A cleaning process for machine-tool exteriors is usually carried by the two or three people who happen to be on shift, and it fails the moment one of them is absent. A written procedure does not fix that by being longer. It earns its place only if it still produces the same result on a shift that did not write it, when an operator who joined recently meets a case the procedure did not expect, and when a second machine model arrives on the floor.
Most shop-floor cleaning documents break for a structural reason rather than a careless one. They are organized around machine names, because that is how the maintenance plan is written, while the work itself is organized around surfaces. A guard panel, a nameplate, a cable entry and a chip conveyor behave the same way on an older manual lathe and on a five-axis machining center; the machines differ, the surfaces mostly do not. Write the process against the surfaces and the machine list becomes a one-page annex. Write it against the machines and every new model doubles the document until nobody follows any of it.
Define what the document has to survive
Before writing anything, decide what the document is being tested against. Three tests cover most of what a facility actually needs. First, a shift that did not write the document can run it without telephoning the author. Second, an operator who has been on the job a short time can find the decision for an unexpected case inside the document rather than in someone else's head. Third, a second machine model can be added by filling in one page, not by rewriting the procedure.
When a document fails these tests, the cause is almost always one of three things. The completion standard is an adjective, so two people legitimately disagree about whether the job is finished. The settings were copied from a brochure instead of measured on our own panels, so the first hard surface defeats the plan. Or the facts that belong to one machine were woven into the steps, so the document cannot be reused and quietly dies with the person who wrote it.
Repeatability is also not the same thing as frequency. A task done every Friday that produces a different result each Friday is not a controlled process; it is a recurring argument with a schedule attached. The point of documenting the work is to make the result independent of who does it and to make the interval something the records can justify.
Write the scope as exclusions
Inclusions grow; exclusions bound. A scope written only as a list of what to clean will absorb one more item every time someone asks a reasonable question. A scope written as exclusions stays one paragraph and keeps its authority.
For machine-tool exteriors the exclusions are usually the same set. The process does not relubricate the things it strips oil from: cleaning a guideway, a ballscrew or a rack removes the film that was protecting it, so returning oil the same day, per the machine's own maintenance schedule, is a task the card must name and hand to a named role. The process does not open electrical cabinets, clean the inside of coolant tanks or chip conveyors, touch a spindle taper or a precision table, or work on the coolant system itself. It does not remove rust, repair a lifting paint edge, or replace lockout and tagout and the site's permit rules. Surface cleaning is an external task; anything that requires a cover off belongs to another procedure.
Each machine annex should then record one honest figure: the cleanable area, which is the visible area minus the protected area. Condensation goes where nobody sprayed - under labels, beneath flanges, behind the machine, along the base - and masking takes time that belongs inside the job, not beside it. A card that counts spray minutes and hides protection minutes will make every method on the floor look unreliable, because the comparison it is feeding into was never fair.
Build the body from surface groups, not machine names
The reusable part of the document is a short set of cards, one per surface group. On machine tools those groups are recognizable from one end of the shop to the other: painted sheet-metal guards, panels and doors; bare machined surfaces, guideways and screws; electrical cabinet exteriors, operator pendants and ventilation openings; nameplates, decals and acrylic windows; hydraulic and pneumatic lines, hose carriers and cable chains; chip conveyors, coolant tank exteriors and the floor directly beneath them.
Every card carries the same fields, in the same order, so that anyone who has read one card can read all of them.
- What "clean" means in observable terms, written so that two people can check it and agree.
- What must be masked, covered, moved out of range, or left to another method entirely.
- The decision rule, written as a response to the state of the deposit rather than as a single path.
- The working distance, steam mode and pass speed that produced the result in our own trial, and who is allowed to change them.
- The drying step and the later check, including who performs them and when.
- The role that signs the line when the surface is handed back.
This structure survives a change of machine because the groups do not change; what changes is which items on a particular machine fall into each group, and that belongs in the annex. It also makes the equipment conversation sharper. When you compare machine tool steam cleaning options, you are not asking whether a machine is "suitable for machine tools" in the abstract; you are asking which of these surface groups a supplier has actually worked on, at what distance, with which attachment, and with what result on painted sheet metal, on bare metal, and near a cabinet vent.
Write decision rules, not adjectives
Delete the words "thoroughly", "carefully", "properly" and "as needed" from every card. Each of them hides a decision, and the decision will be made differently on every shift, which is exactly the problem the document exists to remove. Replace each one with an observation and a response.
The observations are quick and physical: light held at a low angle across the panel rather than overhead light; a white cloth that the whole group has agreed on, so a result means the same thing in different hands; a plastic scraper and the grit felt between the fingers. Those three checks put a surface into one of three states, and the card states the response to each. A fresh film moves with the cloth and offers no resistance, so it is a routine task on a normal interval. A polymerized film is tacky or dull with dark edges and moves less the harder you rub, so it needs heat and dwell, applied as several short passes - and the card should say in plain words that repeat passes are the expected method, not evidence that the operator is doing it wrong. A deposit carrying embedded grit is gritty and does not move under heat alone, so it is removed dry first and then treated as a film; heat on a hard crust only moves material downstream, where it cools, re-deposits and scratches whatever it lands on.
Then write the stop rule, because a procedure without one teaches people that every problem is theirs to solve alone. Lifting paint at an edge, a decal corner rising, moisture at a cable gland that was not there before, or oil returning within a day of cleaning: stop on that surface, write it in the exception log, and treat a fast return as a leak, a venting or a lubrication question rather than a cleaning failure. Where a card names the method for a state - cutting oil removal equipment, a scraper and cloth, a brush - it should name one of them, so the next shift does not improvise.
