Originally Posted By: Steve S
Remember that if you had 1,000 of the exact same engines the good old bell curve would would rear its ugly head. Also depends on what conditions the engine is operating under. 5,000 of w.o.t. track use will show different numbers all things the same over freeway driving without traffic on I80 between Reno and Elko Nevada.
The "bell curve" concept can be a bit misleading, and some use it to obfuscate discussions. It is based on the field of stochastic processes, which makes some fundamental assumptions about the randomness and independence of the natural world. One particular assumption that you have to be careful about is independence. If two things are truly independent, there is no correlation between them, and therefore, if you take a bunch of them and measure them, you'll see something approaching random.
When it comes to automotive engines, not everything is independent between two engines. The design of an engine and it's metallurgy are highly correlated across multiple engines of the same type. Therefore, we treat them as dependent processes and know that if we make enough measurements that depend on them, those measurements will approach a mean. In any engine design, there is a statistical envelope for it's behavior.
Variation from the mean, however, is random, since there is uncertainty from engine to engine in manufacturing, and when you add these uncertainties together across 1000's of engines, yes, you will see something that is like a bell curve. However, in modern engine design, this statistical process curve is extremely tight, unless there are design or manufacturing flaws.
Automotive manufacturers certainly do not test 1000's of engines prior to production. They do not need to since they can extrapolate the testing that they do perform on 10's to 100's of engines to the population as a whole, because they know their manufacturing process controls. The better the controls are, and the better the CAD modeling tools are, the tighter the ransom part of the process distribution is, allowing designers to shorten the testing process, since they don't need as many test engines to fully characterize it's operation and reliability.
Now, in the "real" world, under driving and maintenance conditions, there are certainly other uncorrelated random variables. However, most of these have a minimal influence on modern engines. People talk about the "impact" of tracking and WOT on engines. However, it turns out that most of the engines that this is done with have significant margin by design, and as a result, there is minimal impact.
The 2.0 TFSI engine is one of these. It has an incredible amount of design margin, as evident in the ability of it to be modified to put out impressive levels of power by tuners who warranty their modifications, like Stasis does. (320 HP / 310 lb-ft torque with a 4-year / 50,000 mile warranty.) As a result, I'd expect minimal impact on UOA for an engine that is operating properly. For example, I've seen many UOA results of Audi RS4 engines that have been run on the track vs. just the street, and the differences are essentialy negligible.
There are other uncorrelated random processes that can give rise to differences: fuel, oil, contaminants, cold starting ... etc. And of course, internal design problems can also result in significant uncorrelated differences between engines.
So when looking at these sorts of plots, I generally like to look at all the indicators, and especially for the outliers that pop out. Even with a "bell curve" or more formally a gaussian probability distribution, there are limits to what the real world can produce. Sometimes these outliers, like those I pointed out, are in the range of probability for what might happen between a large number of engines. Maybe they are not. That's why I like to look at curves like these and try to visually identify the trend for the mean over engine mileage. (BTW, automotive designers do exactly the same thing when testing on a dyno and performing near-real-time oil analysis.) Then, when I see an outlier that is significantly outside of the mean I take a closer look.
Might be nothing. However, knowing the problems these engines have had with the high pressure fuel pump lobe on the camshaft, my interest is peaked. No matter what, there is definitely something different about the two engines I pointed out.