Bearing Life L10 Calculator

L10 = (C/P)^p in millions of revolutions and in hours, with p = 3 for ball and 10/3 for roller bearings — plus the load lever, both published a₁ tables, and the rating C a catalogue must show for your target life.

Last updated 24 Sep 2026

Bearing L10 rating life

C, P, speed → rating life in hours

Basic dynamic load rating C (kN)From the bearing’s own catalogue row — C, or C_r for a radial bearing. It is the load for which the basic rating life is one million revolutions, and it is a DEFINITION, not a load the bearing can carry indefinitely.

Equivalent dynamic load P (kN)One number combining the radial and axial load. If you have F_r and F_a separately, get P from the equivalent load page first — P = X·F_r + Y·F_a with X and Y switching at a limit that itself moves.

Speed (rev/min)

Bearing typeThe exponent is the whole difference: 3 for a ball bearing, 10/3 for a roller bearing. 10/3 is a real exponent fitted to roller test data, not a rounding of 3 — at C/P = 4 it gives 1.59 times the life.

Reliability you want90 % is the basis: L10 is the life by which one bearing in ten has failed. Asking for more reliability does not make the bearing better, it just reports an earlier point on the same distribution.

Which a₁ tableThe two editions of ISO 281 publish different a₁ columns and both are still in circulation — 0.64 against 0.62 at 95 %, and 0.25 against 0.21 at 99 %. Mixing them is a real error; this page makes you choose.

Life modification factor a_ISOFrom the bearing maker’s own chart, which this page will not invent. It is read against two arguments — the viscosity ratio κ and the contamination-corrected load ratio e_C·C_u/P — and it is capped at 50. Leave it at 1 and the answer is the basic rating life.

Target life you want (hours)What the machine is meant to last. The page turns it back into the two numbers you can actually shop with: the rating C a catalogue must show, and the load P the bearing you have would reach it at.

Not a circuit: the cube law drawn in normalised coordinates, which is why one curve serves every bearing. The horizontal axis is the load as a multiple of the one you entered and the vertical axis is the life as a multiple of the life you got, so the curve passes through (1, 1) whatever your numbers are. The steep line is the ball bearing's cube law; when you select a roller bearing a second, steeper curve appears for the 10/3 exponent. The shallow line is the linear guess almost everyone makes — that half the load is twice the life. Read the marked halving: at half the load the cube law gives EIGHT times the life and the linear guess gives two. Then read it the other way, which is the way that matters in service: at 1.25 times the load you keep half the life, and at 1.5 times you keep thirty per cent. That is why a machine that grew a heavier attachment eats bearings. The figures on the right are your own bearing's numbers, including the basic dynamic load rating a catalogue must show to reach your target life.

3,157hExample

A 6208-size ball bearing rated C = 32.5 kN, carrying an equivalent load of 5 kN at 1,450 rev/min, with a₁ = 1 (90 % reliability) and a_ISO = 1

The rating life, and the two factors that modify it

L10 = (C/P)^p · p = 3 (ball), 10/3 (roller) · L10h = 10⁶·L10/(60n) · L_nm = a₁·a_ISO·L10 · C_required = P·(L·60n/10⁶)^(1/p)

L10: basic rating life: the life 90 per cent of a population of identical bearings reaches or exceeds, in millions of revolutions. Ten per cent have failed by then
C: basic dynamic load rating — the load for which L10 is one million revolutions. A catalogue number and a definition, not a permissible load
P: equivalent dynamic load: the one radial load that would do the same fatigue damage as the actual combination of radial and axial load
p: 3 for ball bearings, 10/3 for roller bearings. Fitted to test data, not chosen for tidiness
a₁: life modification factor for reliability. 1 at 90 per cent by definition; below 1 for anything more demanding, because you are reading an earlier point on the same distribution
a_ISO: life modification factor for the lubrication and contamination condition. Read from the maker’s chart against κ and e_C·C_u/P, capped at 50, and NOT reproducible from a formula

