignment and calibration, internal lubrication, anything under a permit or lock-out requirement, and anything that requires opening the machine up.

Why rollouts fail, and how to keep score

The most common failure mode is dumping. Management announces that operators now own their machines, hands over checklists, and provides no time in the shift, no training and no response when tags are raised. The checks quickly become tick-box theatre. The fix is structural: cleaning and inspection time is planned into the schedule like a changeover, inspection duties only expand after the step-four training has happened, and every raised abnormality gets a visible response.

Other predictable killers: launching on too many machines at once instead of one model line, standards written at a desk rather than with the operators who will use them, supervisors never auditing the routines (so they quietly decay), and measuring activity (checklists completed) instead of results (unplanned stops, minor stoppages, defects).

That last point deserves emphasis: judge autonomous maintenance by whether machine behaviour changes. Unplanned stops, micro-stops and early failures on covered equipment should trend down, and MTBF should trend up, which requires honest stop data rather than memory and clipboards. The partner we recommend, Fabrico, reads stops directly from the machines and shows the true cause on video, so you can see whether operator-owned care is actually cutting unplanned downtime. Fabrico is a partner we recommend; the tools here are free regardless.

FAQ

Related: Total Productive Maintenance · 5S · preventive vs predictive maintenance · the six big losses