Bearing Preload: How Much and How to Check

Preload is an axial force applied across a bearing set before any working load arrives. It removes internal clearance, pushes every rolling element into contact, and changes the bearing from a slightly loose mechanism into a stiff, precise structure. Designers reach for preload when runout, vibration or shaft stiffness matter, as in spindle bearings, pinion shafts, differentials and wheel hubs. But preload is one of the few parameters where more is genuinely worse after a point, so understanding what it buys and what it costs is essential.

This guide explains the mechanics, the trade curve between stiffness and life, the two ways of applying the force, the verification methods used on the bench, and the thermal trap that ruins more spindles than any bearing defect. The paired-mounting geometry that preload relies on is covered in the angular contact pairing guide.

Play, stiffness and the case for preload

A bearing delivered with clearance has some play. Under light load only a few rolling elements carry the load, and the unloaded balls at the top of the ring let the shaft float slightly. Preload forces all elements into shared contact, which removes the play and dramatically increases both radial and axial stiffness, especially in the low-load region where an unpreloaded bearing is weakest. The gain is real but front-loaded: the first 20 to 50 newtons of preload on a small angular contact pair buy most of the improvement, and beyond a few hundred the curve flattens while the costs keep climbing.

What preload buys and what it costs

In a ball bearing, applying axial preload also increases the effective contact angle, which raises the thrust stiffness further. The measurable benefits are reduced running runout, a higher resonant frequency of the shaft, quieter running and better load sharing among the balls. The costs rise on the same schedule: friction torque climbs, running temperature climbs, and the calculated fatigue life falls, because every ball now carries a standing contact stress before any working load arrives. The life penalty is not linear either - moving from a light to a medium preload can add 30 to 60 percent more standing contact stress, and since calculated life under ISO 281 varies with roughly the cube of that stress, a 10 percent rise in contact stress costs about a quarter of the calculated life. An over-preloaded set can lose half its design life for a stiffness gain it never needed.

Light, medium, heavy: the shape of the curve

The best operating point is usually the smallest preload that removes the play and meets the stiffness target, not a heavy figure chosen for comfort. Stiffness rises quickly at first and then flattens; life falls steadily throughout; temperature follows the friction. A stiffness specification that cannot be met at light preload is a reason to look at the surrounding structure, not to reach for the heavy class - a point developed under "the whole shaft system" below.

Position preload versus spring preload

There are two ways to hold the force, and they behave oppositely under temperature change.

Position (rigid) preload

Rigid preload is set by the geometry of the assembly: matched ring widths, a shim stack, or a ground spacer between the two bearings. It is simple, repeatable and stiff, and it holds the geometry exactly - which is precisely why it is thermally sensitive. Any axial growth of the shaft relative to the housing adds directly to the force, so a set that is correct when cold can become heavily over-preloaded when hot. Rigid preload suits moderate speeds and duties where the temperature swing is small and predictable.

Spring (constant-force) preload

Spring preload uses a wave washer or coil spring pressing one ring of the non-locating bearing, so the force stays nearly constant even as the shaft grows. It is the standard choice for high-speed arrangements where the temperature swing is large, and its only costs are that it does not locate the shaft axially (a second bearing must) and that it adds one more part to assemble. Constant-force preload also tolerates wear better: as faces wear a few micrometres, a spring follows while a rigid geometry goes slack. Deciding when to choose spring preload over rigid geometry comes down to two numbers - the expected speed, and the expected shaft-to-housing temperature difference; when the latter exceeds roughly 10 to 15 °C at any duty point, springs usually win.

Setting and verifying the force

In practice the correct preload is defined by the bearing maker as a force band for the given size and is delivered through the suffix code (light, medium or heavy) or through a matched set. On the bench the applied force is verified indirectly, by one of two measurements.

Displacement measurement is the most direct: load the shaft axially with a known force, read the movement, and compare it with the maker's deflection curve for the pair - bearing makers such as SKF publish these curves for their universal-matching sets. Torque measurement is faster but conflates preload with seal drag and lubricant condition, so it needs a clean baseline. In production, matched pairs and controlled spacer widths replace per-unit measurement entirely.

