A catalogue life number is not the life of your machine
A fan quoted at 50,000 hours does not deliver 50,000 hours in every cabinet. The number belongs to one set of conditions, and the bearing temperature, speed, mounting orientation and start and stop pattern of the real product are usually not those conditions.
Cooling fan life is often discussed as though the bearing were the only variable. In a populated enclosure the fan also sees air that has already passed a heat sink, dust carried by an intake filter, vibration from a thin panel, repeated restarts, and a shaft orientation that differs from whatever the qualification setup used. None of these appears in a single calendar-hour figure, and each of them moves the failure that finally ends its service life.
The useful question is therefore not which bearing carries the largest number. It is which bearing and lubrication arrangement stays suitable at the actual temperature, speed, load, orientation and duty cycle of this product. Framed that way, the answer is a test and a written specification rather than a badge, and the same framing protects both the equipment designer and the fan supplier.
The word life also hides four different things. ISO 281 defines one of them, a grease manufacturer defines another, an electronics reliability handbook defines a third, and a cooling fan catalogue may quote a fourth without saying which one it is selling. Before any number is compared with another number, the definition behind it has to be identified.
Separate bearing fatigue life from grease life and electronics life
The basic rating life of the bearing steel is a fatigue concept. In a compact fan it is rarely the item that fails first.
The basic rating life L10 in ISO 281 is the number of revolutions, or operating hours at a constant speed, that 90 % of a sufficiently large group of identical bearings will complete or exceed before the first evidence of fatigue appears. Its usual form is L10 = (C / P) raised to p, with p = 3 for ball bearings and 10/3 for roller bearings, and the conversion to hours is L10h = 10⁶ / (60 n) multiplied by L10. It is a statistical rating for a population, not a promise about one fan, and it says nothing about temperature or lubricant condition on its own.
Compact fans complicate that picture in a specific way: bearing load is low relative to the bearing's capacity, and the load term drops out of the concern while temperature and lubricant remain. SKF illustrates the reversal with its own worked examples for lubricated-for-life deep groove ball bearings. A sealed bearing running at a speed parameter of 300,000 and 55 °C reaches a grease life L10 of roughly 30,000 hours. A shielded 6302-2Z at 15,000 rpm, C/P = 8 and 90 °C reaches only about 1,500 hours of grease life, while the calculated fatigue life under the same conditions is about 2,100 hours. The grease ran out first.
ebm-papst states the same conclusion in its fan reliability paper without qualification: most fan failures are caused by wear of the bearing systems, and the single most important factor affecting the reliability of a fan is the composition and reliability of the lubrication system, not the fatigue life of the bearing itself. The paper also warns that a bearing can carry a calculated modified rating life of several hundred thousand hours under ISO 281 and still never reach it, because the lubricant fails first.
That is why a cooling fan catalogue can print three different life values for the same product. ebm-papst does: an L10 at 40 °C, an L10 at the maximum rated temperature, and a life expectancy at 40 °C derived from the averaged calculation methods used in the compact fan market. The first two come from in-house endurance tests in which roughly 1,500 fans are run in temperature cabinets at 40 °C and 70 °C until they fail, with failure defined as a deviation from the defined airflow and speed values or a noticeable change in operating noise. Note the definition: a fan that still spins but no longer delivers its specified airflow counts as failed. Its L10 is then read from a Weibull distribution of those failures.
For a new device, the practical instruction that follows is to measure bearing temperature rather than assume it equals ambient. A fan self-heats through winding and drive losses, and it also receives heat by conduction and radiation from nearby electronics. The resulting temperature, not the room label, is the value that belongs in the life evidence.
What the bearing sees is not the room temperature
Life figures are quoted at an ambient temperature. The item that degrades is the lubricant at the bearing, and the two are separated by everything between the air inlet and the rotor.
A catalogue life at 40 °C means the air around the fan was 40 °C in the test. Inside a real enclosure the fan sits downstream of the heat it is removing, inside a wake from its own impeller, and often within a few centimetres of a hot transformer, inductor or power stage. The bearing temperature that results can exceed the inlet air temperature by a margin that is far larger than any rounding in the life figure.