Worked example

A 6208-size ball bearing rated C = 32.5 kN, carrying an equivalent load of 5 kN at 1,450 rev/min, with a₁ = 1 (90 % reliability) and a_ISO = 1
Start with the load ratio, because everything follows from it: C/P = 32,500/5,000 = 6.5. Nothing else on this page matters as much as this one number
The basic rating life is that ratio cubed, in millions of revolutions: L10 = 6.5³ = 274.63 million revolutions. For a roller bearing of the same rating it would be 6.5^(10/3) = 512.5 million, which is 87 per cent more — the exponent is not a detail
Now hours, which is what the question was really about: L10h = 10⁶·L10/(60n) = 10⁶ × 274.63/(60 × 1,450) = 3,157 hours. That is about 132 days of continuous running, or roughly 0.8 years of a single-shift duty
AND HERE IS WHAT L10 MEANS. It does not mean the bearing lasts 3,157 hours. It means that if you fitted a hundred of them, about ten would have failed by then and about ninety would still be running. Half of them are still running at roughly five times that figure. If you need to know when to change the bearing, L10 is the number; if you need to know when THIS bearing will fail, no calculation will tell you
The load lever. Halve the load to 2.5 kN and the life becomes 2³ = eight times — 25,253 hours. Raise it by a quarter to 6.25 kN and you keep 51 per cent of it, 1,616 hours. This is the only cheap lever on the page, and it is usually pulled by moving the load closer to the bearing rather than by buying a bigger one
Turn it round for the answer you can shop with. To reach 20,000 hours at this load and speed you need C = P(L·60n/10⁶)^(1/3) = 60,139 N, which is 185 per cent of the rating you have — so this bearing will not do it and the next size up is the conversation. Or keep the bearing and cut the load to 2,702 N
Two honesty notes. a₁ is a published table and this page carries both of the ones in circulation, because ISO 281:1990 and ISO 281:2007 disagree — 0.21 against 0.25 at 99 % reliability. a_ISO is not a table at all: it comes from a chart read against the viscosity ratio and a contamination factor, it is capped at 50, and this page takes it as an input rather than inventing a curve for it

Life is cubic in load: what a change of load actually costs

This is the most actionable table on the page, and it is worth reading in the direction nobody reads it. Ten per cent more load costs a quarter of the life. Twenty-five per cent more load costs half of it. Fifty per cent more load leaves you with thirty per cent. That is why a bearing arrangement that passed on paper fails on a machine that grew a heavier attachment, a tighter belt or a misaligned coupling — and why the cheapest life improvement available is almost never a bigger bearing, it is a lower load. Tension a belt correctly and the shaft load falls; move the pulley closer to the bearing and the load falls; take out the misalignment and the load falls. Each of those is cubed. This page sizes a part; it does not certify one. Where the answer carries a consequence — a load path, a lifting duty, a pressure boundary, a fastener holding something that can fall — confirm it against the design code that governs the application, and against the manufacturer’s own rating, before relying on it.
Load	Multiplier	Ball bearing life	Roller life	Hours, for this page’s example
half the load	0.50×	8.000×	10.079×	25,253 h
70 per cent of the load	0.70×	2.915×	3.284×	9,203 h
80 per cent of the load	0.80×	1.953×	2.104×	6,165 h
90 per cent of the load	0.90×	1.372×	1.421×	4,330 h
exactly the load	1.00×	1.000×	1.000×	3,157 h
110 per cent of the load	1.10×	0.751×	0.728×	2,372 h
125 per cent of the load	1.25×	0.512×	0.475×	1,616 h
150 per cent of the load	1.50×	0.296×	0.259×	935 h
twice the load	2.00×	0.125×	0.099×	395 h