Matched sets in production

Series production buys its preload in the part number: a universal-matching pair is ground with a controlled stand-out so that clamping the inner rings together delivers the maker's light, medium or heavy class without any measurement at assembly. The savings in bench time are why matched sets dominate spindle and pump production - but they also mean the preload cannot be changed after the fact, so the class must be right at specification time. Matched angular contact sets are ordered as a unit - the angular contact category lists them as paired suffixes, and the locknut and adapter guide covers the hardware that holds the setting.

The differential-growth trap in rigid sets

As a shaft spins, its temperature rises faster than the housing's, and the differential growth shows up as extra preload in a rigidly set pair. A steel shaft 200 mm between bearings that runs 25 °C hotter than its housing grows about 0.055 mm axially - enough to transform a correct cold setting into a heavy overload. A set that was correct when cold can become over-preloaded when hot, which raises friction, raises temperature again, and can run away into seizure. This is the single most common reason a spindle overheats: not a defective bearing, but a rigid preload that did not allow for the operating temperature difference. The physics of the expansion is quantified in the temperature and clearance guide. The standard remedies are to move to spring preload, to reduce the cold-set force so the warm state lands in band, or to design the spacer so it expands with the shaft.

Tapered rollers and end play

The same logic applies wherever two rows face each other. Tapered roller wheel bearings are set to a slight interference or a controlled end-play by tightening the axle nut or shimming the cups - typically 0.01 to 0.05 mm of end play for a road wheel bearing: too tight scorches the rollers, too loose lets the wheel wander. The target is a small controlled clearance or a small preload, verified by nut torque or by measuring rolling resistance, and then rechecked warm. The designation and setting details are in the tapered roller guide, and the heavier two-facing-row cases in gearboxes follow the same rules with the cylindrical roller and tapered roller families.

Preload inside the whole shaft system

Preload is a property of the shaft assembly, not of a single bearing. The stiffness a designer feels at the tool or the encoder depends on the two bearings, the housing, the spacers and the shaft acting together, and the softest element governs. Adding preload to two very stiff bearings cannot rescue a flexible shaft or a thin-walled housing; the shaft and housing geometry guide shows how to find the weak member. Before chasing a preload figure, ask whether the surrounding structure is stiff enough for the preload to matter at all.

Preload also interacts with speed. As a preloaded set speeds up, centrifugal force shifts the balls outward and changes the contact angles, and friction torque rises faster than it would in a clearance bearing. A preload that gives excellent stiffness at 3,000 rpm can overload the same spindle at 12,000 rpm, which is why constant-pressure spring preload is preferred where the speed range is wide. In practice the preload level is always a compromise between the stiffness needed at low load and the heat tolerated at high speed.

Finally, preload has to be maintained. Over a long service life, wear at the spacer faces, creep of a fit, or a loosening lock nut can all reduce the applied force and let the assembly go slack. Locating nuts should be locked, spacers should be hardened and ground flat, and any arrangement that depends on friction to hold its preload should be re-checked at intervals rather than trusted indefinitely.

Choosing the preload level

• Required stiffness and runout of the assembly, measured at the point of use, not at the bearing.
• Speed and the expected temperature rise, which decide rigid versus spring preload.
• Lubrication and available cooling, which set how much standing heat the set can afford.
• The fatigue life target, checked against the standing stress the preload adds.
• Whether the design is adjustable in service or sealed for life, which decides the verification method.

It is worth separating preload from fit, because the two are often confused. A tight interference fit can feel like preload on the bench, but it acts radially on the ring and does not control the axial position of the rolling elements. Preload is axial and deliberate; fit is radial and structural. A bearing can be correctly preloaded and badly fitted at the same time, and it will still fail. The two must be specified and checked independently, and the fit guide handles the radial side. Where friction consequences need quantifying before the preload class is fixed, the friction torque guide gives the M0 plus M1 framework, and the bearing selector filters for universal-matching and paired variants.