The Ins and Outs of Ball and Roller Bearings

If you look into a parts book or bearing catalog, you will see that bearing dimensions are basic: What is the OD, ID, and thickness of the bearing? The bearing type is also specified. A basic “Conrad” ball bearing is usually supplied with a two-piece press-formed steel ribbon separator whose purpose is to space the seven balls so that they cannot rub against each other. Cage halves are usually riveted together, but may also be joined by spot-welds or fold-over tabs.

The raceway grooves naturally have radii larger than the balls, otherwise friction-generating scrubbing would occur. Conceptually, the balls make point contact with inner and outer races, but as load is applied, the elasticity of the 52100 bearing steel allows ball-to-race contact to expand into more like an ellipse. The 52100 is a high-carbon alloy with some chromium, intended to be through-hardened. It has existed since 1905.

There are also levels of bearing quality; the ABEC number is a measure of raceway roundness, surface asperity height, and the roundness and diameter of balls (ABEC = Annular Bearing Engineering Committee). Higher ABEC numbers (5, 7, etc.) are for applications such as high-speed grinder spindles, requiring excellent concentricity and/or freedom from vibration.

Another important bearing variable is internal clearance. Depending on how bearings are installed, the outer bearing race may be slightly compressed, either by pressing it into a counterbore with some interference, or by clamping it between bolted-together elements (as in horizontally split crankcase halves). The inner race is slightly expanded by being pressed onto a shaft, again with some interference. If this type of installation is made with a bearing having standard internal clearance, the result may be preload—the races compressing the balls at all times. To avoid this, bearings with greater internal clearance are made. If you have looked at the main crankshaft bearings of the classic two-stroke Yamaha RD350 or 400 twins, you will see “C3″ stamped into one of the faces of the outer race. This indicates the extra internal clearance such bearings are given.

Many a home rebuilder has gone to the bearing supplier to get notionally higher-quality US-made bearings for his beloved bike, saying, “Gimme the best thing you got.” The clerk sets out an ABEC 7 bearing. Alas, higher grades of precision are not supplied in C3, so when the expensive bearings are installed as the crankshaft is rebuilt, the resulting preload quickly destroys them.

Other types of ball or roller cages are supplied for particular applications. When Yamaha TZ250/350 race crankshafts of early 1970s production fatigued and cracked the ball separators of their inner pair of main bearings, the solution turned out to be a cheap, snap-in plastic separator whose damping quieted down the irregular ball motions that had been cracking steel ribbon separators.

Yet I have also seen those steel ribbon separators go 50,000 miles in Kawasaki 500 H1 Triples operated at eyebrow-raising highway speed. I have never seen a single failure of the snap-in plastic separators.

Ribbon separators are guided by the balls themselves in steel-to-steel contact. Bronze separators, guided by the outer race, are more benign because steel and bronze have got along well since the days of bronze journal bearings on railcar axles. Bearing catalogs explain such options. When in 1970 we received our Kawasaki H1R production racer, we were stunned to see that its crankshaft ran in six bronze-caged roller bearings. Cost to the dealer then was $105—for the whole crankshaft. The miracle of mass production.

When I lucked into a conversation with a bearing engineer, he told me that in some high-speed applications, the balls do not roll smoothly around their races, but can instead quite vigorously track from side to side, stressing the separator. It’s typical for separator cracking to occur where the metal has been highly stressed during manufacture—sharp bends and the punched rivet holes. Look for cracking in such places. I keep a box of variously failed bearings to remind me of the role of imperfection in human endeavor.

Back in the days of hyper-expensive ICBM inertial guidance systems employing actual spinning gyros, ball bearings of unimaginable refinement were required, able to function normally over 24,000 hours of continuous operation. Guidance systems on ICBMs ran continuously to cover the case in which diplomacy failed. The stable platforms were periodically aligned by star sightings, compensating for inevitable “drift.”

Cage or separator failures are the most common, but also seen is so-called “case crushing” in roller bearings. Such bearings are case-hardened rather than through-hardened, and the subsurface shear fatigue caused by the millions of cycles of roller pressure against (typically) the inner race eventually begins to flake tiny bits from the surface. Modern industrial equipment often includes acoustic analysis to predict such bearing failures so they don’t stop production. As the normal acoustic signature begins to shift to the new sound of rollers passing over a deteriorating surface, that bearing can be tagged for replacement at the next scheduled downtime.

Until the mid-late 1950s a major determinant of jet engine life was the durability of main turbine shaft bearings. Because such bearings are fair-sized, and turn at high speeds, most of the load producing raceway fatigue was created by “centrifugal force,” forcing the rollers to travel around their curved races rather than going straight as Isaac Newton long ago understood was their preference.

This brings us to a side issue—making a game of spinning ball bearings to super rpm with an air jet. Frightening rpm are possible in this way, but consider: The force of the balls as they seek to go straight but are constrained to the small circle of the outer race is very large. What if the outer race bursts while the balls are moving at hundreds of feet per second? They become hard steel bullets.

One response to bearing raceway fatigue was to make the balls lighter—by making them from two hollow hemispheres, friction-welded together. A later response was to make the balls of ceramic, such as sapphire or later, silicon nitride (the density of steel is about 7.8, but that of stone is more like 3).

I learned about this by putting on a coat and tie so as to resemble a TA, then strolling into the MIT engineering library where I spent the day reading papers and taking notes. It was very pleasant, but I had a goal: to understand premature big-end needle bearing failure in two-strokes. That was 1971. Today much of that information is available online.

The process of surface fatigue usually begins at a stress concentration, such as occurs around a foreign inclusion in the material. Study of wrought iron rivets recovered from the sunken liner Titanic showed that they contained 12% slag. Imagine stress, such as that applied by collision with a great iceberg, trying to thread its way past all the slag inclusions. Pop goes the rivet.

Sometime in the 1950s the concept of remelting bearing steel in a vacuum, to allow internal impurities to boil away, was made into a practical industrial process. Early examples of bearings whose balls and races were made of vacuum remelt showed fatigue life six times greater than bearings of plain old 52100.

Size for size, a roller bearing has 40% more radial load capacity than a ball. While ball bearings have a bit of end-thrust capacity, cylindrical roller bearings have none. Angular contact bearings (sometimes in past times called “magneto bearings”) are designed to provide accurate axial location of a shaft. Tapered roller bearings are sometimes used to support a motorcycle’s steering head instead of the traditional uncaged balls and races. Their reason to exist was originally as railroad axle bearings.

Install stock outer ball races into the steering head, then coat each one with grease. Next, tediously load all the balls by sticking them in the grease…

Other possible bearing features include snap rings, to provide positive axial location (gearbox ball bearings are seen with this feature). To prevent outer races from rotating in engine cases they may be doweled or otherwise positively prevented from rotation. Yes, I have seen cases that were sprung out of square by being assembled without locating the dowels properly. New crankcase time!

Although your friends will regard you as odd if they see you actually reading the catalogs of rolling bearing manufacturers, there is a lot of practical information there. Have at it.