The reliability factor a₁ — two editions, two columns

Two published columns, both in current circulation, and a figure taken from one and used with the other is simply wrong — by four per cent at 95 % reliability and by nineteen per cent at 99 %. ISO 281:2007’s own published introduction says so in as many words: “the life modification factors for reliability, a₁, have been slightly changed and extended to 99,95 % reliability”. NSK’s and NTN’s catalogues and American Roller Bearing print the 2007 column; AST Bearings and NTN-SNR’s own live Bearing Wizard still print the 1990 one. The last column is this page’s check on which is which, and it is the reason both are offered rather than one being called a misprint: bearing fatigue is conventionally a two-parameter Weibull distribution, for which a₁ = [ln(1/R)/ln(1/0.9)]^(1/β), and at a Weibull slope β = 1.5 that reproduces the 1990 column to the last printed digit at five reliabilities running. The 2007 column is not any single Weibull slope — solve for β row by row and it drifts from 1.61 to 1.70, which is the signature of the truncated distribution the 2007 edition adopted. The same designation can mean different dimensions in different standards families — ANSI against ISO, inch against metric, one national standard against another. The family used here is named beside every figure; check which one your part was made to.
Reliability	Life symbol	a₁, ISO 281:2007	a₁, ISO 281:1990	2007 higher by	A Weibull slope of 1.5 gives
90.0%	L10	1.000	1.000	0.0%	1.0000
95.0%	L5	0.640	0.620	3.2%	0.6189
96.0%	L4	0.550	0.530	3.8%	0.5315
97.0%	L3	0.470	0.440	6.8%	0.4372
98.0%	L2	0.370	0.330	12.1%	0.3325
99.0%	L1	0.250	0.210	19.0%	0.2088
99.9%	L0.1	0.093	0.037	151.4%	0.0448

Why 10/3 is an exponent and not a rounded 3

The two exponents agree only at C/P = 1, and they diverge as fast as the life does. At C/P = 4 — a very ordinary sizing — the roller exponent gives 101.6 million revolutions against 64, which is 59 per cent more life from the same load ratio. Use 3 for a tapered roller bearing and you will undersize nothing, but you will report a life a third short and may buy a bigger bearing than you need; use 10/3 for a ball bearing and you will overstate the life, which is the dangerous direction. Note also that the exponent applies to the LOAD RATIO and not to the load, so the difference between the two bearings’ laws is not a correction you can fold into a service factor. These dimensions come from a published standard’s table, not from a formula. The standard itself is cited below and the printed values are attributed to the catalogue they were taken from; a different publisher may round differently in the last digit.
C/P	L10, ball (C/P)³	L10, roller (C/P)^(10/3)	Roller life higher by
2	8.0	10.1	26.0%
3	27.0	38.9	44.2%
4	64.0	101.6	58.7%
5	125.0	213.7	71.0%
6	216.0	392.5	81.7%
8	512.0	1,024.0	100.0%
10	1,000.0	2,154.4	115.4%

What L10 means, what the load lever is worth, and the two factors that modify it

L10 does not mean the bearing lasts that long. It means one in ten has failed by then. That is the single most misread number in machine design. The basic rating life is a tenth percentile: fit a hundred identical bearings under identical load and about ten will have failed at L10, about ninety will still be turning, and the median life is roughly five times L10. So L10 is the right number for a maintenance interval across a fleet and the wrong number for “how long will this one last”. Nothing you can calculate answers the second question, because the spread is a property of the steel and not of your arithmetic.

Life is cubic in load, and that is the whole of bearing engineering. L10 = (C/P)³ for a ball bearing. Halve the load and you get eight times the life. Add ten per cent and you lose a quarter of it. Add fifty per cent and you keep thirty per cent. Read it that second way and the design consequence is obvious: the cheapest life improvement available is almost never a bigger bearing, it is a smaller load — a correctly tensioned belt, an overhung pulley moved closer to its bearing, a misalignment taken out, a coupling that no longer pulls sideways. Each of those is cubed. Buying the next bearing size up typically raises C by fifteen to twenty-five per cent, which is a life multiplier of 1.5 to 2; moving a pulley in by a third often halves the load, which is a multiplier of 8.

