Roller Bearing

Why a line beats a point

Every rolling-element bearing solves the same problem: keep two pieces of metal that must rotate relative to each other from rubbing along their entire mating surface. A plain journal bearing supports the load over a wide arc and asks lubricant film to do all the work; a rolling-element bearing interposes hardened elements that contact each race over a tiny footprint. That footprint is where the bearing lives or dies. In a ball bearing the footprint is an ellipse a fraction of a millimetre across — an almost-point. In a roller bearing the footprint is a long, thin rectangle running along the length of the cylinder — a line.

Hertz contact theory makes the difference quantitative. For a sphere pressed against a curved surface, the contact patch is an ellipse whose semi-axes grow as the cube root of load, so peak pressure rises as P^(1/3). For a cylinder pressed against a curved surface, the patch is a rectangle whose width grows as the square root of load, so peak pressure rises as P^(1/2). Equate the two for the same applied force and the cylinder always wins on peak pressure — by roughly a factor of two for typical bearing geometries. Subsurface fatigue is driven by the maximum orthogonal shear stress, which lies a few tenths of a millimetre below the contact surface at depths proportional to patch dimension. Lower peak pressure means lower subsurface stress, means later crack nucleation, means longer life or higher allowable load.

The catalogue consequence is stark: for the same bore and outside diameter, a cylindrical roller bearing typically rates two to three times the dynamic capacity C of an equivalent ball bearing. That is why heavy machinery — railway axleboxes, rolling-mill rolls, mining-truck wheel hubs, gas-turbine compressor shafts — is built almost universally on rollers.

Anatomy of a roller bearing

A roller bearing consists of four parts:

• Inner race. A hardened steel ring pressed onto the rotating shaft. The outer surface is finely ground to a precise cylindrical or tapered raceway.
• Outer race. A hardened steel ring pressed into the housing. Its inner surface is the second raceway. Together with the inner race, it defines the rolling envelope.
• Rolling elements. Cylindrical, tapered, spherical, or needle-shaped rollers — typically through-hardened bearing steel (52100, EN31) ground to micron tolerances. They are the only parts that contact both races, and they roll, not slide.
• Cage (retainer). A pressed-steel, brass, or polymer ring that keeps the rollers evenly spaced around the bearing circumference. Without a cage, rollers would bunch up under load and one part of the bearing would carry everything.

A roller bearing is therefore a five-piece assembly: shaft, inner race, rollers, cage, outer race, housing. Two of those pieces (inner race, outer race) rotate relative to each other; the rollers and cage rotate at intermediate speed, with the cage's angular velocity given approximately by the kinematic relation ω_cage = ω_inner × r_inner / (r_inner + r_outer). For an outer-fixed bearing with similar race radii the cage runs at roughly half shaft speed.

Four roller geometries, four jobs

The family splits along the roller shape. Each variant trades capacity, alignment tolerance, and load direction in a different way.

Type	Roller shape	Radial load	Axial load	Misalignment	Typical use
Cylindrical	Straight cylinder	Very high	None (or trivial)	Low (~ 0.04°)	Electric motors, gearboxes, rolling-mill rolls
Tapered	Truncated cone	High	Moderate to high	Low	Car wheel hubs, differentials, pinion shafts
Spherical	Barrel	Very high	Moderate	High (1–2°)	Mining screens, wind-turbine main shafts, paper mills
Needle	Slender cylinder (L/D 3–10)	High per envelope	None	Very low	Connecting rods, planet pinions, universal joints
Toroidal (CARB)	Long barrel	Very high	None	High + axial drift	Vibrating screens, paper-machine rolls (non-locating side)
Crossed-roller	Cylinders alternating ±45°	High	High (both directions)	Low	Robot joints, machine-tool turntables, optical mounts

The cylindrical roller is the workhorse for pure radial duty. The flanges on the inner or outer race control axial location but should never be asked to take significant thrust — the line contact is normal to the axis. Tapered rollers convert combined radial and axial loads into pure rolling because the cones, cups, and roller centre-lines all converge at a single apex on the bearing axis. Spherical rollers run on a spherical outer raceway, so the inner ring can tilt without losing contact uniformity. Needle rollers pack a high-capacity bearing into a thin annulus — invaluable inside connecting-rod small ends or planetary gearsets, where envelope is precious. Toroidal (CARB-style) bearings combine spherical self-alignment with the freedom to accommodate axial displacement of the shaft due to thermal growth. Crossed-roller bearings alternate roller orientation so a single ring takes load in any direction, which is why precision robot wrists and optical mounts depend on them.

