Mechanical Maintenance Methodology: turning ‘fixing the machine’ into ‘managing the system’

1. How a paper machine loses money through downtime

The paper machine is one of the most ‘fragile’ asset forms in process industry: single-line continuous, fully interlocked — if any one section from the headbox to the reel trips, the whole line stops; there is no such thing as ‘half a paper machine keeps producing’. Therefore the economics of maintenance must start with downtime loss. Take a typical graphic-paper machine as an example: design capacity 1,159 t/d, reel design speed 1,200 m/min, design basis weight 90 g/m². Converting to an hourly production figure:

Fifty tonnes of paper per hour means a mid-size failure lasting 8 hours loses not ‘a few hours of electricity’ but the marginal contribution of four hundred tonnes of paper, plus start-up rejects and re-ramping energy. So the mature-mill view is: each percentage-point gain in availability (time-based) is usually worth more than any single process optimization. Top-tier graphic/packaging machines run at 92–96% availability; tissue machines differ slightly because of their blade and coating-management cadence.

The primary goal of maintenance management is therefore clear: minimize the sum of ‘maintenance cost + failure loss + spare-parts capital tied up’, not minimize maintenance cost itself. The money saved by deferring one felt change may fall far short of compensating one accidental web break — which is why methodology must precede craft.

2. Three maintenance strategies: BM / PM / CBM, determined by failure mode

All maintenance actions reduce to three ‘parent strategies’: BM (Breakdown/Corrective — fix it after it fails), PM (Preventive — periodic prevention), and CBM (Condition-Based — triggered by condition). Which to choose is not decided by how expensive the equipment is, but by the failure mode and consequence: is there a pattern to the failure? how severe is the consequence when it fails? how costly is monitoring it? RCM (Reliability-Centered Maintenance) breaks this into a chain of standard questions — what is the function, how does it fail, what is the consequence, which strategy yields the lowest total cost. On the paper-machine floor, the allocation looks like this:

• BM applies to: cheap parts with no warning before failure and no secondary damage — light bulbs, small limit switches, some pipe clamps. Spending condition-monitoring money on these is waste.
• PM applies to: consumables and wear parts with well-defined wear patterns — forming fabric, press felt, doctor blades, refiner plates, screen drums, seal strips. The interval is revised iteratively from life statistics (Kaplan–Meier or simple Weibull estimation).
• CBM applies to: expensive critical parts showing gradual degradation where failure means a major overhaul — main drive gearbox, suction-roll bearings, dryer cylinder journals, hydraulic-unit main pump, fan rotor. The means are the vibration / infrared / oil-analysis trio in the next section.

The ratio among the three strategies is a direct indicator of maintenance-system maturity: junior teams have too high a BM share (fire brigade), mature teams push BM down to small parts whose failure consequence is negligible, with PM/CBM covering most planned-downtime work hours.

3. Inspection: five senses + three instruments; trend matters more than readings

Inspection is the foundation of CBM. Its organizational form is the three standards of ‘inspection route + inspection baseline + inspection record’: divide the machine into several 30–60 minute patrol routes, each with a baseline table (temperature upper limit, pressure range, clearance values, oil-level line), and once recorded into the system what matters is the trend, not single-point readings — a bearing temperature creeping slowly from 58 ℃ to 68 ℃ deserves more alarm than one steady at 72 ℃. The human five senses remain useful but must be paired with instruments:

• Vibration meter: measures the RMS velocity at the bearing housing (mm/s RMS). Reference criteria: for most low/medium-speed paper-machine units, ≤2.3 mm/s is good, 2.3–4.5 mm/s calls for attention and denser monitoring, >4.5 mm/s warrants action (different machine classes use the corresponding ISO 10816 severity band; avoid misapplying across classes). When the vibration trend jumps a step suddenly, first check whether speed and load changed in step, before blaming the bearing.
• Infrared thermometry: bearing housings, dryer cylinder journal covers, hydraulic tank, motor casing, electrical terminal strips. The value of IR is in ‘horizontal comparison’ — among identical equipment under the same duty, the one running 8–10 ℃ hotter is first on the repair queue.
• Oil analysis: sample and send hydraulic and lubricating oil for testing. Cleanliness per the ISO 4406 code (hydraulic servo systems commonly target 16/14/11 to 18/16/13 depending on valve sensitivity); ferrography shows wear-particle morphology, and excessive moisture first points to cooler leakage. Oil is the ‘blood test report’ of rotating equipment, often alerting months ahead of teardown.

