Working principle for a ball bearing

A 4 point angular contact ball bearing

A ball bearing with a semi transparent cage

Wingquist's and SKF's self-aligning ball bearing

Wingquist bearing
A ball bearing is a type of rolling-element bearing that uses balls to maintain the separation between the bearing races.
The purpose of a ball bearing is to reduce rotational friction and support radial and axial loads. It achieves this by using at least two races to contain the balls and transmit the loads through the balls. In most applications, one race is stationary and the other is attached to the rotating assembly (e.g., a hub or shaft). As one of the bearing races rotates it causes the balls to rotate as well. Because the balls are rolling they have a much lower coefficient of friction than if two flat surfaces were sliding against each other.
Ball bearings tend to have lower load capacity for their size than other kinds of rolling-element bearings due to the smaller contact area between the balls and races. However, they can tolerate some misalignment of the inner and outer races.

Common designs

There are several common designs of ball bearing, each offering various trade-offs. They can be made from many different materials, including: stainless steel, chrome steel, and ceramic (silicon nitride (Si3N4)). A hybrid ball bearing is a bearing with ceramic balls and races of metal.
[edit]Angular contact
An angular contact ball bearing uses axially asymmetric races. An axial load passes in a straight line through the bearing, whereas a radial load takes an oblique path that tends to want to separate the races axially. So the angle of contact on the inner race is the same as that on the outer race. Angular contact bearings better support "combined loads" (loading in both the radial and axial directions) and the contact angle of the bearing should be matched to the relative proportions of each. The larger the contact angle (typically in the range 10 to 45 degrees), the higher the axial load supported, but the lower the radial load. In high speed applications, such as turbines, jet engines, and dentistry equipment, the centrifugal forces generated by the balls changes the contact angle at the inner and outer race. Ceramics such as silicon nitride are now regularly used in such applications due to their low density (40% of steel). These materials significantly reduce centrifugal force and function well in high temperature environments. They also tend to wear in a similar way to bearing steel—rather than cracking or shattering like glass or porcelain.
Most bicycles use angular-contact bearings in the headsets because the forces on these bearings are in both the radial and axial direction.
[edit]Axial
An axial ball bearing uses side-by-side races. An axial load is transmitted directly through the bearing, while a radial load is poorly supported and tends to separate the races,so that a larger radial load is likely to damage the bearing.
[edit]Deep-groove
In a deep-groove radial bearing, the race dimensions are close to the dimensions of the balls that run in it. Deep-groove bearings can support higher loads.
[edit]Construction types