The life equation

The catalogue rating for any rolling-element bearing is its basic dynamic load rating C — by ISO definition, the constant radial load that gives a basic rating life of one million revolutions for 90 percent of a population of identical bearings. The relation between actual life and applied load is

L₁₀ = (C / P)^p × 10⁶  revolutions

where P is the equivalent dynamic load (a combination of radial and axial components weighted by the bearing's geometry factors) and p is the load-life exponent: 3 for point-contact ball bearings, 10/3 ≈ 3.33 for line-contact roller bearings. The exponent comes from Lundberg-Palmgren fatigue theory: the volume of stressed material under a line contact scales differently with load than under a point contact, and the integral over the resulting stress distribution gives the empirically validated 10/3.

To convert revolutions to hours, divide by the shaft speed n (rev/min) and multiply by 60:

L₁₀ₕ = (C / P)^(10/3) × 10⁶ / (60 × n)  hours

Modern ISO 281:2007 augments this with a life-modification factor a_ISO that accounts for lubricant viscosity, contamination, and the fatigue load limit C_u; the modified life is L_nm = a₁ × a_ISO × L₁₀, where a₁ adjusts for reliability levels other than 90 percent (e.g. a₁ = 0.21 for L₁ life at 99 percent survival).

Worked example: gearbox shaft on cylindrical rollers

Consider an industrial gearbox shaft running at n = 1500 rpm under a steady radial load of P = 25 kN. The designer selects a cylindrical roller bearing (NU 314 ECP) with C = 178 kN, C_u = 24 kN. The basic life is

L₁₀ = (178 / 25)^(10/3) × 10⁶
    = 7.12^3.333 × 10⁶
    ≈ 666 × 10⁶ revolutions

L₁₀ₕ = 666 × 10⁶ / (60 × 1500)
     ≈ 7 400 hours

About 7 400 hours of operation before 10 percent of bearings will have failed by spalling — roughly one year of continuous duty. If we now double the load to P = 50 kN, the life falls by 2^(10/3) ≈ 10, to about 740 hours — a single month. Compare a ball bearing at the same C and P with p = 3: doubling load shortens life by 2³ = 8. Roller bearings are slightly more punitive of overload than balls, a fact that catches inexperienced designers out.

Tapered rollers and the apex condition

Tapered roller bearings deserve a closer look because they are the dominant geometry in automotive wheel hubs, differentials, and any application where radial and axial loads coexist. The defining geometric requirement is that the axes of the inner raceway (the cone), the outer raceway (the cup), and every roller all meet at a single point on the bearing axis. When that apex condition is satisfied, the rollers can roll without sliding along their entire length — the linear velocity at every point along the contact line is consistent with pure rolling.

If the apex condition is violated — by, say, an outer race manufactured to the wrong half-angle — the roller is forced into a mixed rolling/sliding regime, contact patches heat up, and the bearing fails prematurely. This is one of the reasons tapered bearings are produced in matched sets (cone, cup, and rollers ground together) and why an installer must never mix cones from one bearing with the cup from another even of the same designation.

The classic automotive application is the SKF 32213 (and equivalents from Timken, Koyo, NTN, NSK): a 65 mm bore, 120 mm OD tapered roller with a 21.5 mm assembled height, used in opposed pairs on the front wheels of light trucks. One bearing takes radial weight plus cornering thrust outward; the other takes weight plus cornering thrust inward. Pre-load is set by a torqued spindle nut so neither bearing is loose at any wheel angle. The same family scales up to the multi-tonne tapered rollers in mining-haul truck wheel ends and down to the small tapered rollers in pinion shafts of car differentials.

How roller bearings fail

Subsurface fatigue is the canonical end-of-life mode, but bearings die many other ways — and field engineers learn to read the raceway after a failure the way a pathologist reads tissue.

• Subsurface fatigue spalling. The intended failure mode. After enough load cycles, cracks initiate at non-metallic inclusions a few tenths of a millimetre below the raceway and propagate to the surface, dislodging a flake. Spalls are crater-like, with sharp edges and metallic debris in the lubricant. Time to spalling is what L₁₀ predicts.
• Surface-initiated spalling. When the lubricant film is too thin (κ < 1 in the SKF nomenclature, where κ is the viscosity ratio), asperity contact roughens the surface and cracks start at the surface rather than below it. Far faster than subsurface fatigue.
• Brinelling. Plastic indentation of the raceway by a stationary roller under shock or static overload — a classic dropped-shaft fault. False brinelling is the vibration-induced micro-fretting variant, common in equipment shipped by rail or stored adjacent to vibrating machinery.
• Smearing. When a roller suddenly stops rolling and slides — e.g. on entry to the loaded zone in a lightly loaded bearing — material from one surface welds to the other. Roller skidding in low-load high-speed bearings (jet-engine main-shaft bearings, EV traction-motor bearings) is the classic cause.
• Electric arcing / fluting. A bearing in the rotor circuit of a variable-frequency-drive motor sees high-frequency common-mode voltages that arc across the lubricant film. The result is electrical discharge machining (EDM) damage: pin-prick craters or, with sustained current, a fluted washboard pattern circumferentially around the raceway. The classic fix is an insulated bearing or, increasingly, a ceramic hybrid.
• Corrosion / etching. Water ingress, mismatched lubricants, or condensation from cyclic temperature produce surface oxide that initiates pitting. Marine and outdoor applications need sealed or shielded bearings and water-resistant grease.
• Cage failure. The cage is often the weakest part. Under high acceleration, vibration, or in a bearing operating well above catalogue speed, cage pockets fatigue and a roller breaks free, jamming the bearing in seconds.