There are three separate contributions and they have to be separated during measurement. The first is the temperature of the air arriving at the fan inlet, which is an aerodynamic and thermal-design question. The second is the fan's own self-heating, which rises with speed and with supply voltage. The third is heat arriving by conduction through the mounting bracket and by radiation from surrounding components, which is present even when the air path is good.
Measurement practice matters more than instrument cost. A contact sensor bonded to the bearing housing or the stator winding, allowed to reach steady state, tells more than an infrared reading taken across a grille, because infrared instruments average whatever is in their field of view and a small dark target behind a bright grille is the worst case for them. If only a stopped-machine measurement is possible, the reading is dominated by the surrounding metal and not by the bearing.
The same logic explains why a fan that passes a 25 °C laboratory run can fail in a sealed industrial cabinet. The laboratory run establishes function: that the fan starts, reaches speed, and delivers airflow at the specified pressure. It does not establish life, because the bearing temperature during that run is not the bearing temperature in the product.
Two published temperature rules, and they do not agree
When a bearing runs hotter than the rated condition, something has to convert the temperature difference into a life effect. There is more than one published way to do it, and they give different answers.
SKF approaches it from the lubricant side. For a standard grease, the relubrication interval obtained from its diagrams is halved for every 15 °C increase in operating temperature above the 70 °C baseline. Below that baseline the interval may be extended, but SKF also caps the extension at a factor of two and warns that the lower temperature performance limit of the grease should not be approached. The same document notes that for a typical grease life distribution, L10 will be roughly 2.7 times the L01 value.
The compact fan market approaches it from an acceleration-test side. The IPC-9591 zero-failure method used in the interactive model above accelerates time with the factor 1.5 raised to the power of the temperature step divided by 10 °C. A 30 °C step therefore compresses roughly 1.5³, or about 3.4, hours of use into one test hour.
ebm-papst publishes the disagreement openly rather than resolving it. Its catalogue prints side-by-side factors under five different assumptions, including 2 per 10 K, 1.65 per 10 K, 2 per 15 K, a 1.5 per 10 K maximum, and an in-house standard. At 60 °C the factors range from about 0.5 down to about 0.1 relative to the 40 °C reference point in the same table.
Take one concrete case to see how much this matters. A fan whose rated L10 is 70,000 hours at a 40 °C ambient, installed so that its bearing runs at 75 °C. The 15 °C halving rule applied over a 35 °C step gives about a fifth of the rated value. The 1.5-per-10 °C step gives about a quarter. The two published methods, applied to the same fan and the same installation, differ by the better part of 10,000 hours, and both of them are more credible than ignoring the temperature difference entirely.
| Stated as | What it means | What it does not tell you |
|---|---|---|
| L10 at 40 °C | The life 90 % of a tested population reached, with the fan run at a stated speed and supply | The life when the bearing is hotter, faster, dirtier or mounted differently |
| L10 at maximum temperature | The same statistic evaluated at the highest rated ambient | Whether the product actually operates at that temperature or below it |
| Life expectancy | An extrapolation from high-temperature tests using a declared temperature factor | Which factor was used, unless the catalogue says so |
| MTTF | A mean time to failure for a population, which is not the same statistic as L10 | The time by which any stated fraction has failed |
Temperature, speed and duty cycle belong in one test matrix
A single bench run at room temperature cannot represent a fan in a sealed industrial cabinet. The qualification has to cover the corners of the real operating range, not the middle of it.
Test at the lowest and the highest supply voltage, at the required operating point, with hot and cold inlet conditions, across the expected speed range, and in the mounting orientation that the product actually uses. Include the real grille, filter, heat sink and bracket, because they set both the aerodynamic load on the fan and the heat transfer into it.
The duty cycle then has to be reduced to something explicit. Consider an illustrative fan that runs 70 % of the time at 3,000 rpm and 30 % at 1,500 rpm. The time-weighted average speed is 2,550 rpm, and it is worth writing that figure down so that everybody on the project is using the same definition. That average is not a substitute for a life calculation: bearing temperature, acceleration events, start-up torque and the high-speed portion of the cycle each still have to be evaluated separately.
