Wait you mean the bar of Dr. Squatch special edition soap made with Sydney Sweeny's bathwater for ten bucks was not an investment?
Determining the Proper Oil for Your Engine - Email from Lake Speed Jr
- Thread starter rstcso
- Start date
Well, I'm waiting for someone to send it off for analysis......but for true research purposes we will need a virgin sample of said bath water. If not.....well, it's all for nothing, ya know.
Just compare the differences in results from a few different labs with the same sample to see the futility of small sample size UOA. On my last sample there was a full order of magnitude difference in the LSJ “were metal per 1000 miles” metric between wearcheck, which reported 0.5 ppm/1k and SpeeDiagnostix which reported 5.1 ppm/1k.
As time passes I have less and less faith in the equipment accuracy and lab practices of these UoA vendors. I would like to to think that at least a huge 50+ ppm element spike would still show up and allow a preemptive maintenance investigation. However, I’m starting to wonder if they even send results for the correct sample or if they “smooth” any anomalous data manually based on an expected range of values. Fuel Dilution test is another one where there seems to be a 2-3x difference between labs that all claim to be using gas chromatography.
As time passes I have less and less faith in the equipment accuracy and lab practices of these UoA vendors. I would like to to think that at least a huge 50+ ppm element spike would still show up and allow a preemptive maintenance investigation. However, I’m starting to wonder if they even send results for the correct sample or if they “smooth” any anomalous data manually based on an expected range of values. Fuel Dilution test is another one where there seems to be a 2-3x difference between labs that all claim to be using gas chromatography.
Personally, I would expect an 8 cylinder to shed more than a 4, and a 12 cylinder more again.
OVERKILL
$100 Site Donor 2021
Swept ring area, for sure. My 6.4L has 4.1" bores, there's a lot of swept ring area there.
As would I. More cylinders relative to displacement means the engine has a surface-area-to-volume ratio that would cause higher wear metal generation.
In other words, a 4.0 V8 should, ceteris paribus, have higher nominal wear metals than a 4.0L V6 or (theoretical) 4.0L four-cyl engine (which nobody builds for NVH reasons).
Not that I have any real gravitas to add to the discussion here, but I wanted take a side here with @dnewton3. I do engines and data analysis for a living and while I'm not an SPC engineer, I'm fairly conversational in the language of the trade. My machine has JMP and Minitab on it and I know how to use them. I have done reliability studies and Weibull modeling with mixed-censored data. I never learned Matlab so I end up brute force processing data sets the hard way using JMP or Minitab or (Lord help me) Excel. It has sufficed for the last couple decades.
I like LSJr and I'm sure I'd get along with him personally. But knowing chemistry and oil doesn't mean you know data interpretation per se. It's one thing to know your own sample data (i.e. all the analyses SpeedDiagnostix has run) and to know what is typical for a given engine within your data. That's the macro view and I think is where Lake is probably most authoritative.
But the methods @dnewton3 is talking about-- and the flaws he correctly identified in LSJr's approach-- are rooted in micro analysis. You aren't trying to pick an oil for all chevy trucks all over the world or all Volvos in Sweden. You are trying to pick an oil for YOUR vehicle-- a very tiny sample size. This is why Dnewton's approach focuses so much on baseline and analyzing your own data. Knowing what some oil does in someone else's duty cycle, burning someone else's local fuel, in someone else's car really doesn't offer you much in the way of conclusive data. Those drivers aren't you and your car isn't theirs.
This is why the macro view is of limited use. Your particular car or truck is just one engine within a massive dataset.
The micro view is more a "Case study" approach. Depth, not width. It's the difference between knowing the surgeon general's nutrition recommendations and knowing your own diet and exercise habits and exactly how your health is responding (or not) to those habits. It's the difference between reading the latest nutritional study vs knowing your own blood work in light of your own diet and activity routine.
One innovation that I'd like to see from oil analysis labs in their data presentation is a publication of percentiles within their data sets. Presumably the data is not normally distributed (normal distributions fail where there is a practical zero bound), so instead of mistakenly thinking in terms of "sigma" and standard deviations, I think a simple percentile would be most helpful and insightful. Heck, even decile-level resolution would likely be useful for many of us.
