Generator, Pump and Motor Preventive Maintenance

If you want a fast read on the quality of a maintenance programme, do not start with the CMMS reports. Read the PM task lists for the motors, the pumps and the standby generators. Those three classes carry most of the operational risk in a facility, they have well-understood failure modes, and there is decades of engineering consensus on how to look after them. If the tasks are vague, the measurements missing, and the generator has never seen a load bank, you have learned everything you need to know about the rest of the programme. This guide sets out what those task lists should contain and why each line is there.

The message up front: rotating equipment PM only works when tasks produce numbers. "Check bearing" tells you nothing next year. "Record bearing temperature, vibration overall velocity and grease quantity applied" builds a trend that catches the failure before it happens. And rotating equipment is the single best candidate class for condition monitoring, so the goal of a good PM programme here is not to keep doing fixed-interval work forever, it is to build the baseline that lets you move the critical units onto condition.

1. Why rotating equipment deserves its own PM discipline

Rotating equipment differs from static plant in one important way: it degrades along a path you can watch. A bearing does not go from perfect to seized in an instant. It develops a defect, the defect generates a high-frequency signature, that signature grows, heat follows, lubricant degrades, clearance opens up, and eventually the machine fails. That progression can run for weeks or months. Compare it with a control board that works until the moment it does not.

This is why the preventive tasks on these assets should be built as measurement tasks rather than inspection tasks. Every reading is a point on a trend. A single reading is nearly worthless; twelve monthly readings tell you whether the machine is stable or drifting, and drift is the whole game. For how condition data turns into intervention decisions, see the predictive maintenance and failure prediction pillar.

The second reason is consequence. A chilled water pump, a fire pump, a main LV supply motor and a standby generator each sit on a path where failure stops something that matters, which puts them high on any sensible criticality ranking, and criticality is what should decide how much PM they get. The asset criticality classification pillar covers that ranking.

2. Electric motor preventive maintenance: the core tasks

An electric motor is a simple machine with a small number of well-characterised failure modes. Industry failure studies have consistently found that bearings account for the largest share of motor failures, with stator winding insulation the next largest, followed by rotor problems and everything else. That distribution should drive the task list directly: most of your effort belongs on bearings and on insulation condition, because that is where the failures are.

The core electric motor preventive maintenance tasks, in the order I would put them on a task list:

• Insulation resistance test. A megohmmeter reading from winding to earth at the voltage appropriate to the machine rating, corrected to a standard temperature, usually 40 degrees Celsius. The absolute value matters less than the trend: a motor that read 500 megohms last year and 80 this year is telling you something even though 80 still passes. Record test voltage, ambient and winding temperature and humidity, because without those the reading is not comparable to the next one.
• Polarisation index. The ratio of the ten-minute insulation resistance reading to the one-minute reading. It separates genuinely aged insulation from insulation that is merely damp or dirty, because moisture and contamination hold the reading flat while healthy insulation keeps climbing as the dielectric polarises. The most informative electrical test on a large motor, routinely skipped because it takes ten minutes instead of one. IEEE publishes the recognised guidance on insulation testing of rotating machinery; the catalogue is at ieee.org .
• Winding temperature. From embedded RTDs where fitted, or a thermographic survey of the frame where not. Insulation life is dominated by temperature: the long-standing rule of thumb is that a sustained rise of around ten degrees Celsius above design operating temperature roughly halves insulation life. A motor running hot is consuming its remaining life at double rate.
• Bearing condition and lubrication. Temperature, vibration and grease. The most commonly botched task in rotating equipment maintenance, and it gets its own section below because the usual failure is over-greasing rather than under-greasing.
• Vibration reading against baseline. Overall velocity in millimetres per second RMS at the drive end and non-drive end, horizontal, vertical and axial, at the same points with the same mounting every time, compared to the machine's own commissioning baseline first and a published severity chart second. ISO 20816 is the current reference family for vibration evaluation on non-rotating parts; the catalogue is at iso.org .
• Running current and voltage balance. Current on all three phases under a known load, plus voltage balance. Unbalance is punishing: a small percentage of voltage unbalance produces a much larger percentage of current unbalance and a disproportionate rise in winding temperature. Cheap to measure, commonly ignored.
• Cooling path. Fan cover clear, fins free of dust and lint, airflow unobstructed, filters clean on enclosed machines. A motor coated in process dust runs above its design temperature whether or not anyone has measured it.
• Mounting, alignment and soft foot. Hold-down bolt torque, base condition, grout, coupling alignment. Its own section below, because alignment is where a large share of premature bearing and seal failures originate.
• Terminal box and connections. Thermographic scan under load where possible, torque check when de-energised, gland and cable entry integrity, earth continuity. The electrical preventive maintenance pillar covers the wider EPM discipline around this.

