Do automakers REALLY test long-term durability?

Status
Not open for further replies.
Originally Posted By: d00df00d
Guys, any big manufacturer makes TONS of engines. Try doing some quick math to figure out how low their failure rate would have to be so that you wouldn't easily be able to think of a few duds.


If you're trying to say that no manufacturer has ever released an engine that wasn't properly tested/engineered and turned out to be an enormous failure, you couldn't be more wrong.

In the case of the FORD Powerstroke 6.0 engine most failures happened in the very first years at under 100,000 miles. Failures were ridiculously rampant. They clearly did not do their homework here. If they had done any major amount of testing before releasing it they would have seen high failure rates and pulled back.
 
Originally Posted By: OilFilters
If you're trying to say that no manufacturer has ever released an engine that wasn't properly tested/engineered and turned out to be an enormous failure, you couldn't be more wrong.

Ugh, no! He's definitely not saying that!

Yes, automakers do long term reliability testing.
No, they don't use customers as beta testers!

Unless we're talking about Jeep.
In that case, yes, the consumers are definitely the beta testers!
28.gif


And d00df00d is right about the availability heuristic as well.
 
Thanks for the assist, but I wouldn't worry about it. Let the non-sequitur stand for what it is.
 
If any company strives for 99.5% success, that would be 500 problems per 100,000 units or cars in this case. A million cars made puts the number at 5000.
 
Originally Posted By: PimTac
If any company strives for 99.5% success, that would be 500 problems per 100,000 units or cars in this case. A million cars made puts the number at 5000.

This is equivalent to a sigma of approximately 2.5 and a Cpk of 0.83.
This is really not that great (actually it stinks).

World class component manufacturing strives for 6.0 sigma, Cpk 2.0, defect rate 3.4 parts per million.
Cpk 1.33 or greater (0.0032% defect rate) is generally required to qualify a process as capable for a feature.

Engines and other large assemblies are a little different but manufacturers strive to attain the 6-sigma quality levels in order to drive overall quality.
 
Originally Posted By: Imp4
Originally Posted By: PimTac
If any company strives for 99.5% success, that would be 500 problems per 100,000 units or cars in this case. A million cars made puts the number at 5000.

This is equivalent to a sigma of approximately 2.5 and a Cpk of 0.83.
This is really not that great (actually it stinks).

World class component manufacturing strives for 6.0 sigma, Cpk 2.0, defect rate 3.4 parts per million.
Cpk 1.33 or greater (0.0032% defect rate) is generally required to qualify a process as capable for a feature.

Engines and other large assemblies are a little different but manufacturers strive to attain the 6-sigma quality levels in order to drive overall quality.

So let's assume 3.4 failures per million on a low-ish part count engine, say 500 parts. And a moderate production run of 100,000 engines. 0.0000034 * 500 parts * 100,000 engines = 170 defective engines. And that's just engines alone, not even the whole powertrain, let alone the whole car. And it ignores failures from unlucky tolerance stack-up. So it's an insane understatement, likely 1-2 orders of magnitude lower than the defect rate for whole cars -- even assuming every part in the car is made by a process that achieves the optimistic defect rate you mentioned.

If we take the 0.0032% part defect rate, we end up with 1600 defective engines. Then you layer on the aforementioned adjustments...

Is that all roughly correct?
 
Originally Posted By: macarose
I have studied long-term reliability for nearly 20 years at this point.

Here's a study that offers a healthy synopsis. Hope you enjoy it.

http://www.dashboard-light.com/




I didn’t see a study at that link, only a car rating website with no acknowledgments.
 
Originally Posted By: d00df00d
Originally Posted By: Imp4
Originally Posted By: PimTac
If any company strives for 99.5% success, that would be 500 problems per 100,000 units or cars in this case. A million cars made puts the number at 5000.

This is equivalent to a sigma of approximately 2.5 and a Cpk of 0.83.
This is really not that great (actually it stinks).

World class component manufacturing strives for 6.0 sigma, Cpk 2.0, defect rate 3.4 parts per million.
Cpk 1.33 or greater (0.0032% defect rate) is generally required to qualify a process as capable for a feature.

Engines and other large assemblies are a little different but manufacturers strive to attain the 6-sigma quality levels in order to drive overall quality.

So let's assume 3.4 failures per million on a low-ish part count engine, say 500 parts. And a moderate production run of 100,000 engines. 0.0000034 * 500 parts * 100,000 engines = 170 defective engines. And that's just engines alone, not even the whole powertrain, let alone the whole car. And it ignores failures from unlucky tolerance stack-up. So it's an insane understatement, likely 1-2 orders of magnitude lower than the defect rate for whole cars -- even assuming every part in the car is made by a process that achieves the optimistic defect rate you mentioned.

If we take the 0.0032% part defect rate, we end up with 1600 defective engines. Then you layer on the aforementioned adjustments...

Is that all roughly correct?