Repeated start and stop cycles deserve their own test rather than a share of a continuous run. A fan that survives thousands of continuous hours may show a different failure mode after thousands of restarts, because start-up torque, commutation current and bearing friction are all highest before the rotor reaches speed. Define the number of cycles, the dwell time in each state, the temperature of each start, and the pass criteria before the sample is built.
For a fan inside a battery cabinet, a machine-tool panel or an outdoor telecommunications enclosure, the matrix usually also includes at least one accelerated thermal test. That is where the temperature models from the previous section are used, and it is the reason the chosen model should be written into the test plan rather than left to whoever analyses the failures.
| Item to define | Why it changes the life result | What to write down |
|---|---|---|
| Bearing temperature | Sets grease viscosity, oxidation rate and the applicable life curve | Measured temperature at the bearing or a validated proxy, with the enclosure closed |
| Speed profile | Changes friction, self-heating and the number of revolutions per year | RPM range and the time fraction spent at each operating point |
| Mounting orientation | Sets the lubricant distribution and the load path in the bearing | Shaft orientation, bracket stiffness and any preload condition |
| Environment | Dust, moisture and vibration accelerate degradation independently of temperature | Exposure class, filter condition and measured vibration profile |
| Start and stop cycle | Applies the highest torque and the highest transient current repeatedly | Cycle count, dwell time, and whether each start is hot or cold |
Orientation decides where the lubricant sits
The same fan, the same temperature and the same speed can behave differently when the shaft is turned through ninety degrees.
The reason is mechanical. A ball bearing in a fan is normally preloaded during manufacture, with a spring or wave washer pressing the rolling elements against the raceways. NMB describes the consequence directly in its ball-versus-sleeve comparison: orientation does not affect the life of a ball bearing fan, because the preload is built in, and the bearing can be mounted in any position.
A sleeve bearing has no such equaliser. Its shaft rides on a film of oil distributed by rotation and gravity, and NMB states that a sleeve bearing fan holds its life only when mounted in the vertical position, with life falling away in other orientations. SKF applies the same physical reasoning to relubrication more generally: bearings on vertical shafts lose grease faster because gravity pulls lubricant away from the contact zone, and its standard correction is to halve the interval calculated for horizontal mounting.
For an equipment designer the practical consequences are easy to state and easy to get wrong. A fan mounted on a vertical wall of a cabinet, blowing horizontally, is not in a vertical-shaft orientation. A fan mounted in the base of an enclosure pushing air upward, or in a roof panel pushing it down, is. The two installations can share a part number and a datasheet and still not share a life figure.
Orientation also interacts with the aerodynamic penalty. A fan installed against a grille, close to a wall, or in the wake of another fan runs at a different operating point, which changes both its speed command and its self-heating. That is a clearance and system-effect question, and it belongs in the same review as the bearing question rather than in a separate one.
Ball, sleeve and fluid bearings do not rank in one fixed order
There is no single ordering in which one bearing construction is always better. The ordering changes with temperature, and that is measurable.
NMB's comparison gives the temperature dependence in numbers. Across 25 °C to 60 °C, ball bearing fans outlasted sleeve bearing fans by roughly 50 % on average. Above 70 °C the comparison stops being a ratio: the ball bearing fans ran about 45,000 hours in the same evaluation, while the sleeve bearing fans were reported as no longer operable. At low ambient temperatures the two constructions lasted about equally well. The ordering is therefore a function of the operating temperature, and a statement that one construction lasts longer without a temperature attached to it is not a specification.
Sleeve bearings have also improved beyond the simple bushing. ebm-papst describes its sintered sleeve bearing as a porous structure in which the pore volume occupies roughly 15 % to 30 % of the bearing, filled with lubricant by vacuum soaking and served by a recirculating reservoir. Delta's fluid-bearing family uses a different mechanism again, with an anti-gravity oil pump circulating oil inside a chamber sealed by a double labyrinth. Those constructions exist precisely because the plain sleeve bearing did not hold its life in every orientation and temperature.
The lesson for a selection exercise is not that one type wins. It is that the comparison has to be made between two specific constructions, at the temperature, speed, orientation and contamination level of the product, using evidence from the supplier of each. A general ranking taken from a blog post or a distributor summary is not evidence, and the manufacturers who publish test data publish it for their own parts.