A note on frequency. Insulation testing on a running production motor generally requires an outage, so it lands annually for most machines and at every available shutdown for critical ones. Vibration and temperature readings need nothing but access, so they can be monthly or continuous. Do not let the hardest task set the rhythm for the easy ones.

3. Bearing lubrication: the most commonly botched task on the list

This deserves a full section because it is the task most likely to be done, most likely to be signed off, and most likely to be doing harm. Grease-lubricated bearings fail from lubrication problems more often than anything else, and the problem is at least as often too much grease as too little.

Over-greasing fails a bearing by a specific mechanism. A housing packed full of grease leaves no space for grease to be churned out of the rolling path, so the rolling elements plough through it continuously and generate heat. The heat thins the base oil and drives it out of the thickener, and the remaining soap dries into a stiff residue that no longer lubricates. Worse, on a motor the excess is pushed along the shaft into the windings, where it contaminates the insulation and attracts dust. A technician who diligently pumps until grease appears has, in one action, cooked the bearing and contaminated the stator.

One more honest point: sealed-for-life bearings are fitted to a great many small and medium motors and have no regrease task at all. If your plan has a quarterly regrease on a motor with sealed bearings, the technician is either skipping and signing it, or forcing grease into a housing that cannot accept it. Verify which bearings are actually fitted before writing the task.

4. Alignment, soft foot and the failures they cause

A large proportion of premature bearing and mechanical seal failures on coupled machines trace back to misalignment, which is almost always maintenance-induced rather than a wear-out fault. Somebody rebuilt the pump, refitted it, coupled it up by eye, and handed back a machine that will destroy its own bearings over the following months.

• Laser alignment, not straightedge. A dial indicator is acceptable in skilled hands; a straightedge and feeler gauge is not, on anything that matters. Laser tools have been affordable for years.
• Soft foot first, always. Soft foot is the condition where the machine feet do not all sit flat on the base, so tightening the bolts distorts the frame and pulls the shaft out of position. Check it by loosening each foot in turn with an indicator on the shaft and observing movement. Aligning before clearing soft foot produces an alignment that changes the moment anyone touches a bolt.
• Thermal growth. A pump handling hot fluid and a motor at ambient will not stay aligned once at operating temperature. Cold alignment on hot service must apply the manufacturer's thermal offset targets, otherwise you have aligned the machine in the one state in which it never runs.
• Pipe strain. Pipework pulled into place by the flange bolts distorts the casing and shifts the shaft. Check alignment with flanges loose and again tight; a shift between the two is pipe strain, and no amount of shimming fixes it.
• Record the final numbers. Angular and offset values in both planes and the shim thicknesses used. That record is the baseline for the next intervention and the evidence when a bearing fails early.

Alignment belongs in the plan twice: as a verification task at a suitable interval on critical coupled machines, and as a mandatory step in the corrective procedure for any unit removed and refitted. The second matters more. If corrective work orders do not require an alignment record before closure, you are generating misaligned machines faster than the PM programme can find them. The work order types pillar covers how corrective and preventive types should carry different mandatory completion fields.

5. Pump preventive maintenance: seals, suction and the performance curve

Pumps add a set of failure modes that motors do not have, because a pump has to deal with a fluid. The dominant pump failures are mechanical seal failure, bearing failure and cavitation damage, and all three are influenced heavily by conditions upstream of the pump rather than by the pump itself.

Mechanical seal versus packing. These are different maintenance regimes and confusing them produces bad task lists.

