Posts

Bearing Type Selection: A Decision Matrix with a Worked Example

Bearing Type Selection: A Decision Matrix with a Worked Example

Most “how to choose a bearing” pages hand you a comparison table and stop there. A table tells you
what a deep groove ball bearing is; it does not tell you whether that bearing survives
20,000 hours at 720 rpm under 8 kN of radial load, 2.5 kN of axial load, and 0.4° of shaft
misalignment. That second question is the one that actually costs money when it is answered wrong.

This guide does both. First, a decision matrix that narrows ten-plus bearing
families down to a shortlist using five measurable inputs. Then a fully worked selection
example
: we compute the equivalent dynamic load P, the ISO 281 basic rating life
L10h, and the required dynamic load rating C, and compare three
candidate arrangements numerically. You get the arithmetic, not another checklist.

About the numbers: all load ratings, factors and tolerance angles
in the worked example are illustrative magnitudes for teaching the method, not catalogue
values for any specific manufacturer’s product. Always substitute the ratings from the catalogue of
the bearing you intend to buy.

1. Why a comparison table is not a selection method

A table is a lookup device: you already know the answer, and you want to confirm its properties.
Selection runs in the opposite direction — you have a load case and need to arrive at the answer.
Three failure modes show up repeatedly when people skip the arithmetic:

  • Capacity is not a single number. “Roller bearings are stronger” is true for
    radial load and misleading for speed, friction, and misalignment. A spherical roller bearing
    that thrives in a crusher is a poor choice for a high-speed spindle.
  • Life is dominated by one exponent. Because L10 scales with
    (C/P)p, and p is 3 for ball bearings but 10/3 for roller bearings,
    a modest change in equivalent load or rated capacity moves life by multiples — not percentages.
    Guessing here is expensive.
  • The binding constraint is often not load. Misalignment, speed, envelope,
    minimum load, and lubrication frequently eliminate a bearing that passes the life calculation
    comfortably. In our worked example below, the arrangement with the longest calculated
    life is rejected for exactly this reason.

The fix is a sequence: define inputs → eliminate by hard constraints → compute
P and life for the survivors → check the secondary constraints → only then pick a
part number.

Single row deep groove ball bearing
The deep groove ball bearing is the default answer for moderate loads and high speed — but “default” is a starting hypothesis, not a conclusion.

2. The five inputs that decide the type

Before looking at any catalogue, fix these five. Everything downstream is a consequence of them.

InputWhat to establishWhy it decides the type
Load direction & ratioRadial Fr, axial Fa, combined, shock content, and the ratio Fa/FrSeparates pure-radial types (cylindrical, needle) from combined-load types (deep groove, angular contact, tapered, spherical) and pure-axial types (thrust)
SpeedContinuous and peak rpm; n·dm where dm is the mean diameterBall bearings tolerate higher speed and run cooler; roller types generate more heat and have lower limiting speeds
MisalignmentExpected angular error between shaft and housing, in degreesRigid types (tapered, cylindrical, angular contact) tolerate only a few arcminutes; self-aligning types tolerate ~1–2°
EnvelopeBore (set by the shaft), max outside diameter, max widthWhen radial space is tight, needle rollers or thin-section bearings win; when one bearing must do everything, a compact combined-load type wins
Life, temperature & environmentTarget L10h, operating temperature, dust/water, grease vs oil, sealingDrives required C, clearance class (C3/C4), seal type, and lubricant choice

Note that target life sits in this list, not in the calculation section. Life is a design
input you choose; the calculation then tells you what rated capacity is needed to deliver it.

3. The bearing type decision matrix

Use this as the first-pass filter. Load direction gets you to a block; speed and misalignment cut
it down; envelope and cost pick the winner. Ratings are qualitative and typical — the worked
example in sections 5–8 shows how to replace “high” and “moderate” with numbers.