Lubrication regimes

Bearings fail far more often from lubrication problems than from fatigue. The role of the lubricant is to maintain a film between roller and race that separates the two surfaces — elastohydrodynamic lubrication (EHL) — and to dissipate the heat generated by rolling and sliding friction.

Two practical regimes dominate. Grease is used for low-to-medium speeds, sealed-for-life applications, and anywhere oil supply is impractical: domestic appliances, automotive wheel hubs, small electric motors, conveyor rollers. A typical lithium-complex grease works from −30 to +120 °C with a base oil viscosity around 100 cSt at 40 °C. Greases are characterised by their NLGI consistency (000 to 6, with 2 being the universal default), thickener type, base-oil viscosity, and operating temperature window.

Oil is used for high speeds, high heat, large bearings, or where the bearing is integral to a circulating lubrication system: industrial gearboxes, machine-tool spindles, gas turbines, large pumps. Oil bath, oil mist, oil jet, oil splash, and circulating-oil systems all see use depending on bearing speed factor n × d_m (where d_m is the pitch diameter in mm); above about n × d_m = 500 000 mm·rpm, grease becomes thermally infeasible and oil takes over.

The key lubrication metric is the viscosity ratio κ = ν / ν₁, where ν is the actual kinematic viscosity at the operating temperature and ν₁ is the reference value required for full-film separation at the bearing's pitch-line velocity. κ ≥ 4 gives long, fatigue-limited life; κ = 1 marks the transition into mixed-film operation where surface-initiated spalling dominates; κ < 0.4 means metal-on-metal contact and seizure risk within hours.

Materials and modern variants

• Standard bearing steel. Through-hardened chromium steel (AISI 52100, EN31, 100Cr6) at 60–64 HRC remains the workhorse. Vacuum-arc-remelt (VAR) and vacuum-induction-melt + vacuum-arc-remelt (VIM-VAR) processes reduce non-metallic inclusion density and double rolling-contact fatigue life over air-melt steel.
• Case-hardened steel. Carburised SAE 8620 or similar produces a hard wear-resistant case over a tough core — useful where shock loads would crack a through-hardened ring. Standard for tapered roller bearings and large spherical rollers.
• Stainless steel. AISI 440C or X65Cr14 for corrosive environments, food machinery, marine, semiconductor processing. Slightly lower fatigue capacity than 52100 but tolerates moisture.
• Ceramic hybrid. Silicon nitride (Si₃N₄) rolling elements with steel races. Lower density (60% of steel) reduces centrifugal roller load at high speed; higher elastic modulus reduces Hertzian deformation; electrical insulation breaks the bearing-current path in inverter-fed motors. Now standard in EV traction-motor main-shaft bearings, motorsport gearboxes, and high-speed machine-tool spindles.
• Full ceramic. Si₃N₄ or zirconia for both rings and rollers. Used in extreme corrosion, high-temperature, or vacuum applications (semiconductor wafer-handling robots, medical pumps).

Configurations and adjacent technologies

• Double-row. Two rows of rollers on a single inner/outer race pair. Doubles the dynamic capacity for the same bore at a small axial-length penalty. Common in railway axleboxes and gearbox shafts.
• Self-aligning. Spherical rollers on a spherical outer raceway. Accept up to ~2° of misalignment. Standard for long-shaft and structurally flexible applications.
• Toroidal (CARB). A 2002 SKF innovation combining spherical self-alignment with axial freedom of the inner ring. Used on the non-locating side of paper-mill rolls and vibrating screens.
• Crossed-roller. Cylindrical rollers alternating at ±45° between a pair of V-grooved rings. Take radial and bidirectional axial load in a single thin ring. Used in robot joints, optical mounts, machine-tool indexing tables.
• Yoke / cam-follower. A heavy thick-walled outer ring runs directly on cam profiles, with the inner ring or stud bolted to the structure. Common in linkage and cam-driven machinery.
• Slewing rings. Very large diameter (often metres) single- or double-row bearings used as the central pivot in cranes, excavators, wind-turbine yaw drives, and tunnel-boring machines.

