Radial, Thrust and Combined Loads

Why this matters

A bearing catalogue gives one big number, the dynamic load rating, and that number is quoted for one direction only. Push on a bearing in a direction its geometry was not built to resist and the rating stops describing it. The failures that come from this do not look like overload: they look like a bearing that ran quietly for a year and then came apart with damage on a raceway shoulder nobody expected to be in contact. This article is mostly about what each bearing type will refuse to do, because that is the part the rating never tells you.

Before you go looking for the thrust

Isolate and lock out under 29 CFR 1910.147 before opening a guard, and remember the wheel coasts after the contactor drops: wait for zero rotation before reaching in, and restrain a spring-loaded motor base before pulling belts, because the base is stored energy under the same standard. Guarding on shafts, sheaves and couplings goes back before the machine runs, per 29 CFR 1910.219 in general industry or 29 CFR 1926.300(b) on a construction site. Axial thrust from air or fluid pressure across a wheel or impeller exists only while the machine runs, so it is something you infer from geometry and pressures, never something you go feel for by hand on a turning shaft. If you open a terminal box to read current as part of the same trip, that is electrical work under 29 CFR 1910.333(b)(2), proved dead by the live-dead-live sequence in NFPA 70E-2021, 120.5.

The three directions, and what physically resists each

Radial load acts perpendicular to the shaft axis. It is resisted by the rolling element being squeezed between two raceways whose curvature faces it, which every bearing type has.

Axial load, also called thrust, acts along the shaft axis. It is resisted only if there is raceway geometry angled to catch it. A ball sitting in a deep groove climbs the groove wall and generates an angled contact, so it can take thrust. A cylindrical roller sitting on two flat raceways has nothing angled to catch it, so it cannot.

Moment load is a couple trying to tilt the ring planes relative to each other. A single row of rolling elements resists this poorly because it has one contact plane. Two rows spaced apart resist it well, which is why a moment load is really a pair of opposed axial loads once you draw it.

The distinction is not about strength. It is about whether the contact geometry exists at all.

What each type will not do

This table is the useful half of a bearing catalogue.

The tapered roller entry is the one that catches experienced people. Apply a pure radial load to a single-row tapered roller bearing and the geometry produces an induced axial force of roughly 0.5 x Fr / Y, where Y is the axial load factor for that specific bearing from its own catalogue page. That force has to be reacted by the opposing bearing in the arrangement, which means you cannot analyse one tapered roller bearing in isolation. The relationship is derived for the standard single-row tapered geometry with the load spread normally across the rollers, and it is one of the few places where a bearing manufactures a load rather than just carrying one.

Converting a combined load into something you can compare

When both radial and axial load are present, you cannot compare either one against the rating. You convert them into an equivalent dynamic load, P, which is the pure radial load that would do the same damage:

P = X x Fr + Y x Fa

X and Y are read from the bearing's own catalogue page. They are not universal constants and they are not derived in the field. They depend on the bearing type, on its nominal contact angle, and for a deep-groove ball bearing on the ratio of axial load to the static rating C0. There is also a threshold, e, below which the axial load is small enough that the bearing behaves as if purely radially loaded, and above which the full X and Y apply.

Never carry an X or Y value from one bearing series to another. A 15-degree angular contact bearing and a 40-degree one have different Y factors for exactly the reason the table above exists: the contact angle is the geometry doing the work.

The worked case: a fan shaft with a duct-side pressure

A belt-driven fan shaft, drive-end bearing, with 400 lbf of radial load from belt pull and wheel weight worked out by statics. The wheel is a single-inlet type, so there is a static pressure difference across it and an axial force pushing the wheel toward the inlet. Call the resulting thrust at this bearing 150 lbf, taken from the fan manufacturer's thrust figure for this wheel at this pressure rise rather than estimated.

Ratio Fa / Fr is 150 / 400, which is 0.375.

The catalogue page for this bearing series, at this axial load relative to its C0, gives e = 0.30, X = 0.56 and Y = 1.6. Because 0.375 exceeds e = 0.30, the combined case applies rather than the radial-only case.