Bearing typeRadialAxialSpeedMisalignmentStiffnessRelative costTypical fit
Deep groove ballModerateModerate, both directionsVery highLow (~0.15°)Moderate$Electric motors, pumps, gearboxes
Angular contact ball (paired)ModerateHigh, one direction per bearingVery highVery lowHigh with preload$$Spindles, pumps, precision shafts
Self-aligning ballLight–moderateLightHighHigh (~2°)Low$Long shafts, agricultural, conveyors
Four-point contact ballModerateHigh, both directionsMediumLowModerate$$Compact combined load, turntables
Cylindrical roller (NU/N/NJ)Very highNone (or limited, by design)HighVery lowHigh$$Gearboxes, motors, rolling mills
Tapered roller (paired)HighHigh, one direction per bearingMediumVery low (~0.05°)Very high$$Wheel hubs, gearboxes, rolling mills
Spherical rollerVery highModerate, both directionsLow–mediumHigh (~1.5°)High$$$Crushers, fans, paper machines, mining
Needle rollerHigh per section heightNoneMediumVery lowModerate$Transmissions, compact radial space
Thrust ball / thrust rollerNoneVery highLowLowHigh axially$$Vertical shafts, hooks, screw presses
Slewing / crossed rollerHighHigh + tilting momentLowLow–moderateVery high$$$$Cranes, excavators, robotics, radar

Two families sit in the “ball” block, two in “rigid roller”, one in “self-aligning”, and one at
the top for moment loads. Their mechanics are worth a picture each.

Angular contact ball bearing
Angular contact ball bearing: contact angle carries axial load in one direction, so these are almost always mounted as a matched pair.
Double row self-aligning ball bearing
Self-aligning ball bearing: two rows in a spherical outer-ring raceway tolerate shaft deflection at the cost of load capacity.

4. Worked example — the load case

A geared drive shaft on a welded frame. The shaft diameter is fixed at 70 mm by the gear bore, the
housing allows a maximum outside diameter of 150 mm and a width of 40 mm, and the frame is known to
deflect. The specified duty is:

ParameterSymbolValue
Radial loadFr8.0 kN
Axial load (one direction, from a helical gear)Fa2.5 kN
Rotational speedn720 rpm
Application / load factor (moderate shock)fd1.3
Target basic rating lifeL10h20,000 h
Expected misalignment—0.4° (welded frame, long shaft)
Operating temperature—75 °C, grease lubricated
Envelope—bore 70 mm, OD ≤ 150 mm, width ≤ 40 mm

Load ratio: Fa/Fr = 2.5 / 8.0 = 0.313. Combined loading, radial dominated,
with a non-trivial axial component. We now walk the sequence.

5. Step 1 — eliminate by load direction and envelope

Fa/Fr = 0.313 means the bearing must carry both components. That immediately rules out:

  • Cylindrical roller (NU/N) and needle roller — high radial capacity, but no
    axial capacity in the standard design. They would need a separate thrust bearing, adding width,
    parts, and a lubrication split. With a 40 mm width budget and a single housing, that is
    unattractive here.
  • Pure thrust ball / thrust roller — they cannot take the 8 kN radial load at
    all, and they also need a minimum axial load to avoid skidding.
  • Slewing and crossed roller rings — technically capable of combined load plus
    tilting moment, but sized and priced for low-speed rotational structures, not a 720 rpm shaft.
    See our rolling bearing vs slewing bearing comparison for
    where that boundary sits, and our slewing bearing selection overview
    if the application is a rotating structure rather than a shaft.

Survivors: deep groove ball, angular contact ball (paired), tapered roller (paired),
spherical roller, four-point contact ball
. Five candidates, all available in a 70 mm bore.

NU design cylindrical roller bearing
Cylindrical roller (NU) designs are excellent for pure radial load and allow axial shaft movement — but here the 2.5 kN axial load rules them out as a standalone solution.

6. Step 2 — eliminate by misalignment before you calculate life

This is the step most selection guides skip, and the one that decides this example. The welded
frame gives 0.4° of angular error. Typical tolerable misalignment (illustrative order-of-magnitude
values, confirm against the catalogue):

CandidateTolerable misalignment (typical)0.4° case
Tapered roller pair~0.05°Fail — edge loading, rapid fatigue
Angular contact pair~0.03°Fail — preload-sensitive
Four-point contact ball~0.1°Fail
Deep groove ball~0.15°Marginal / fail at this magnitude
Spherical roller~1.5°Pass with margin

So on the frame as built, the spherical roller bearing is the only survivor that comfortably
accepts the misalignment. But we should still do the life arithmetic for all three representative
options — because it quantifies how much we are giving up, and it tells us exactly what
changing the frame would buy us. That trade is the useful engineering output.