The exponent is 3 for balls and 10/3 for rollers, and 10/3 is not a rounded 3. It is a fitted exponent, and it matters: at C/P = 4 the two give 64 and 101.6 million revolutions, a difference of 59 per cent. The reason is the contact. A ball touches its raceway at a point and the stressed volume grows one way with load; a roller touches along a line and it grows another. Use the wrong one and you are out by a third to a half in the direction you chose by accident.

a₁ is a table this page will give you twice, and a_ISO is a chart it refuses to invent. ISO 281’s modified rating life is L_nm = a₁·a_ISO·L10. The reliability factor a₁ is published — and published DIFFERENTLY by the 1990 and 2007 editions of the standard, both of which are still in circulation, so this page makes you say which you want and shows you the other. a_ISO is a different kind of object: it is read off a chart against the viscosity ratio κ and a contamination-corrected load ratio, there is one chart per bearing family, and it is capped at 50. There is no honest formula for it. This page takes it as an input, tells you what the chart’s two arguments are, and computes one of them on the lubrication page. Inventing a curve for a_ISO would be the worst thing this page could do, because it is the factor that can move the answer by a factor of ten.

What this calculation cannot see. Rating life models subsurface fatigue and nothing else. It does not see a bearing run too lightly, which skids and smears — the failure mode that gets worse as the load falls, and the one the equivalent load page checks with a minimum load. It does not see an interference fit that has eaten the internal clearance and preloaded the bearing, which the preload and axial play page computes. It does not see contamination, which is what a_ISO is for and which is the commonest real cause of early failure. It does not see the shaft’s own slope at the bearing, which the shaft deflection and slope calculator gives and which misaligns the rings. And it does not see the grease drying out, which the relubrication interval addresses. Most bearings that fail early fail for one of those and not for fatigue.

Frequently asked questions

Does L10 mean the bearing will last that long?

No, and this is the most important thing on the page. L10 is the life by which ten per cent of a population of identical bearings has failed — ninety per cent are still running. The median life is roughly five times L10. Use L10 to set a maintenance interval or to compare two bearing selections; do not use it to predict when a particular bearing will fail, because the scatter is a property of the fatigue process and no calculation removes it.

Why is the exponent 10/3 for roller bearings?

Because a roller contacts its raceway along a line and a ball contacts at a point, so the stressed volume under the contact grows differently with load. 10/3 is a fitted exponent from roller bearing test data and it is a real number, not a rounding of 3: at C/P = 4 it gives 101.6 million revolutions against 64, which is 59 per cent more life. Cylindrical, tapered, spherical and needle roller bearings take 10/3; every kind of ball bearing takes 3.

What actually improves bearing life the most?

Reducing the load, by a distance. Because life goes as the cube of the load ratio, a twenty per cent load reduction is a ninety-five per cent life increase, while the next bearing size up usually raises C by fifteen to twenty-five per cent and so roughly doubles it at best. In practice the load reductions available are: tension the belt correctly, move an overhung load closer to its bearing, remove misalignment, and stop the coupling pulling sideways. After load, the biggest lever is cleanliness, which acts through a_ISO and can be worth a factor of several on its own.

What is a_ISO and why will this page not compute it?

a_ISO is ISO 281’s life modification factor for the lubrication and contamination condition, and it is read from a chart with two arguments: the viscosity ratio κ = ν/ν₁, and the contamination factor times the fatigue load limit over the load, e_C·C_u/P. There is a different chart for each bearing family, the published implementations cap it at 50 and cap κ at 4, and there is no closed form. Reproducing that from memory would be inventing a curve, and since a_ISO can move the answer by a factor of ten in either direction it is the worst place on the page to guess. It is an input here, with its two chart arguments named so you know what to look up.

Which a₁ table should I use?

The one that matches the standard your calculation claims to follow, and say which it is. ISO 281:2007 gives 1.00, 0.64, 0.55, 0.47, 0.37 and 0.25 at 90 to 99 per cent reliability; ISO 281:1990 gives 1.00, 0.62, 0.53, 0.44, 0.33 and 0.21. Both are printed by current manufacturer sources. The 1990 column is exactly a two-parameter Weibull distribution with a slope of 1.5, which is how this page verified it; the 2007 column comes from a truncated distribution and is the more generous of the two at every reliability.

My bearing is barely loaded. Does that mean infinite life?