The settings line is where most documents go wrong, because it is tempting to copy it from a data sheet. On a machine that generates saturated steam, temperature and pressure are two readings of one condition, not two separate capabilities, and the value that reaches the panel is set by the hose, the distance and the pass speed. Ask a supplier whether published figures were measured at the generator or at the nozzle; the answer changes how a card is written. What belongs on the card is what our own trial produced: dry steam near electrical items and where the surface must be back in service quickly, wet steam where more heat transfer is worth the water it leaves and where drainage and drying are under control, plus the distance, the dwell and the number of passes, obtained by treating half a panel and leaving the other half as the comparison, changing one variable at a time.
One annex page per machine, filled before the first clean
The annex is the only place model-specific facts are allowed to live, and it is one page. It holds the machine identifier and model; a photograph or sketch with the surface groups marked on it; the distances measured from the position the machine will actually occupy, with each surface marked reachable, reachable with a platform, or out of reach; the power circuit and the name of the electrician who confirmed it; the water point and how far the operator walks to it; where the runoff goes and who owns oily water, which is a question for the site and the local drainage authority rather than for an equipment supplier; the sensitive items unique to that machine; the relubrication points and the role that performs them; the fixed reference points for photographs; and the measured time for each stage, including protection, drying and the later checks.
Reach is a physical limit and it should be measured, not assumed. A fixed machine stays at one point and the hose is the working radius, so the annex records the distance from that point to the farthest surface on the machine, not to the nearest wall. It also compares the machine's footprint against the door widths, aisles and turning space on the route it will travel: the published overall size of 1,200 x 800 x 1,200 mm including handles is a number to hold against a doorway before the delivery is booked, not after. If a surface is out of reach, the honest options are moving the position, reducing the scope, or changing the method - and that choice belongs in writing before the job is scheduled, not in the middle of it.
The blank annex is useful in a second way. It is the form for a machine that has not arrived yet, and therefore the list of questions to answer before the layout is frozen: what the machine will need at its position, what it will make oily, what must not get wet, and what the crew will have to walk to. It is also the sheet a supplier can be sent with our records attached, which turns a vague inquiry into a small number of specific ones.
Verification is part of the procedure, including the checks that report late
A completion standard is only useful if two people can reach the same answer without discussing it. Write it as observations: no visible film on the front panel under low-angle light; no oil transfer on the agreed cloth at the door handle and along the latch line; no standing water on any ledge, in any seam, or under a fastener head; the floor beneath the machine dry while the machine is running. Photograph the same two or three reference points before and after, from the same angles, so the record does not depend on anyone's memory of how it looked.
Some findings arrive after the crew has gone, and a procedure that ignores them produces a pass that is false by the next morning. Flash rust on bare metal shows up hours later. Water hiding in a gap announces itself as a streak overnight. A film that returns within days says the source is still active and the cleaning was only a reset. Put those checks in the document as their own lines, with a named role and a time, so that a late finding is a scheduled observation rather than a complaint.
Records follow the same structure as the cards: one line per surface group per machine, with the state found, the rule applied, the settings used and the outcome. "Machine 3 clean" cannot be compared, closed or audited, and it cannot tell you an interval. Intervals come out of the records. The time a film takes to return is the interval; the calendar is only how you book it. Comparing two shifts' records on the same machine is the only evidence that the document works, and it is also the cheapest quality system a facility can run.
The exception log runs the document
A process document that never changes is not well controlled; it is abandoned. The exception log is where operators write what the card did not cover, in one line: the surface, what happened, what they did instead. If writing it takes longer than doing the work, nobody will write it, so keep it on the same sheet as the record and keep it to three columns.
Read the log rather than reacting to it. A finding that appears once is a note. The same finding twice is a missing decision rule, and the fix belongs in the document, not in a conversation with the person on shift. The same finding three times is a scope problem or a source problem, and the honest answer may be that the surface should not be in this procedure at all.
Name an owner for the document and one place where the current version lives - printed at the machine, or in the folder that travels to it. No email attachments, no second copy on a shared drive, no file named with the word final. The version standing at the machine is the version. Then name the triggers that force a revision: a new model or a machine that has been moved; a change of coolant, lubricant or additive package, which changes what the deposit is made of; a repeated finding; an operator's suggestion; an audit or customer-visit comment; a change in how the site handles oily water. Announce a revision in two minutes at the start of the next shift, not by re-issuing a manual. One thing is not changed on the fly: the completion standard. If that drifts, the records stop being comparable and the interval loses its basis.
Publish a first version and let someone else run it
Write the first version for one surface group on one machine, and hand it to a colleague who did not write it and has not been briefed. Watch from a distance and note only two things: where they stop to ask, and where they do something the card did not say. Those places are the missing decision rules. They are not evidence that the person needs training; they are the parts of the process that were still living in the author's head. Fold them in, run one timed job with the annex filled in - including protection, drying and the later checks, because those are the stages an estimate always forgets - and issue the version that the timed run supports.
Then compare two shifts before extending to the next machine. If the results differ, look first at the three causes that produce almost all of these differences: a completion standard written as an adjective, settings copied instead of measured, and machine-specific facts buried inside the steps. Fix the document before concluding anything about the method or the crew.
Extend one surface group at a time, filling in that machine's annex as you go. A set of cards that covers a whole shop because it was written against surfaces ends up smaller than a set that covers two machines because it was written against machine names, and it is the smaller one that survives a change of staff.
What a facility should send a supplier at the end of this is the record, not a summary: the surface groups, the states found on each, the settings that worked, the times each stage took, and the photographs from the fixed points. That single attachment turns an open-ended question into a short list of answerable ones - what the steam temperature is at the nozzle end of the hose being supplied, the hose and wand lengths, the warm-up time and the runtime per tank fill, and a written statement of which surface groups the machine is intended for. If you want that conversation started with our engineering team, use the contact page and send the record with the question.