Typical on-site setting of inspection parameters (example, for building the baseline table; each mill re-baselines by machine type):

4. Failure analysis: RCFA, 5Why, FMEA and the three-column fault table

80% of paper-machine failures are repeat failures — meaning one analysis benefits a whole chain. A mature machine manual always has a chapter ‘Fault finding (Troubleshooting)’ whose format is a three-column table: Problem → Possible Cause → Solution. Taking a pulper as an example, the table looks like this:

This table is the fast-lookup layer of ‘symptom → cause → action’, but it does not answer ‘why the bearing wore out’. To go one level up, you need RCFA (Root Cause Failure Analysis):

• Preserve the scene: keep the failed part; don’t rush it into the scrap bin — fracture surface, wear debris, and sludge are all evidence.
• Collect data: DCS history trend (24 h before failure), vibration and temperature trends, oil report, recent maintenance records, comparison with identical equipment.
• Build a fault tree: decompose from ‘frequent web breaks’ or ‘bearing overheating’ layer by layer, drilling vertically with 5Why, each layer supported by evidence rather than assumption.
• Confirm root cause and countermeasures: countermeasures split into corrective (replace part) and preventive (change design / procedure / interval); lacking the latter guarantees recurrence.
• Standardize: write the conclusions into the inspection baseline, maintenance procedure, and spare-parts technical agreement — this is the step that distinguishes RCFA from ‘fix and forget’.

Two quantitative indicators close this chapter: MTBF (Mean Time Between Failures) and MTTR (Mean Time To Repair), which together determine inherent availability.

5. Maintenance windows & safety: turning emergency repair into planned work

The same job done in a monthly planned shutdown versus an ad-hoc emergency repair differs in cost by 3–5 times: emergency repair requires pulling people for overtime, sourcing spares on the fly, and working on a site still carrying stock and steam, amplifying both safety and quality risk. So the final link of maintenance management is to plan everything that can be planned:

• Downtime-window tiers: daily stop (1–2 h, handle queued items found by inspection), weekly stop (4–8 h, felt change / blade change / periodic servicing), annual stop (5–10 days, overhaul and retrofit). Work hours within a window are scheduled to the minute — every extra hour of machine downtime is tens of tonnes of paper.
• Job package: work order + drawings + spare-parts list + special tools + safety measures + acceptance criteria, prefabricated as a bundle. The installation binder in the equipment delivery package (section drawings, parts list, installation notes) is the raw material for the job package — the ‘shutdown maintenance’ lecture in Part D will specifically cover how to turn it into your own shutdown-management system.
• LOTO five steps: notify → stop → energy isolation (electric / steam / water / air / stock) → lockout/tagout → verify zero energy. A paper machine’s ‘energy’ is far more than electricity: steam condensate, residual pressure in hydraulic accumulators, trapped stock pressure, and dryer-cylinder condensate flash — each can kill. Putting ‘work safety’ in Chapter 1 of the manual is not a ritual; incidents of stock pumps starting under pressure and dryer manholes opened before depressurization repeat in the industry every year.
• Start-up verification: finishing maintenance is not ‘just get it turning’ — clearance recheck, interlock test, creep-speed trial, gradual acceleration to process speed, each step with an acceptance record. The majority of maintenance-quality problems surface within 24 hours of start-up; this is the statistical window for the first-pass-acceptance rate.

6. Maintenance organization & performance: turning methodology into the crew’s daily actions

Writing methodology on paper is easy; landing it in the crew is hard. The difficulty is in two places: organization and metrics. Organizationally, paper-mill maintenance comes in three common forms: centralized (one plant-wide maintenance team — deep skills but one layer of response lag), decentralized (maintenance workers assigned to machines — fast response but narrow skill range and scattered spares), and hybrid (a machine-resident small team handles daily work and inspection, while specialist teams handle overhauls and special projects). For most mid-size mills the optimum is hybrid — the resident team guarantees response, the specialist team guarantees depth, and the annual shutdown merges the two teams. Whatever the form, one iron rule holds: operators do not repair equipment, and maintenance workers do not do inspection on the operators’ behalf — inspection is part of the operator’s responsibility (whoever starts the machine does the inspection), while the maintenance team handles professional upkeep and repair; blur the boundary and inspection becomes nominal.

The work order is the carrier of methodology. A work order’s life cycle: discovery (inspection / alarm / shift handover) → creation (symptom + location + severity) → scheduling (into which downtime window) → spare issue → execution (job package) → acceptance → close and archive. Archiving is not a formality — the work-order history is the machine’s medical record, and the ‘re-baseline the interval with this plant’s historical data’ from Section 2 of M01 draws its raw material from work-order statistics. The most common point where work-order flow jams is ‘creation’: inspection finds a problem but nobody bothers to create an order, handing it over verbally and forgetting it three days later — so order creation must be ‘easier than speaking’, which is why mobile scan-to-create was born.