Single row tapered roller bearing
Tapered roller bearings give outstanding combined-load capacity and stiffness, but only when the shaft and housing stay aligned to a few arcminutes.
Spherical roller bearing
Spherical roller bearings self-align inside the outer ring, which is why they dominate crushers, fans and long-shaft installations.

7. Step 3 — equivalent dynamic load P

ISO 281 reduces the combined load to a single equivalent dynamic load:

P = fd · (X·Fr + Y·Fa)

The factors X and Y depend on the bearing type and on whether Fa/Fr
exceeds a limiting value e; for deep groove ball bearings, e and Y are
themselves functions of Fa/C0. The table below uses illustrative factors to keep
the method transparent.

Candidate (illustrative)C kNC0 kNeFa/Fr vs eXYP kN
A) Deep groove ball 6314104680.240.313 > e0.561.811.67
B) Tapered roller 30214 pair170 (pair)—0.370.313 ≤ e1.0010.40
C) Spherical roller 22214178—0.230.313 > e0.672.916.39

Worked longhand for each:

  • A: Fa/C0 = 2.5/68 = 0.037 → illustrative e = 0.24,
    Y = 1.8. Since 0.313 > 0.24, use X = 0.56:
    P = 1.3 × (0.56×8.0 + 1.8×2.5) = 1.3 × 8.98 = 11.67 kN.
  • B: for an opposed pair, 0.313 ≤ e = 0.37, so the axial term drops
    out and the pair behaves as X = 1, Y = 0:
    P = 1.3 × 8.0 = 10.40 kN. (The pair rating C is taken as
    1.62 × the single-bearing rating, an illustrative pairing factor.)
  • C: 0.313 > e = 0.23, so X = 0.67, Y = 2.9:
    P = 1.3 × (0.67×8.0 + 2.9×2.5) = 1.3 × 12.61 = 16.39 kN.

Notice the counter-intuitive result: the spherical roller bearing has by far the highest
C yet the highest P, because its Y factor penalises axial load heavily.
This is precisely why “just pick the biggest C” is not a method.

8. Step 4 — L10 life and the required dynamic load rating

ISO 281 basic rating life, in million revolutions, is L10 = (C/P)p,
with p = 3 for ball bearings and p = 10/3 for roller bearings. Converting to hours
at speed n (rpm):

L10h = (106 / 60n) · (C/P)p

At 720 rpm, the factor 106/(60×720) = 23.15 hours per million
revolutions
. Rearranging for the capacity needed to hit a target life gives
Creq = P · (L10h / 23.15)1/p. For a 20,000 h
target, (20,000 / 23.15) = 864 million revolutions, so the required ratio C/P is
8641/3 ≈ 9.52 for ball bearings and 8640.3 ≈ 7.60 for roller bearings.

CandidateC kNP kNpC/PL10h (h)Creq for 20,000 hVerdict
A) Deep groove ball 631410411.6738.9116,367111.2 kNLife ~18% short
B) Tapered roller 30214 pair17010.4010/316.35257,09179.1 kNHuge margin
C) Spherical roller 2221417816.3910/310.8665,623124.6 kNPasses with margin

Three useful conclusions fall out of this table:

  • A fails, but narrowly. It needs C = 111.2 kN and has 104 kN. Stepping
    up to a heavier 64-series bearing (illustrative C = 143 kN) gives ~42,500 h — but its 180 mm
    outside diameter breaks the 150 mm envelope. The life problem is really an envelope problem.
  • B is massively over-capacity — 257,000 h against a 20,000 h target. That
    margin is worthless here because step 2 already rejected it on misalignment.
  • C clears the target by 3× while also being the only type that tolerates
    0.4°. Its higher P is more than offset by the roller exponent p = 10/3.

These are illustrative magnitudes, not catalogue values. The factors e,
X, Y, the pairing factor 1.62, and all C/C0 figures
must be taken from the actual manufacturer’s catalogue before any purchasing decision.