[edit]Conrad
The Conrad-style ball bearing is named after its inventor, Robert Conrad, who was awarded British patent 12,206 in 1903 and U.S. patent 822,723 in 1906. These bearings are assembled by placing the inner race into an eccentric position relative to the outer race, with the two races in contact at one point, resulting in a large gap opposite the point of contact. The balls are inserted through the gap and then evenly distributed around the bearing assembly, causing the races to become concentric. Assembly is completed by fitting a cage to the balls to maintain their positions relative to each other. Without the cage, the balls would eventually drift out of position during operation, causing the bearing to fail. The cage carries no load and serves only to maintain ball position.
Conrad bearings have the advantage that they are able to withstand both radial and axial loads, but have the disadvantage of lower load capacity due to the limited number of balls that can be loaded into the bearing assembly. Probably the most familiar industrial ball bearing is the deep-groove Conrad style. The bearing is used in most of the mechanical industries.
[edit]Slot-fill
In a slot-fill radial bearing, also referred to as a full complement design, the inner and outer races are notched on one face so that when the notches are aligned, balls can be slipped in the resulting slot to assemble the bearing.  A slot-fill bearing has the advantage that the entire groove is filled with balls, called a full complement, resulting in a higher radial load capacity than a Conrad bearing of the same dimensions and material type.  However, a slot-fill bearing cannot carry a significant axial load on the loading slot side.  Also, the slots cause a discontinuity in the races that has a small but adverse effect on strength.  Note that an angular contact bearing can be disassembled axially and therefore is a full complement design.
[edit]Split-race
The outer race may be split axially or radially, or a hole drilled in it for filling. These approaches allow a full complement to be used, but also limit the orientation of loads or the amount of misalignment the bearing can tolerate. Thus, these designs find much less use.
[edit]Rows
There are two row designs: single-row bearings and double-row bearings. Most ball bearings are a single-row design, which means there is one row of bearing balls. This design works with radial and thrust loads.[1]
A double-row design has two rows of bearing balls. Their disadvantage is they need better alignment than single-row bearings.
[edit]Flanged
Bearings with a flange on the outer ring simplify axial location. The housing for such bearings can consist of a through-hole of uniform diameter, but the entry face of the housing (which may be either the outer or inner face) must be machined truly normal to the hole axis. However such flanges are very expensive to manufacture. A more cost effective arrangement of the bearing outer ring, with similar benefits, is a snap ring groove at either or both ends of the outside diameter. The snap ring assumes the function of a flange.
[edit]Caged
Cages are typically used to secure the balls in a Conrad-style ball bearing. In other construction types they may decrease the number of balls depending on the specific cage shape, and thus reduce the load capacity. Without cages the tangential position is stabilized by sliding of two convex surfaces on each other. With a cage the tangential position is stabilized by a sliding of a convex surface in a matched concave surface, which avoids dents in the balls and has lower friction. Caged roller bearings were invented by John Harrison in the mid-18th century as part of his work on chronographs.[2] Caged bearings were used more frequently during wartime steel shortages for bicycle wheel bearings married to replaceable cups.
[edit]Ceramic hybrid ball bearings using ceramic balls
Ceramic bearing balls can weigh up to 40% less than steel ones, depending on size and material. This reduces centrifugal loading and skidding, so hybrid ceramic bearings can operate 20% to 40% faster than conventional bearings. This means that the outer race groove exerts less force inward against the ball as the bearing spins. This reduction in force reduces the friction and rolling resistance. The lighter balls allows the bearing to spin faster, and uses less energy to maintain its speed.
While ceramic hybrid bearings use ceramic balls in place of steel ones, they are constructed with steel inner and outer rings; hence the hybrid designation.
[edit]Self-aligning
Self-aligning ball bearings, such as the Wingquist bearing, are constructed with the inner ring and ball assembly contained within an outer ring that has a spherical raceway. This construction allows the bearing to tolerate a small angular misalignment resulting from deflection or improper mounting.
[edit]Operating conditions

[edit]Lifespan
The calculated life for a bearing is based on the load it carries and its operating speed. The industry standard usable bearing lifespan is inversely proportional to the bearing load cubed. Nominal maximum load of a bearing (as specified for example in SKF datasheets), is for a lifespan of 1 million rotations, which at 50 Hz (i.e., 3000 RPM) is a lifespan of 5.5 working hours. 90% of bearings of that type have at least that lifespan, and 50% of bearings have a lifespan at least 5 times as long.[3]
The industry standard life calculation is based upon the work of Lundberg and Palmgren performed in 1947. The formula assumes the life to be limited by metal fatigue and that the life distribution can be described by a Weibull distribution. Many variations of the formula exist that include factors for material properties, lubrication, and loading. Factoring for loading may be viewed as a tacit admission that modern materials demonstrate a different relationship between load and life than Lundberg and Palmgren determined .[3]
[edit]Failure modes
If a bearing is not rotating, maximum load is determined by force that causes non-elastic deformation of balls. If the balls are flattened, the bearing does not rotate. Maximum load for not or very slowly rotating bearings is called "static" maximum load.[3]
If that same bearing is rotating, that deformation tends to knead the ball into roughly a ball shape, so the bearing can still rotate, but if this continues for a long time, the ball fails due to metal fatigue. Maximum load for rotating bearing is called "dynamic" maximum load, and is roughly two or three times as high as static max load.[3]
If a bearing is rotating, but experiences heavy load that lasts shorter than one revolution, static max load must be used in computations, since the bearing does not rotate during the maximum load.[3]
[edit]Maximum load
In general, maximum load on a ball bearing is proportional to outer diameter of the bearing times width of bearing (where width is measured in direction of axle).[3]
[edit]Lubrication

