Friction Torque: Where Bearing Losses Come From

Friction torque is the price a bearing pays for carrying load, and although it is small compared with the power a machine transmits, it is not negligible: it sets the starting torque of a drive, the heat generated in a high-speed spindle, and the efficiency of a gearbox. Understanding where the losses come from, and which of them dominate at a given speed and load, is what separates a rough estimate from a reliable calculation - and it is the first step in reducing parasitic losses and running temperatures.

This guide breaks bearing friction into its four mechanisms, shows how the standard two-part model separates them, and lists the levers that actually reduce torque in practice. The thermal consequences of the same losses are treated in the thermal reference speed guide.

Four mechanisms behind one torque number

Bearing friction is a mixture of four distinct mechanisms, and each scales differently with speed and load. Confusing them is why "reduce viscosity" sometimes fixes a hot bearing and sometimes does nothing at all.

Rolling friction: elastic hysteresis

As a rolling element passes through the contact, the material ahead of it compresses and the material behind it recovers, and a small fraction of that energy is lost as heat rather than returned. The loss scales mainly with load, hardly with speed, and it is irreducible at the material level: it is a property of steel under cyclic stress. Modern clean steels have lowered it over the decades, but the designer's only lever is the load itself.

Sliding friction at ribs and cages

Wherever surfaces slide rather than roll - between roller ends and guide ribs, between cage pockets and elements, and at the land-guided cage surfaces - a conventional sliding friction acts, scaled by both load and speed. Its share depends heavily on bearing type: deep groove ball bearings have almost none, while tapered and cylindrical rollers pay for their rib guidance with real sliding work. Crowned rollers and optimised cage pocket geometry reduce it, which is part of what premium internal designs actually sell.

Seal friction: the hidden giant at low speed

The drag of a contact lip seal against its counterface is often the single largest term at low and moderate speed, and two seals double it. Non-contact shields trade a little exclusion for a large reduction in drag; labyrinth and gap seals sit between. The seals and shields guide details the options.

Lubricant drag: churning and shear

Churning oil or working grease around the chamber grows steeply with speed - roughly with the cube of speed for bath lubrication - and with viscosity and fill volume. It is the term that overheats fast gearboxes and the reason oil levels and grease fill percentages are specified so precisely. At high speed it dominates everything else, which is why oil-air systems meter individual drops instead of filling a sump.

The two-part model: M0 plus M1

Most engineering models split bearing torque into a load-independent part and a load-dependent part. The load-independent term M0 captures lubricant drag, seal friction and the small losses that exist even under no load; the load-dependent term M1 captures the hysteresis and sliding losses that scale with the equivalent load. Written as M = M0 + M1, this decomposition underlies the formulas in bearing catalogues and in ISO-based efficiency calculations, and its practical value is diagnosis: at low speed and high load, M1 dominates and lubricant viscosity or seal choice matters less; at high speed, M0 dominates and reducing drag is everything. When a warm bearing refuses to respond to a viscosity change, the machine is probably in the M1 regime and the load, or the internal sliding, is the real driver.

Starting torque and breakaway

A bearing's starting torque is considerably higher than its running torque because the lubricant film has not yet formed and the contacts are in boundary or mixed lubrication. This is why a large machine can be hard to break free from rest and why drives must be sized for starting. Grease thickeners raise starting torque further, and a low-temperature application may need a softer consistency precisely so the drive can overcome the initial resistance - a point the grease selection guide treats in detail. Where a machine must start unloaded anyway, the penalty is small; where it starts under load and in the cold, breakaway torque can be the sizing case.

Speed, churning and the thermal rating

As speed rises, lubricant drag grows faster than linearly, because the oil or grease is sheared at higher rates and more of it is flung around the chamber. In an oil bath this becomes churning, which can overheat a bearing long before its dynamic load limit is reached. This is the physical reason behind maximum speed ratings and behind the ISO 15312 thermal speed rating, which defines the speed at which heat generated equals heat dissipated for a defined reference condition. Exceeding it does not cause immediate failure, but it causes thermal growth, clearance change and reduced grease life - the sequence the temperature and clearance guide follows through. A bath filled to the correct line instead of double depth can be worth 15 to 25 percent of torque on its own at speeds above 3,000 rpm.

Seals: the largest hidden term

Seals are overlooked when efficiency is analysed, yet a contact lip seal can add more torque than the bearing itself at moderate speed, and a standard NBR lip starts to suffer above about 100 °C of chamber temperature. The remedies, in order of effectiveness, are: drop to one seal instead of two, switch to a non-contact shield, or move to a labyrinth design. In high-speed spindles the seal choice can dominate the total loss, and designers sometimes accept lower contamination protection precisely to recover efficiency and reduce heat. The comparison is worth making at design time, because it costs nothing on the drawing and is expensive to retrofit.

Levers for cutting torque

Each lever interacts with life and reliability, so the goal is a balanced design rather than a single minimum-torque number. The life rating guide shows how to check that the thin-oil and right-sizing choices still deliver the target L10 life under ISO 281, and the deep groove deep dive explains why that family remains the low-friction default.

A five-point torque audit

When a drive runs hotter or hungrier than its calculation says it should, work this checklist before touching the design. Each point targets one of the four mechanisms, in the order they usually turn out to be guilty.

• Fill volume. Open the chamber and check the grease fill against 30 to 50 percent of free space, or the oil level against the maker's line; surplus lubricant is pure churning loss.
• Viscosity at temperature. Compare the base-oil grade at the real operating temperature with the minimum the film calculation allows, not with the room-temperature sheet.
• Seal count and type. Count the contact lips and ask whether a shield or a single seal would hold the contamination margin at this speed.
• Clearance after the fit. Confirm the interference fit did not close the internal clearance to zero - a pinched bearing burns its losses in the M1 term.
• No-load baseline. Run the machine empty at speed and log the power; what remains after known gear and fan losses is the bearing M0, and it should match the estimate within about 20 percent.

Efficiency in machine context

For a single well-lubricated, correctly loaded ball bearing the friction coefficient is extremely low - on the order of 0.0010 to 0.0018 for deep groove designs - and the torque is a small fraction of the transmitted load torque. Roller bearings have higher friction because of their line contact and rib sliding; a needle roller in a rockers arm or a thrust bearing under pure axial load pays more. In a gearbox, bearing losses are usually a small share of the total, with gear meshing and seals accounting for most of the rest - but in a high-speed spindle or a lightly loaded, fast-rotating machine the bearings can be the dominant consumer, which is why electric motor efficiency rules have pushed low-friction bearing arrangements into the standard motor design. The applications side is covered in the motor efficiency rules article, and family comparisons start at the needle roller and thrust bearing categories for the higher-friction specialists. The SKF catalogue publishes measured torque bands for its low-friction variants, which is a useful cross-check before committing to a design.

A practical audit closes the loop: measure the drive's no-load power at speed, subtract the known seal and gear contributions, and compare what remains against the M0 plus M1 estimate. If the measured figure is far above the estimate, suspect overfilled grease, dragged seals or a fit that closed the clearance - all cheaper to fix than to redesign.