Where the application has a hard reliability requirement, the review usually ends up asking for a specific construction rather than a category. That is a legitimate request, and it should come with the temperature and operating assumptions attached.
Load, contamination and speed decide which failure arrives first
Beyond temperature, three more variables decide whether the lubricant, the raceway or the electronics ends the life.
Load matters, but only up to a point. SKF's relubrication guidance treats a load ratio C/P below roughly 15 as a significant accelerator of grease consumption, and a load ratio at or above that value as not a significant modifier of the interval. Compact fans normally operate well inside the benign region, which is another way of saying that the load they were designed for is not what limits them.
Bearing type scales the interval further. SKF's criteria place a deep groove ball bearing as the baseline, a cylindrical roller bearing at roughly twice the regreasing frequency, and a spherical roller bearing at roughly five times. Those are not fan bearings, but the ratio is the point: a claim about grease life belongs to a construction, not to bearings in general.
Contamination is the variable most often missing from a life calculation and most often present in the field. Dust that passes an intake filter reaches the bearing, moisture condenses during off periods, and a humid environment changes the chemistry inside the grease. Contamination both degrades the lubricant and removes it from the contact zone, and the effect is not linear in the amount of contamination.
Speed enters twice, and the two effects push in opposite directions. Higher speed increases friction, self-heating and the number of revolutions per hour, which shortens life in hours. It also improves the hydrodynamic film in a sleeve bearing, which is why a sleeve bearing fan may be comfortable at speed and fail when it is run slowly or stopped. A control strategy that parks fans at very low speed for long periods is therefore not automatically the gentle option.
Starting and stopping the fan is a separate qualification
Continuous hours and repeated starts are different loads on the same bearing, and passing one does not demonstrate the other.
In a continuous run the bearing spends almost all of its time in a steady hydrodynamic or rolling state. In an application that starts and stops it spends a large fraction of its life accelerating, which is the condition of highest friction and highest start-up torque, and it spends the remainder stopped, which is when the lubricant drains back and when moisture has its best chance to condense.
The electrical side of a start is a second, independent stress. Delta states on its own approval sheets that the maximum value of the consuming current is not the peak value, and that the peak has to be measured with an oscilloscope. A fan started once at commissioning can tolerate a large inrush without heating anything significantly. A fan started twice an hour for ten years cannot, and the same peak current appears every time.
The operating pattern has to be written before the test, not after the failures. A usable definition states the number of cycles, the on time and off time, the temperature at the start, whether the start is cold or warm, and what counts as a failure. Without those five items, a test with repeated starts produces a result that cannot be compared with anything, including the next test of the same fan.
What to ask a supplier to state in writing
Most of the ambiguity in a comparison of cooling fan life figures comes from two suppliers using the same phrase for different measurements.
The first item is the definition of the number. Ask whether the quoted value is a bearing rating life, a grease life, a life expectancy from an accelerated test, a mean time to failure, or a model-specific result. These are different statistics, and only one of them describes the point by which a stated fraction of the population has failed.
The second item is the condition attached to it. A life figure without an ambient temperature, a speed, a supply voltage and an orientation is not a specification. ebm-papst prints the temperature in the column heading of every life value in its catalogue, which is the standard worth asking others to meet.
The third item is the failure criterion. ebm-papst defines a failed fan as one that deviates from its defined airflow and speed values, or whose operating noise has become noticeable, and it derives L10 from a Weibull distribution of those failures. A supplier that instead reports the time until rotation stops is measuring something real, but it is not the same quantity and it will not compare with a catalogue figure.
The fourth item is the evidence behind the number. Ask whether the value comes from an endurance test on that construction, from an accelerated test with a declared acceleration model, or from a calculation. All three can be useful, and they need to be labelled, because an accelerated test carries the assumptions of its model with it and a calculation carries the assumptions of its inputs.
The fifth item, and the one that connects this article to the rest of a thermal design, is the operating point at which the fan will actually run. Life is quoted at a condition; the condition has to be the one the enclosure produces, which is a question of airflow, pressure, clearance and control rather than of bearings alone.
From a life statement to a validated production sample
For a customized fan, the useful deliverable is not a longer life number. It is a sample that was selected around the real duty cycle and then measured against it.