• Gland packing is designed to leak. A controlled drip carries away the frictional heat at the packing face. A gland tightened until it stops dripping will run hot, glaze, score the shaft sleeve and fail. The task is therefore "verify leakage is a steady drip, adjust gland nuts by a defined small increment if excessive, record", not "tighten until dry". Packing also needs periodic replacement rather than endless adjustment.
• Mechanical seals are designed not to leak. Any visible leakage is a failure in progress, not a tolerance. The task is visual inspection plus, on seals with a support system, verification of the barrier or flush fluid: seal pot level, pressure, temperature and cleanliness, and flow through the flush line. Seals fail overwhelmingly because their environment failed first: dry running, flush blockage, contaminated barrier fluid, or vibration from misalignment.
• Conversion. Sites often convert packing to mechanical seals for water conservation. If you do, update the task list. I have seen packed-gland adjustment tasks still running on pumps converted to cartridge seals three years earlier.

Suction conditions, NPSH and cavitation. Cavitation is vapour bubbles forming in the low-pressure region at the pump inlet and collapsing violently as pressure recovers, pitting the impeller and hammering the bearings. It happens when the net positive suction head available falls below the net positive suction head required by the pump at that flow. Technicians should be trained to recognise the symptoms: a sound like gravel being pumped, erratic discharge pressure, fluctuating motor current, elevated vibration of broadband rather than tonal character, and pitting on the impeller leading edge at strip-down.

The point for a PM programme is that cavitation is almost never fixed at the pump. The causes are upstream: a blocked or partially closed suction strainer, a suction valve not fully open, a dropping supply level, a suction line air leak, fluid hotter than design, or a pump run far out on its curve. So the task list must cover the suction side, not just the machine: strainer differential pressure, suction valve confirmed fully open, suction pressure, and supply level.

Wear ring clearance and performance curve drift. This is the pump equivalent of a vibration trend and it is criminally underused. A centrifugal pump has close-clearance wear rings between impeller and casing. As those clearances open up, internal recirculation increases and the pump delivers less head at the same speed. Record suction pressure, discharge pressure and motor current at a known valve position each month and you have a performance trend. A pump whose developed head at the same duty point has fallen noticeably over a year has worn clearances, so you can plan the overhaul rather than discover it. Manufacturers publish a maximum allowable wear ring clearance, typically around twice the new clearance, and that is the number that triggers replacement at strip-down.

Priming and dry-run protection. A centrifugal pump run dry destroys its seal in seconds. Non-self-priming pumps in suction-lift service must be verified primed before start, with the foot valve and priming arrangement on the PM list. On any pump that can lose suction, verify the dry-run protection: low-suction-pressure trip, minimum-flow protection, or a thermal or power-based dry-run relay. Test it rather than inspect it. A trip that has never been proven is not protection.

6. Standby generator testing: the weekly run is not enough

Standby generators have the widest gap between the PM that is done and the PM that is needed. Almost every site runs a weekly no-load exercise; very few run a proper annual load bank test. The weekly run creates a comfortable feeling of coverage while leaving the most important question unanswered.

Why the weekly no-load run is insufficient on its own. It proves a narrow set of things: the batteries can crank, the starter works, the engine starts and reaches speed, the fuel system delivers, the alternator produces voltage. Those are worth proving weekly. What it does not prove is that the set can carry its rated load, that the cooling system can dissipate full-load heat, that the fuel supply sustains full consumption, that the exhaust and turbocharger are healthy under load, or that the transfer actually works.

Worse, repeated no-load or very light-load running damages a diesel engine. Cylinder temperatures never reach the point where combustion is complete, so unburnt fuel and carbon accumulate on the valves, in the exhaust and around the turbocharger, and oil is drawn past rings that never properly seat under low cylinder pressure. The result is wet stacking, visible as oily residue at the exhaust outlet. A generator maintained only by weekly light running is being degraded by the very task intended to protect it.

The tiered generator maintenance schedule I would recommend:

• Weekly, no load or light load, short duration. Starts, reaches speed and voltage, no alarms, fluid levels, no leaks, charger in float, block heater working. Record start time in seconds and any abnormality. Keep it short precisely because light running is harmful.
• Monthly, on load. Transfer the actual building load to the set via the transfer switch and run long enough for the engine to reach full operating temperature. This proves the transfer switch, load-carrying capability at real load, the cooling system, and the engine under genuine cylinder pressure. It is the test that turns a weekly exercise into meaningful assurance. NFPA 110, the standard for emergency and standby power systems, sets out the recognised regime for life-safety installations; see nfpa.org for the current edition alongside your local authority requirements.
• Annually, resistive load bank to full rating. Where the connected load cannot exercise the set properly, and in most buildings it cannot, a load bank is the only way to prove the set at its rating. A staged test ramping through the load range to full rated output, held long enough for temperatures to stabilise, proves cooling, fuel delivery, governor and voltage regulator stability and turbocharger condition, and burns off accumulated wet-stacking deposits. Record voltage, frequency, current per phase, coolant temperature, oil pressure and exhaust temperature at each step.
• Fuel management. Diesel degrades in storage: it oxidises, forms sediment and gums, and absorbs water from tank condensation, which supports microbial growth at the fuel and water interface. That growth produces sludge that blocks filters at exactly the moment the set must run for hours. Sample and test the fuel at least annually for water, sediment and microbial contamination; drain the tank low point; polish or filter on a defined cycle; maintain a biocide and stability treatment regime; and replace fuel filters on schedule rather than on failure.
• Batteries and the starting system. Battery failure is consistently the leading cause of standby generator start failures, which is disproportionate to the attention batteries receive. Check terminal cleanliness and torque, electrolyte level and specific gravity where the type permits, float voltage from the charger, and run a load or conductance test rather than reading voltage. A battery at 12.7 volts at rest can still be incapable of cranking. Replace on age rather than on failure, and record the installation date on the battery itself.
• Coolant and block heater. Condition matters as much as level: test inhibitor concentration and freeze or boil protection annually, because depleted inhibitor allows internal corrosion and cavitation erosion of cylinder liners. The block heater is what lets a cold set accept load quickly, and its failure is invisible until the day it matters. Verify it by measuring block temperature at rest, not by looking at the element.
• Lubrication and air. Oil and filter changes on hours or calendar, whichever comes first, because a standby set accumulates few hours but the oil still degrades with time and contamination. Air filter restriction indicator checked, intake path clear.
• Transfer switch. Exercise it, do not just inspect it. An automatic transfer switch that has not operated under load is an untested single point of failure between a working generator and the load it feeds. Exercise transfer and retransfer, verify timing settings, and thermographically survey the power connections under load.

7. The combined task table: motors, pumps and generators

This is the reference table I would put in front of a planner building or reviewing these PM plans. Frequencies are a starting point for typical commercial and light industrial service on a moderately critical asset. Adjust them up for critical duty, harsh environment or continuous operation, and adjust them down for low-criticality intermittent units. The measurement column is the part that matters most: if a task has no measurement, it produces no trend, and if it produces no trend it cannot tell you anything next year.

If you want a broader library of task lists in this format for other asset classes, the PM checklists, templates and examples pillar covers how to structure them so that they stay usable at scale.

8. Why rotating equipment is the best candidate for condition monitoring

Everything in the table above is fixed-interval work, and fixed-interval work is blind to actual condition, so it over-maintains the healthy units and can still be surprised by the one that degrades between intervals. For most asset classes you live with that. Rotating equipment is the exception, for three reasons.

• The dominant failure modes are detectable with useful warning. Bearing defects, misalignment, imbalance, looseness, lubrication distress and rotor faults all produce measurable signatures well before functional failure, and the interval between detection and failure is usually long enough to plan around. That is the precondition for prediction being possible at all.
• The techniques are mature and inexpensive. Vibration analysis, thermography, oil analysis, ultrasound and current signature analysis have decades of practice behind them, published severity criteria, and a wide market of tools and trained analysts. You are not pioneering anything.
• There are usually many similar units. A site with forty similar pumps generates comparative data, so the outlier stands out against its peers before it crosses an absolute threshold. That population effect is exactly what a site with three unique large assets does not have.

So the sensible trajectory is disciplined measurement-based PM first, used to build the baseline and clean the history, then move the critical units onto condition-based intervals while leaving low-consequence units on simple time-based tasks or run-to-failure. That progression is what the preventive versus predictive versus reactive comparison lays out, and the mechanics of the condition layer are in the failure prediction pillar. One caution: vibration analysis on an unlubricated bearing tells you the bearing is failing, it does not grease it. The condition layer sits on top of good fundamentals, it does not replace them.

9. Setting frequencies without guessing

The most common question on these task lists is how often. The frequency should come from five inputs, in this order of authority.