No — it means a different failure mode, and one this calculation cannot see. Below roughly two per cent of the basic dynamic load rating the rolling elements stop rolling and start to skid, which wipes through the lubricant film and smears the surfaces. Smearing kills a bearing much faster than fatigue does. The fix is more load, not less: a preload spring, a deliberate axial load on an angular contact pair, or a smaller bearing. The published minimum-load rules disagree by a factor of three, so use the bearing maker’s own formula for your type.

How do I turn a target life back into a bearing size?

Invert the equation: C = P·(L·60n/10⁶)^(1/p), which this page computes as “rating a catalogue must show”. That is the number you search a catalogue on. The companion figure, “load this bearing would carry”, is the same equation solved the other way and is what you need if the bearing is fixed and the layout is not. Both assume you have the equivalent load P right, which for anything with an axial component means going through the equivalent load page first.

Related calculators

References

ISO 281:2007, Rolling bearings — Dynamic load ratings and rating life. Cited by number, not reproduced. What the freely published preview does confirm, in its own introduction, is the thing that matters most here: “the life modification factors for reliability, a₁, have been slightly changed and extended to 99,95 % reliability” — so the a₁ column differs between the 1990 and 2007 editions, and a figure taken from one edition and used with the other is wrong. Both columns are printed on this page and both are attributed. The preview stops before clause 9, so the a₁ table itself was taken from manufacturers’ catalogues, below.
NTN, Ball and Roller Bearings, catalogue 2203-E, technical section 3, “Load rating and life” (pages A-20 and A-21). The source for L₁₀ = (C/P)^p with p = 3 for ball bearings and 10/3 for roller bearings, for L_10h = 10⁶/(60n) × L₁₀, and for the ISO 281:2007 a₁ column (1.00, 0.64, 0.25, 0.093 at 90, 95, 99 and 99.9 %).
NSK, Bearing life calculation tool — contents of HELP in the new life calculation equation. The source for L_nm = a₁ a_ISO L₁₀ and for the ISO 281:2007 a₁ column with its L_5m to L_1m designations.
NTN-SNR, Bearing Wizard — Lifetime calculation. A live manufacturer’s calculation tool, and the source for the ISO 281:1990 a₁ column (0.62, 0.53, 0.44, 0.33, 0.21 at 95 to 99 %), which it still publishes. It is also the source for what a_ISO actually depends on — a_ISO = f(e_CC_u/P, κ) — and for its two hard limits: a_ISO ≤ 50, and κ capped at 4 however good the film is. Those two limits are the reason this site takes a_ISO as an input rather than inventing a curve for it.
AST Bearings, Radial Ball Bearing Life & Load Ratings, and its angular contact companion note. The second independent printing of the ISO 281:1990 a₁ column. NOTE: the summarised fetch of this table returned twelve reliability headings against twelve values in which 0.64 sat under 94 % and 0.62 under 95 % — a shift of one, because 0.64 belongs to the 2007 column and not to this one at all. The rest of the row is the 1990 column exactly.
American Roller Bearing, Bearing Life Calculation — Bearing Loads & Speeds. The source for the minimum load rule P_min = 0.02C and for a second, independent printing of the ISO 281:2007 a₁ column. NOTE: its own life equation is printed as L₁₀ = (C/P)^e × 10⁶/N, which is a life in MINUTES, not hours — the factor of 60 is missing. Recorded because it is the commonest slip on this calculation and it is a factor of sixty.
ISO 15243:2017, Rolling bearings — Damage and failures — Terms, characteristics and causes. Cited by number. It is the document that gives the failure modes this batch keeps naming — surface-initiated and subsurface fatigue, smearing, fretting corrosion, false brinelling — their agreed names, which matters because a defect frequency identifies a LOCATION and not a cause.
Tribonet, Rolling bearing lubrication — ISO 281 kappa factor in practice, and Interlub’s κ-value note. The sources for κ = ν/ν₁, for the practical target band of 1 to 4, and — importantly for what this site will not do — for the fact that ν₁ comes from a manufacturer’s chart: “use the bearing manufacturer’s tools, ISO 281 guidance or validated engineering references to estimate the reference viscosity”.