This is a long post capture from "BigNate", dated 2010, who is a Mod at Anandtech. An excellent review of the the "which bearing" issue right in line with what you have provided but perhaps more depth as he is a turbine engine bearing designer. His view on the cleanliness of automotive engine oil relative to roller bearings is interesting as well. If this is the wrong place for this please advise.
"I do bearing design for turbine engines (they are all high speed ball or roller bearings) at work so I can give a few insights...
1. Load is pretty high, you can design a roller bearing to withstand the load pretty easily but it would be relatively large, noticeably larger than a typical journal bearing. You might be able to fit it in the same axial space on the crankshaft but at the very least you'd have to greatly increase the size of the bottom of the connecting rod to fit in a roller bearing. If you need more axial room than a typical journal bearing you'll probably end up having to lengthen the crankshaft which would force you to lengthen the block. Packaging a correctly sized bearing could significantly affect the rest of the engine design in a bad way. You don't want your bearing design to drive the rest of the engine.
2. Because of reason 1 you're increasing the rotating mass of the engine which isn't a good thing. Also, the balance of the crankshaft will likely be thrown off. Trying to balance a shaft assembly that has 4, 6 or 8 roller bearings on it would be a complete pain, you'd have to make sure each bearing weighed a pretty exact weight or you'd have a crankshaft that was out of balance. I can tell you right now that even bearings for aerospace can differ in weight enough from one bearing to the next that this would be a problem. It's not insurmountable but it would be a concern. Requiring the mass of bearings to be within a certain spec would increase the cost.
3. Roller bearings have limited life. For turbines we can get bearing lives in the 10k hours, 20k hours, or even upwards of 100k hour range depending on the design. However, they pretty much all have a limited life. While they could design bearings beefy enough to effectively have a longer life than the car the size and cost of those bearings would be horrendous. The bearing life for realistically sized bearings would probably be quite low. On the other hand a normal journal bearing has effectively infinite life as long as it has a constant supply of good oil. They have no contacting parts during normal operation and only really wear during starting, stopping, operation with very cold oil, or if something is wrong with the engine. Most modern cars never need their bearings replaced.
4. The life mentioned above for roller bearings is only valid if the oil is extremely clean. The oil in car engines is relatively filthy compared to what is needed for roller or ball bearings running under high loads. There's hard particles all over the place. However, for a car engine the journal bearings have such a thick oil film that the typical containment particle size is smaller than the thickness of the oil film. That lets the particle flow right through without actually being crushed between the inner and outer parts of the journal bearing. On a roller bearing the oil film between the raceways and rollers is much smaller. A particle that flows right through the journal bearing with no problem would get pinched between the roller and raceway. This can create a little dent in the metal and create a point for a crack to begin forming. A large enough dent can take a bearing that has a 100k hour predicted life down to a few hundred hours. This is one of the reasons why transmissions which have rolling element bearings (ball or roller bearings) do not share oil with the engine, it would significantly lower the life of the transmission bearings.
5. As was mentioned before there's no good way to install the bearings without a split ring. That's a poor design for a highly loaded bearing. You could try to make the crank out of a material that would be a suitable raceway material (EXPENSIVE) and then thread the outer ring around the bends in the crank. That may not be possible because the outer ring really isn't that much larger than the other bearing features. You'd then have to install the rollers by hand and use a two piece cage. I'm not a huge fan of two piece cages and the amount of hand assembly would be very expensive and prone to assembly errors. It's not worth the trouble, you're more or less stuck with a split ring.
6. More parts, harder to assemble, and more things to break. None of these things are what car companies want. They'd also be more sensitive to how they're handled before assembly. Many bearing steels used for highly loaded bearings rust like you wouldn't believe. If you let it sit out in the open it will rust. If you touch it with bare hands you might have a rusty finger print form (the oil in your hands attracts water). If you let the vehicle sit for a few months the oil film left on the bearing can attract moisture. That will rust too. Car companies wants the parts to be easy to handle, assemble, and don't want to add any additional service to the car.
7. COST, COST, and did I mention COST? Journal bearings are cheap. I mean ridiculously cheap. I can guarantee that car companies are paying very little for your typical journal bearing. A roller bearing would be significantly more expensive to buy and more difficult to install. Our aerospace bearings can cost hundreds per bearing (I've got one on my desk for a development engine that cost over $3k per bearing to make ), an automotive bearing would be cheaper but it could still be 10x or 100x more than it costs for a journal bearing."
__________________
Ugly can be fixed,stupid is forever.
AnandTech Moderator