• Statutory and insurance requirements. Fire pumps, life-safety generators and certain pressure and electrical equipment carry mandated regimes. Not negotiable, and they set the floor.
• Manufacturer recommendation. The starting point for everything else, and the defensible position if something fails. Where you deviate, document the engineering reasoning.
• Criticality. Identical assets can legitimately carry different PM plans when their consequence of failure differs. A programme that ignores this is over-maintaining the unimportant units or under-maintaining the important ones.
• Operating context and failure history. Duty cycle, ambient, dust, humidity, corrosive atmosphere, and above all your own failure history for that class. If a failure mode recurs inside the current interval, the interval is wrong. That feedback loop needs disciplined coding; see the failure codes pillar.
• Condition data, once you have it. A machine with a stable trend over two years does not need the same intrusive interval as one that drifts. Extending intervals on evidence is the reward for having built the trend.

The frameworks for choosing between time-based, meter-based and condition-based triggers sit in the preventive maintenance strategies pillar, and the overall programme structure in the complete guide to preventive maintenance.

10. The recurring mistakes on these three asset classes

Across the plans I have reviewed, the same errors appear again and again.

• Regrease tasks on sealed bearings. Signed off monthly on a bearing with no grease nipple. Nobody has walked the plant against the plan.
• Grease guns with no quantity control. The instruction says "lubricate", the technician pumps until grease appears, and the result is cooked bearings and contaminated windings.
• Insulation readings taken without temperature. Insulation resistance varies strongly with temperature, so a reading without it cannot be compared with any other and the whole exercise produces a number used once and discarded.
• Alignment not required after a refit. The plan verifies alignment annually but the corrective procedure does not mandate an alignment record, so misalignment is introduced faster than it is found.
• Pump PM that stops at the pump. No strainer differential, no suction valve check, no supply level. Cavitation damage accumulates and is written up at strip-down as impeller wear rather than a suction problem that will destroy the replacement too.
• Performance readings not taken. The cheapest condition indicator on a pump, and most sites do not record it, so wear ring clearance is discovered only when the pump is opened.
• Weekly generator run treated as the whole programme. No monthly on-load run, no load bank, no fuel analysis, and the compliance report looks green throughout.
• Batteries checked by voltmeter only. Resting voltage says almost nothing about cranking capability, yet it is the only test on most plans, on the component causing the most start failures.
• Transfer switch inspected but never exercised. The one component that must operate to connect the set to the load is never proven.
• Standby units never rotated into duty. The redundancy exists on the drawing and has seized in reality.

None of these is a technology problem and none needs a budget. They are task-quality and procedure problems, entirely within the control of whoever writes the plan.

The idea to walk away with

The preventive tasks that actually protect motors, pumps and generators share one property: they produce a number. Insulation resistance with its temperature. Grease in grams. Vibration in millimetres per second against a baseline. Developed head at a known duty. Battery conductance rather than resting volts. Load bank output at rated load. Tasks that produce numbers build trends, and trends let you intervene on your schedule instead of the failure's.

The three things I would fix first on any programme covering these classes: rewrite the bearing lubrication tasks with a quantity in grams and a purge instruction, add suction-side and performance-point readings to every pump task list, and put a monthly on-load generator run and an annual load bank test into the plan, with a documented risk acceptance if the load bank genuinely cannot be arranged. Those three changes cost very little and close the largest gaps in a typical plan.

Final thoughts

This is one of the few areas where the engineering consensus is settled and the gap between good practice and common practice is a matter of discipline rather than knowledge. Everyone knows a generator should be load tested, that over-greasing is harmful, that misalignment destroys bearings. The plans still do not say so, because plans get written once, copied between sites, and never walked against the actual equipment.

So if you do one thing after reading this, take the task list for your most critical pump set and walk it at the machine with a technician. Do the bearings named in the task exist, is the grease quantity stated, is the suction strainer on the list, can the alignment record from the last refit be produced, and can the last twelve months of readings be laid out as a trend? What that walk turns up will tell you more about the programme than any report the CMMS can print, and it produces a specific, fundable list of fixes rather than a general feeling that maintenance could be better.

Related reading: Preventive maintenance: the complete guide, PM checklists, templates and examples, Electrical preventive maintenance checklists, Predictive maintenance and failure prediction, Preventive vs predictive vs reactive, Asset criticality classification.