Let's say I send in a lab sample. Wouldn't it be neat to see a column in each wear metal row showing the percentile for the oil duration? Not just say, 8ppm iron in 4000 miles, but that your 8ppm in 4000 miles is the 73rd percentile for all vehicles in our data set with the same engine. Or the percentile in our data set of all vehicles in your ZIP code or state, or percentile of all vehicles in our data.
I'd love to see a normalized (per 1000mi) table of percentiles for each wear metal. This would go a long way towards putting the wear rates in context. So I have 8ppm of iron, is that normal? For my Toyota engine? For my area? For my oil life?
This would take that large macro-data set and help make it more relevant to the micro-trend your are trying to establish. This insight would let you, as the consumer of the UOA, know where your particular data falls within the context of all other engines, all others using your same engine, or all others using the same oil, etc etc. It addresses one of the main shortcomings of UOA by answering the "compared to what?" question in a useful way.
It's possible that the UOA that is normal for a Hemi might be alarmingly high for a GM 3800 or some other engine notorious for being "easy on oil." But what does "easy on oil" even mean? Well, it hints at what we really care about-- how is my engine and my oil doing compared to some other engine, some other oil, some other user?
If you only have micro, you forfeit the "compared to what" part and you're left basically with "compared to itself" which, while supremely valid for trending purposes, allowed very little comparison and context. But without the micro, your have no visibility on your particular data within the broader context.
This is why I think the provision of "percentiles" for key elements of the wear metal analysis would be super, super helpful. And why I feel like that absence is a real missed opportunity.
Forgive me for not reading beyond page one but it seems to me that paying for UOAs on a daily driver before it reaches 30k is just a waste of money. One might switch between the dealer oil and any other oil within this window and note the wear metals going down. Then they might conclude that their choice of oil was sound even though all that they are really seeing is normal break-in behavior.
What you said is so True, no need for the Forgive Statement.
A "waste of money" is in the eye of the beholder.
But a prominent member here recently had a low-mile Duramax in a Chevy tahoe show evidence of imminent major engine failure in a "waste of money" oil analysis.
I bet many folks wrestling with GM over 6.2L engine failures are likely to have a different assessment on the "waste of money" of early life oil analysis. They probably would have liked to know their engines were in the process of failing.
To the contrary, it is well-known in Reliability engineering that product failure rates follow a "bathtub curve" and they are more likely to fail BEFORE 30k miles than in the span of time from 30-60k, for example. This is because the book ends of the "bathtub curve" are "infant mortality" and then "wear out fail modes."
Thus, I'd argue that the early life sampling is MORE critical than sampling after 30k. The thing is you need to see the early life wear dropping steadily and stabilizing. IN @wwillson and the Tahoe, he caught the wear metals spiking back up after they initially leveled off from the normal elevated break in levels.
It's worth reading his post about his Tahoe and duramax. I don't see how someone could possibly read his post and conclude the early life oil analysis is a waste of money.
OK, it's probably a waste of money on a Honda, but on a GM, definitely not
I kid.For normal daily driven cars, it's a complete waste of money IMHO - 5K miles changes with Walmart stuff and it will run as long as it will run. If it's interesting/fun for you, then the UOAs have a value. If you are using these in a HD use case, also not a waste.
I'm sure he's on the same level as LSPjr correct?
No Mr. Newton is not at the same level as LSJ when it comes to knowledge and experience with statistics, he far far exceeds LSJ. I've never one time heard LSJ claim to be an expert in statistics.
It shouldn't be up to any of us to spend our money to discover a design or manufaturing problem. UOAs are useful for fleet operators or imdividuals wanting to push it towards 200k.
There’s the ideal world and there’s the real world. The world you occupy might determine your course of action.
Lol! Too funny.
Remember that people don't change their mind in public. They do it in private. The most you can do is make an argument civilly and with evidence, but many times there's not much else you can do. If someone arrives at a viewpoint based on emotion and not logic, you're not going to change their mind with logic.
Is this all about the specific engine or the oil?
The main concern with Mr. Speed's advice, JMHO, was that his recommendation was flawed because it didn't take into account that different engines wear differently and that <5ppm wear metals / 1000 mi. for one engine may not be considered a normal/acceptable amount of wear for other engines.
Similar threads
- Locked