MEGA Tech builds custom DC and EC cooling fans, including axial and backward-curved centrifugal designs, and selects the bearing and lubrication arrangement around the operating point rather than around a catalogue line. The engineering sequence starts with the equipment heat load and air path, derives the temperature and pressure the fan will actually see, and only then chooses the frame, the impeller, the bearing construction and the control interface. A production-representative sample is then validated in the customer's own enclosure or against its measured operating point, because a fan that passes on an open bench and fails behind a filter has not been validated.
To do that selection properly, we need a defined set of inputs from the equipment side: the required airflow and the static pressure at that airflow, the temperature and altitude basis for the air density, the supply voltage and its tolerance, the control interface and the feedback signal expected, the connector and harness arrangement, the mounting orientation and available space, the acoustic target, and the life assumption the product is designed against. Where those are not yet fixed, the first useful output of the project is the measurement plan that will fix them.
We are explicit about the boundary as well. A fan supplier does not select the coil, design the liquid loop, specify the cabinet controller, take responsibility for the electrical safety of the enclosure, or provide emergency ventilation. Those belong to the equipment design, and a fan meeting its airflow and life figures inside a thermal design that does not work will still fail in the field.
We also do not accept the common shortcut that a fan with the same frame size, voltage and airflow figure is an equivalent replacement. Equivalent installed performance has to be demonstrated at the same operating point, in the same orientation, with the same life assumption attached. That demonstration is what a custom cooling fan project produces, and it is the only version of the answer that survives a design review.
Engineer FAQ: cooling fan bearing life
A reliable fan specification makes its life assumptions visible before production rather than after a field failure.
Q: Does L10 mean the fan will fail at that hour? A: No. L10 is a statistical rating life under defined assumptions, and it describes the point that 90 % of a population reaches or exceeds. The product's own life also depends on grease, temperature, speed, contamination, orientation, vibration and electronics.
Q: Does a ball bearing automatically last longer than a sleeve bearing? A: Not universally. Across 25 °C to 60 °C, ball bearing fans outlasted sleeve bearing fans by roughly 50 % in the comparison NMB publishes, and at lower temperatures the two were about equal. The comparison has to be made between two specific constructions at the temperature, speed and orientation of the product.
Q: Is ambient temperature enough for a fan-life check? A: No. Measure the fan inlet and a bearing-side proxy with the enclosure closed and the real load applied, because recirculation, self-heating and conduction from nearby components all raise the lubricant temperature above the inlet air temperature.
Q: Why do start and stop cycles matter? A: Start-up torque, friction and transient current are all highest before the rotor reaches speed, and a fan that survives continuous operation can show a different failure mode after thousands of restarts. Cycle count and dwell time belong in the test plan.
Q: How much does a hotter bearing actually cost in life? A: It depends on the model, and the published models disagree. Over a 35 °C step above a 40 °C rating, the SKF 15 °C halving rule gives about a fifth of the rated life and the 1.5-per-10 °C acceleration factor gives about a quarter. Either answer is a reason to measure the bearing temperature rather than assume it.
Q: What should a production sample prove? A: Airflow and static pressure at the real operating point, temperature, current, RPM, noise, vibration, start and restart behaviour, and the defined environmental exposure, with the bearing and lubrication construction written into the release documentation.
Questions for a design review
Five questions that separate a cooling fan life specification from a statement about bearing life.
1. What is the defined life statistic, and at what ambient, speed, voltage and orientation was it measured? If the answer is a single number without those four attachments, the number cannot be used in a comparison.
2. What is the bearing-side temperature in the closed enclosure at the worst-case ambient, and how was it measured? An infrared reading through a grille and a bonded contact sensor at steady state are not the same measurement.
3. Which temperature model converts the rated condition to the installed condition, and does the test plan say so explicitly? The 15 °C halving rule and the 1.5-per-10 °C acceleration factor are both defensible, and they disagree.
4. What is the operating pattern, including start count, dwell time, and hot or cold start? A life figure quoted for continuous operation says nothing about a fan that restarts forty times a day.
5. What is the failure criterion that ends the test? A fan counting as failed when it stops rotating and a fan counting as failed when its airflow drops below specification produce two different life numbers for the same part.