For a bearing to operate properly, it needs to be lubricated. In most cases the lubricant is based on elastohydrodynamic effect (by oil or grease) but working at extreme temperatures dry lubricated bearings are also available.
For a bearing to have its nominal lifespan at its nominal maximum load, it must be lubricated with a lubricant (oil or grease) that has at least the minimum dynamic viscosity (usually denoted with the Greek letter ) recommended for that bearing.[3]
The recommended dynamic viscosity is inversely proportional to diameter of bearing.[3]
The recommended dynamic viscosity decreases with rotating frequency. As a rough indication: for less than 3000 RPM, recommended viscosity increases with factor 6 for a factor 10 decrease in speed, and for more than 3000 RPM, recommended viscosity decreases with factor 3 for a factor 10 increase in speed.[3]
For a bearing where average of outer diameter of bearing and diameter of axle hole is 50 mm, and that is rotating at 3000 RPM, recommended dynamic viscosity is 12 mm?/s.[3]
Note that dynamic viscosity of oil varies strongly with temperature: a temperature increase of 50–70 °C causes the viscosity to decrease by factor 10.[3]
If the viscosity of lubricant is higher than recommended, lifespan of bearing increases, roughly proportional to square root of viscosity. If the viscosity of the lubricant is lower than recommended, the lifespan of the bearing decreases, and by how much depends on which type of oil being used. For oils with EP ('extreme pressure') additives, the lifespan is proportional to the square root of dynamic viscosity, just as it was for too high viscosity, while for ordinary oil's lifespan is proportional to the square of the viscosity if a lower-than-recommended viscosity is used.[3]
Lubrication can be done with a grease, which has advantages that grease is normally held within the bearing releasing the lubricant oil as it is compressed by the balls. It provides a protective barrier for the bearing metal from the environment, but has disadvantages that this grease must be replaced periodically, and maximum load of bearing decreases (because if bearing gets too warm, grease melts and runs out of bearing). Time between grease replacements decreases very strongly with diameter of bearing: for a 40 mm bearing, grease should be replaced every 5000 working hours, while for a 100 mm bearing it should be replaced every 500 working hours.[3]
Lubrication can also be done with an oil, which has advantage of higher maximum load, but needs some way to keep oil in bearing, as it normally tends to run out of it. For oil lubrication it is recommended that for applications where oil does not become warmer than 50 °C, oil should be replaced once a year, while for applications where oil does not become warmer than 100 °C, oil should be replaced 4 times per year. For car engines, oil becomes 100 °C but the engine has an oil filter to continually improve oil quality; therefore, the oil is usually changed less frequently than the oil in bearings.[3]
[edit]Direction of load

A ball bearing with a semi transparent cage

Most bearings are meant for supporting loads perpendicular to axle ("radial loads"). Whether they can also bear axial loads, and if so, how much, depends on the type of bearing. Thrust bearings (commonly found on lazy susans) are specifically designed for axial loads.[3]
For single-row deep-groove ball bearings, SKF's documentation says that maximum axial load is circa 50 % of maximum radial load, but it also says that "light" and/or "small" bearings can take axial loads that are 25 % of maximum radial load.[3]
For single-row edge-contact ball bearings, axial load can be circa 2 times max radial load, and for cone-bearings maximum axial load is between 1 and 2 times maximum radial load.[3]
If both axial and radial loads are present, they can be added vectorially, to result in total load on bearing, which in combination with nominal maximum load can be used to predict lifespan.[3] However, in order to correctly predict the rating life of ball bearings the ISO/TS 16281 should be used with the help of a calculation software.
[edit]Avoiding undesirable axial load
The part of a bearing that rotates (either axle hole or outer circumference) must be fixed, while for a part that does not rotate this is not necessary (so it can be allowed to slide). If a bearing is loaded axially, both sides must be fixed.[3]
If an axle has two bearings, and temperature varies, axle shrinks or

Working principle for a ball bearing

 expands, therefore it is not admissible for both bearings to be fixed on both their sides, since expansion of axle would exert axial forces that would destroy these bearings. Therefore, at least one of bearings must be able to slide.[3]
A 'freely sliding fit' is one where there is at least a 4 ?m clearance, presumably because surface-roughness of a surface made on a lathe is normally between 1.6 and 3.2 ?m.[3]