9. Step 5 — compare the three solutions, not just the three bearings

A bearing selection is really an arrangement selection. The honest comparison includes
what you would have to change elsewhere in the machine:

SolutionL10hHandles 0.4°?Fits envelope?What else changesRelative cost
A — Deep groove ball (64-series)~42,500 hNo, marginalNo (OD 180 > 150)Redesign housing; still need to fix alignment$
B — Tapered roller pair + frame rework257,000 hOnly if frame is machined to ≤0.05°Yes (OD 125)Machine/weld-jig the frame, add preload/endplay setting procedure and shims$$ + machining
C — Spherical roller 2221465,600 hYes (up to ~1.5°)Yes (OD 125)Nothing; standard fit, standard grease$$$

Recommendation for the frame as built: solution C. It is the only option that
satisfies life, misalignment and envelope simultaneously without redesigning the machine. Its higher
unit price is cheaper than machining a welded frame and adding a shimming procedure.

But if the frame can be corrected to ≤0.05° (line-bored housing, machined
pedestal, or a shorter shaft), solution B becomes very attractive: it fits the same 125 mm OD, is
stiffer, costs less per bearing, and its enormous life margin means you could downsize the bearing
series further. This is the trade worth quoting both ways — and it only becomes visible once you
run the numbers. Our tapered roller bearing selection guide covers the
preload and endplay setting that a paired arrangement then requires.

For a deeper look at the two roller families in this comparison, see our
spherical and cylindrical roller bearing notes; for the high-volume
rolling-mill case where four-row tapered bearings are the norm, see
rolling mill bearing selection.

Four row tapered roller bearing
Four-row tapered roller bearings are the rolling-mill default: enormous radial capacity with axial load handled by the paired geometry.

10. The checks that come after life

Passing the life calculation is necessary, not sufficient. Run these before issuing a purchase
spec:

CheckWhat to verifyThis example
Speedn·dm against the catalogue limiting speed for the lubricant and cagedm = (70+125)/2 = 97.5 mm; n·dm = 70,200 mm·rpm — comfortable for grease
Minimum loadRoller and thrust types need a minimum load to prevent roller/ball skidding8 kN radial is well above typical minimums for this size
Operating clearanceInterference fit + inner/outer ring temperature difference consume clearance; may push you from CN to C3/C475 °C with an interference inner-ring fit → check whether C3 is needed; see bearing clearance guidance
LubricationGrease vs oil, base-oil viscosity at operating temperature, relubrication interval, EP additives for shock loadsGrease, EP2-class, interval set by 75 °C and contamination; see lubricant selection notes
Sealing & contaminationOpen, shielded (ZZ) or sealed (2RS); dust/water ingressWelded-frame industrial environment → sealed or well-protected housing
Fits & mountingShaft/housing tolerances, shoulder squareness, mounting method (never hammer through rolling elements)Verify ISO 286 fits and shoulder perpendicularity before assembly

11. Five mistakes the arithmetic catches

  • Selecting by bore diameter alone. Bore is set by the shaft; capacity, speed and
    misalignment decide the type. A 6314 and a 22214 share a 70 mm bore and behave nothing alike.
  • Ignoring the axial component because it is “small”. At Fa/Fr = 0.313
    it crossed the e threshold and changed both X and Y, raising
    P by up to ~58% depending on type (10.40 kN for the tapered pair vs 16.39 kN for
    the spherical roller). That is a life change of several multiples.
  • Assuming the highest C wins. Candidate C had the highest C and the
    highest P; only the exponent made it competitive.
  • Sizing life before fixing alignment. Candidate B computed 257,000 h and was
    still the wrong bearing for this frame. Hard constraints outrank life.
  • Treating a single bearing as the whole arrangement. Axial load in one
    direction, thermal expansion, and shaft locating/floating duties often require a matched pair
    or a dedicated thrust bearing. See angular contact vs deep
    groove ball bearings
    and where thrust bearings are used.
Thrust needle roller bearing
Needle and thrust rollers solve space- and axial-dominated problems — but each needs the right load direction, and thrust types need a minimum load.