P = 0.56 x 400 + 1.6 x 150 P = 224 + 240 P = 464 lbf

Now compare that to the mistake. A tech who sizes on radial load alone works with 400 lbf. Using the ball bearing load-life exponent of 3, the life ratio between the two is (400 / 464) cubed, which is 0.64. The bearing will reach about 64 percent of the life the radial-only calculation predicted, which means that calculation overstated its answer by more than half. Notice which term did the damage: the axial 150 lbf, only 38 percent as large as the radial load, contributed 240 lbf to P, slightly more than the radial term's 224 lbf, because the Y factor is larger than the X factor. In a combined load the smaller force is frequently the bigger contributor.

What flips this result: if the fan runs mostly against a much lower pressure rise, Fa falls, the ratio drops under e, and the radial-only calculation becomes correct. That is why a fan sized for a dirty-filter condition and operated clean does not behave like the calculation, and why a damper closing on a running fan changes the bearing's duty rather than just its amp draw.

What P deliberately leaves out

This is the section worth remembering. The equivalent dynamic load carries no information about any of the following, and every one of them affects life:

• Lubrication adequacy. The film condition lives in a separate life-modification factor, not in P.
• Contamination. Solid particles in the lubricant are handled by a contamination factor applied outside P.
• Temperature. Raise the operating temperature and the lubricant thins; P is unchanged.
• Shock and vibration. A drive that hammers is applying a peak load P never saw, and the peak is what damages the raceway.
• Misalignment. A shaft slope that the bearing type cannot tolerate concentrates the contact at one end of a roller. The rating assumed the contact was uniform.

A reader who computes P carefully and then treats it as the whole story has done the arithmetic that is easiest and skipped the four factors that decide most real service outcomes. P tells you whether the bearing is the right size for the forces. It does not tell you whether the machine will look after it.

Who in the arrangement takes the thrust

Almost every two-bearing machine has one bearing that locates the shaft axially and one that floats. The locating bearing takes all the thrust and its share of radial load. The floating bearing takes radial load only and must be free to move axially in its housing as the shaft grows with temperature.

Two failures come out of getting this wrong, and they look nothing alike. Locate at both ends and the shaft's thermal growth has nowhere to go, so it turns into an axial load that rises as the machine warms and peaks at operating temperature. Locate at neither end and the shaft hunts axially, which loads a deep-groove ball bearing's shoulders in both directions and rubs the seals. If a machine's bearing failures start after a rebuild, check which end was made the locating end against how the manufacturer drew it.

How to verify you got this right

Check the ratio against e before you use X and Y at all. If Fa / Fr is below e, the combined factors do not apply and using them overstates P. This is the single most common arithmetic error in the calculation, and it goes in the conservative direction, which is why it survives.

Check that Y came from the same catalogue page as the bearing. If you can find your Y value in an article rather than in a catalogue, it is a typical value for some contact angle, not the value for yours.

Check the load directions physically. Fa should have a source you can name and point at: a pressure difference across a wheel, a helical gear's separating force, a thermal growth reacting against a locating shoulder, or an induced force from a tapered arrangement. If you cannot name the source, you have either missed a thrust that is really there or invented one that is not, and both change P.

References

• ISO 281 for the equivalent dynamic load formulation and the load-life exponents; the X, Y and e values themselves are bearing-specific and come from the manufacturer's own catalogue page
• ISO 76 for the basic static load rating C0, which is the reference the axial-load ratio is taken against for deep-groove ball bearings
• 29 CFR 1910.147 for hazardous-energy control and stored energy in a tensioned drive; 29 CFR 1910.219 and 29 CFR 1926.300(b) for power-transmission guarding in general industry and construction respectively
• Fan or pump manufacturer documentation for the axial thrust produced at a given pressure rise or head, which is not estimable from the bearing side
• See related: What a Bearing Load Actually Is; What Preload Does in a Bearing Arrangement