At the metrics level, four numbers are enough for the maintenance team to watch:

• Planned-work ratio: planned hours ÷ total maintenance hours. Healthy value above 80% — below this you are still firefighting, and every improvement starts by raising the planned ratio.
• MTBF / MTTR: track by equipment category, watch the trend not the absolute value; split MTTR into ‘wait time / disassembly-assembly time / commissioning time’ segments — a high wait-time share means fix spares and process first, don’t rush to polish craft.
• First-pass acceptance rate: the share not reworked within 24 hours after maintenance. It measures maintenance quality and the rigor of start-up verification, and is also a mirror of how good the job packages are.
• Spare-parts capital turnover: spare issue value ÷ inventory capital. It ties maintenance and the warehouse to the same boat, preventing the inertia of ‘stock more for fear of shortage’ — hoarding is a black hole of inventory capital, while shortage is a downtime risk; both ends need someone accountable on the books.

Finally, back to people. Every rule of maintenance methodology — strategy, inspection, RCFA, KPI — is ultimately executed by specific people, so the skill matrix is where the system closes: list skills like ‘inspection, lubrication, disassembly-assembly, alignment, hydraulics, vacuum system, automation’ as columns and people as rows, rating each person per skill in four levels (can do / independent / can teach / can revise standards). The blank cells in the matrix are next year’s training plan — send people out for training, pair masters with apprentices internally, or self-study with books; invest in the cells, not in people already maxed out.

7. Ten pitfalls of the maintenance system (a condensed version of on-site hard lessons)

Having covered the positive side of the methodology, the negative side saves more tuition. The following ten are a list of pitfalls paid for repeatedly on the paper-machine floor — check yourself against how many you have:

1. Report MTBF as a score — MTBF diluted by ‘unrecorded minor repairs’ loses meaning; the definition must precede the number — what counts as a ‘failure’ needs one plant-wide definition table.
2. Inspection becomes checkbox-ticking — routes unchanged for ten days, records nobody reviews, trends nobody re-baselines; inspection lives in the closed loop, not in the form.
3. Replace critical parts by the calendar — replacing monitorable parts by time discards remaining life and introduces ‘early failure after replacement’ (re-shuffled installation-quality risk).
4. Run consumables to failure — dragging felt and fabric to the ‘performance cliff’ before replacement sacrifices a week of output and quality; a consumable’s value is in its performance phase, not its residual-value phase.
5. Emergency-repair heroism — fast emergency repair gets rewarded while planned maintenance goes unnoticed, so nobody wants to plan; the reward structure defines the maintenance culture.
6. ‘More spares is always better’ — a black hole of inventory capital plus scrap of dead stock; a spare-parts store without a reorder point and consumption statistics is just a warehouse, not a system.
7. One drum of lube for everything — mixing viscosity/additive systems across positions, no oil-grade chart on the wall, no dedicated filling cans; cross-contamination is gradual and its payback is delayed.
8. Alignment by feel — coupling alignment without a gauge, belt tension by thumb; the vibration bill arrives punctually at month end; alignment is the foundation of every rotating-equipment KPI.
9. LOTO as a formality — lock hung but zero energy not verified, one lock for multiple workers; lock count and energy list must match the job — this is the bottom line, with zero flexibility.
10. Failure analysis stops at part replacement — RCFA closes when it reaches ‘the bearing failed’; the three-column table treats symptoms, RCFA treats root cause, FMEA treats what has not yet happened — miss any layer and the failure will surely return with the next shift.

Print these ten and pin them up for self-check at maintenance crew meetings, self-assess once a quarter — three or fewer means healthy, five or more means the system is still at the ‘rule-by-person’ stage; first rebuild the foundation of M01 Sections 1–3 before talking about system software.

8. Putting M01 to work: how to use a self-check list

Methodology’s greatest fear is ‘nod while reading, forget on closing the book’, so this lecture gives a minimal instruction manual: take one week to do three things. First, calculate downtime loss once for your own machine — using the Fig. 1 waterfall, fill in your plant’s speed, width, basis weight, and margin to get the single number ‘1 hour of downtime = how much money’, then post it mill-wide with a unified definition; this one number is worth ten ‘take maintenance seriously’ slogans. Second, sample 50 work orders for a planned-ratio health check — randomly pull 50 maintenance work orders from the last three months, count the planned vs unplanned ratio, then ask of each unplanned order ‘could this failure have been seen in advance’; the answers to that question are your next batch of CBM point lists, more accurate than any consulting report. Third, give the inspection records a ‘trend health check’ — pick any inspection route, pull the vibration/temperature record of a critical pump into a line chart, and see whether anyone watches it, whether there is a threshold, whether there is action; if the answer is three no’s, the inspection system is merely ‘running’, not ‘working’.

Once these three things are done, this lecture’s methodology is no longer a model but three reviewable actions. The reading path of the three lectures also closes here: M01 gives the framework (strategy, inspection, analysis, organization), M02 gives the special account for the most expensive assets (rolls), M03 gives the landing account for the most consumable-dense area (wet end) — framework + two object types validate the same method’s generality across different assets. The later grade-specific volumes (dry end, board, tissue, pulp line) merely extend these two accounts further; the skeleton does not change.