Where roller bearings show up

• Automotive wheel hubs. Opposed tapered roller pairs (SKF 32213 family in light trucks; integrated hub-bearing units in modern passenger cars) carry the entire vehicle weight, cornering load, and brake reaction.
• Railway axleboxes. Cylindrical or tapered roller bearings rated for 250 000 km between overhauls under multi-tonne axle loads.
• Wind-turbine drivetrains. Spherical roller main shaft bearings (up to 2.5 m bore), planetary gearbox bearings, generator bearings — the most failure-sensitive sub-system in the entire turbine.
• Steel and paper mills. Rolling-mill rolls, paper-machine rolls, calender stacks — multi-tonne loads on bearings the size of small cars.
• Mining and construction. Spherical rollers in vibrating screens, conveyor pulleys, mill trunnions; large tapered rollers in mining-truck wheel ends.
• Aerospace. Cylindrical rollers in gas-turbine compressor and turbine shafts; tapered rollers in helicopter transmission output stages; needle rollers in flight-control mechanisms.
• EV traction motors. Hybrid ceramic-roller bearings to defeat inverter-induced bearing currents, with deep-groove ball or cylindrical roller geometries depending on speed and load.
• Machine tools. Cylindrical roller bearings (NN and NNU series) and crossed-roller bearings for spindle support and rotary table indexing — micron-level runout requirements.

Ball versus roller — when to pick which

Criterion	Ball bearing	Roller bearing
Contact	Point (elliptical patch)	Line (rectangular patch)
Radial capacity (same envelope)	Baseline 1×	2–3×
Axial capacity	Moderate (angular contact) to high (thrust)	Cylindrical: none. Tapered: high. Spherical: moderate.
Speed limit (n × d_m)	Higher (lower friction)	Lower (slightly more drag)
Friction torque	Lower	Slightly higher
Misalignment tolerance	Self-aligning ball: high. Standard: low.	Spherical: high. Cylindrical/tapered: low.
Life exponent p	3	10/3
Sensitivity to overload	2× load → 8× shorter life	2× load → 10× shorter life
Cost (typical)	Lower	10–40% higher
Canonical pick	Light loads, high speeds, low cost	Heavy loads, large shafts, demanding life

The pragmatic rule: pick a deep-groove ball bearing first; switch to a cylindrical roller bearing when radial load exceeds ball-bearing capacity for the available envelope; switch to a tapered roller when you also need to take significant axial load; switch to a spherical roller when shaft alignment cannot be guaranteed; switch to a needle roller when axial length matters more than misalignment tolerance.

Common pitfalls

• Applying thrust to a cylindrical roller. NU and N series cylindrical roller bearings have flanges on one race only; the rollers can drift axially. Asking the flange to take steady thrust generates heat and scuffs the roller ends. Use NUP, NJ + HJ collar, or a separate thrust bearing for axial loads.
• Forgetting the apex condition with tapered rollers. Mixing cones and cups from different bearings even of the same designation, or relying on imprecise spindle geometry, breaks the geometric constraint that allows pure rolling. Always replace matched sets together and verify roller-end clearance after assembly.
• Under-loading a high-speed roller bearing. Lightly loaded high-speed bearings — common in jet-engine main shafts and EV traction motors — can let rollers skid rather than roll. Skidding generates surface smearing and rapid spalling. Designers use a minimum load criterion (Carmin in SKF nomenclature) and sometimes intentionally pre-load the bearing.
• Ignoring κ when picking the lubricant. Catalogue C and L₁₀ assume κ ≥ 1 (mixed film or better). At κ < 0.4 the bearing operates in boundary friction and life is reduced to a small fraction of catalogue regardless of load. Always check operating viscosity against pitch-line velocity.
• Treating the cage as decoration. Cage stresses can be the limiting factor at high speed or high vibration. Steel cages survive higher temperatures but are sensitive to corrosion; polymer cages are quieter and self-lubricating but limited in temperature; brass cages tolerate harsh conditions but are heavier. Match cage choice to environment, not just to catalogue default.
• Mounting force through the wrong race. Pressing the outer ring onto a shaft (or the inner ring into a housing) routes the press load through the rolling elements and brinells the raceways before the bearing has run a single revolution. Always press on the ring that is being interference-fit.
• Neglecting bearing currents in inverter-fed motors. The most common single cause of premature bearing failure in modern variable-speed AC drives. Insulated bearings, ceramic hybrids, or shaft-grounding brushes are all valid mitigations — but the problem must be addressed at the design stage, not after the first failure.