12. Quick application → type map

Application profileFirst type to evaluateWatch out for
High speed, moderate radial + some axial, cleanDeep groove ballCheck Fa/Fr against e; C3 if hot
Combined load, high stiffness, precise axial locationPaired angular contact or paired tapered rollerPreload/endplay setting; alignment
Heavy radial, shock, shaft deflectionSpherical rollerSpeed limit, minimum load, cost
Very heavy radial, no axial, high speed, axial float neededCylindrical roller (NU/N)Needs a separate axial locating bearing
Overwhelmingly axial, vertical shaftThrust ball / thrust roller / spherical roller thrustNo radial capacity; minimum load
Very limited radial spaceNeedle roller or thin-sectionShaft hardness, alignment sensitivity
Combined load + tilting moment, slow rotationFour-point contact / crossed roller slewingBolt pattern, mounting-face flatness
Rolling mill / very heavy radial, moderate speedFour-row tapered or four-row cylindricalLubrication, sealing, mounting sequence

Browse our deep groove ball bearings,
tapered roller bearings,
spherical roller bearings,
cylindrical roller bearings,
thrust bearings,
thin-section bearings and
slewing bearings to see the ranges available for each family
in the matrix.

Open thin section bearing
When radial space is the binding constraint, thin-section and needle designs deliver capacity in a very small cross-section.

Send us the load case, get a sized comparison

The method above is exactly what our engineers run when a customer sends an application. Give us
Fr, Fa, speed and duty cycle, target life, misalignment, temperature, and your
envelope limit, and we will compute P and L10h for the viable types,
then quote the shortlist side by side — including what would have to change in the surrounding
structure. Contact our engineering team with your operating data.

Related reading

Crossed roller slewing bearing
For combined load plus tilting moment at low speed, crossed roller slewing rings take over from conventional shaft bearings.

Frequently Asked Questions

What is the first thing to determine when choosing a bearing type?

The direction and ratio of the load. Compute Fa/Fr: pure radial points to cylindrical or needle rollers, combined load to deep groove, angular contact, tapered or spherical types, and predominantly axial load to thrust bearings. Then apply speed, misalignment, envelope and life as successive filters.

Are roller bearings always stronger than ball bearings?

No. Roller bearings carry more radial load for the same envelope because line contact spreads the load, but ball bearings run faster, cooler, quieter and cheaper. Strength is conditional on load direction, speed and alignment, so a spherical roller bearing is a poor substitute on a high-speed precision spindle.

What is equivalent dynamic load P and why does it matter?

P is the single radial load that would produce the same fatigue life as the actual combined radial and axial loading. It is computed as fd*(X*Fr + Y*Fa), where X and Y come from the bearing type and the Fa/Fr ratio relative to the limit e. Because life scales with (C/P)^p, a small error in P becomes a large error in predicted life.

Why do ball and roller bearings use different life exponents?

ISO 281 uses p = 3 for ball bearings (point contact) and p = 10/3 for roller bearings (line contact), reflecting how contact stress and fatigue life scale with load. The higher roller exponent means roller life is more sensitive to the C/P ratio.

What is the C3 clearance and when do I need it?

C3 denotes greater-than-normal internal radial clearance. Interference fits and a hotter inner ring consume clearance during operation, so C3 (or C4) is used for electric motors, hot applications, heavy interference fits and many roller bearings. Choosing the wrong class can cause thermal runaway and seizure even when the size is correct.

Can I replace a deep groove ball bearing with a spherical roller bearing of the same bore?

Sometimes, but not as a blind swap. The spherical roller bearing handles misalignment and much higher radial load, but it has a lower speed limit, higher friction, a different axial-load factor Y, and usually a different width. Re-run the P and L10h calculation and check the housing fits and speed before substituting.

How much misalignment can each bearing type tolerate?

As an order of magnitude: tapered roller and angular contact pairs only a few arcminutes (~0.05 deg or less), deep groove ball around 0.15 deg, four-point contact around 0.1 deg, self-aligning ball up to ~2 deg, and spherical roller up to ~1.5 deg. Confirm exact values in the manufacturer's catalogue.

What should I send a supplier to get a correct bearing recommendation?

Radial and axial loads (including shock), speed and duty cycle, target L10h life, expected misalignment, operating temperature and lubricant, sealing needs, and the bore/OD/width envelope. With those inputs an engineer can compute P, required C and the shortlist of viable types instead of guessing from a part number.