Yes, I like where you're going with this, but honestly i didn't check your math. I'll trust you here.
But you lost me at 'unlucky tolerance stackup'.
There really should be no such beast if the design engineer is worth his salt and did a proper stackup analysis.

We also need to define 'defective'.
Does that mean my XJ 0331 cylinder head cracked and needs replaced?
Or does that mean a coolant passage is undersized and slightly restricts flow with no other material impact?

This is a question of defects versus defective, and these are important differences.
A product may have many defects – imperfections. But a product is not defective unless the defects prevent the product from functioning. If a product is not usable, it is considered defective.
So engines with 1600 defects definitely does NOT mean 1600 defective engines.

Other than that, well done!
 
Originally Posted By: PimTac
I didn’t see a study at that link, only a car rating website with no acknowledgments.

Then you didn't look into the database, read how data is collected, analyzed and reported and their (relatively) transparent efforts at highlighting what they are and are not measuring.

Here's another link for you so you have to do even less work to read up on their approach.
Dashboard Light FAQs Link

As always, analysis can only be as good as the data set so, YMMV.
 
Originally Posted By: d00df00d
Originally Posted By: E365
Because there’s only so much testing you can do. Even with 50 or 100 test cars running basically non-stop for maybe a year. That’s peanuts compared to 500,000 cars in consumers hands for a decade.

^ This.


Yep - NEVER EVER EVER underestimate the ability of an ignorant "populace" to screw things up.
 
Originally Posted By: FlyNavyP3
Originally Posted By: Imp4
Would you prefer any new car by any major manufacturer out there today?
Or would you prefer a brand new 1978 Chrysler Cordoba with Corinthian leather?

Pretty easy choice, I'd say.

To your original question, of course they do, just maybe not in the ways you think.
If you think the only way to do long term testing is to test something for 10 years, then the answer is an emphatic No because it's not fiscally responsible.

There are ways to perform accelerated life cycle testing by applying statistical methods to test protocols.
This probably isn't the straightforward answer many here are seeking but it is well understood and proven to work quite well.

That being said, there are always examples of OEMs screwing up, but don't be tempted to throw the baby out with the bathwater in these situations.

ALWAYS ask the proponent of the notion that OEMs don't do long term durability testing if they would like that shinly new '78 Cordoba over a new 2018 Malibu, Accord, etc...


Call me crazy, but if I could buy a brand new 1978/2018 Chrysler Córdoba with a 400 big block Chrysler and a 727. AC, cruise all the options. I could maintain it on a shoestring budget and fix absolutely everything on the entire car with common tools at my house. I'd take that over a 2018 Accord all day every day.


+1
laugh.gif
 
I'll never understand how the auto manufactures turn $300,000.00 worth of parts into a $50,000 piece of junk.
 
Originally Posted By: 14Accent
Originally Posted By: littleant
Originally Posted By: exranger06
Originally Posted By: littleant
. Ford explorer 2001-2010 4.0 3-4 timing chains in that engine that had issues. Ford 5r55s/w transmission servo problems steel piston s moving in aluminium bores. The two Ford issues i mentioned were resolved 9 years later when they stopped making them.

Ford 4.0 timing issues were 1997-2004. 1997 was the first year they made that engine, and they installed upgraded parts starting in 2005. I have a 2006 Ranger with that engine and I fully expect it to go at least 200k with no timing chain issues.
Unfortunately the timing chain issues did continue till 2010 but to a lesser degree. Ford did not update the rear cassette that requires the engine to be removed to repair.2006-2010 can go 200,000 miles. Maintenance is everything.


Those timing chain issues were blown out of proportion. Yes, they were an issue no doubt, but not the issue they're made out to be. The 4.0 SOHC can easily go 200k, even the early models. A friend of mine has a 2001 Sport Trac pushing 300k, all on regular oil every 3k miles. Does it rattle? Sure. It's been rattling for the last 150k miles. Still runs like a top.

Automakers can only test for so many variables. They use salt chambers to subject cars to 10 years worth of Midwest winter road conditions in a matter of days. The problem is they just don't have the time to let these things play out. You can't possibly plan for everything, even with computer simulations and exhaustive testing. One could argue it's more difficult to make a vehicle last 10 years than it is to make an airliner last for 25.
Agree many have. Owners need to do their part with service. You do not want to sludge this engine or any engine for that matter. Those hydraulic tensioner can plug up. No extended oil changes. Do not follow the tachometer at 6200 redline. Remember ford DID NOT back spec this engine to 5w20. It still Requires 5W30.
 
Originally Posted By: Imp4
Originally Posted By: d00df00d
Originally Posted By: Imp4
Originally Posted By: PimTac
If any company strives for 99.5% success, that would be 500 problems per 100,000 units or cars in this case. A million cars made puts the number at 5000.

This is equivalent to a sigma of approximately 2.5 and a Cpk of 0.83.
This is really not that great (actually it stinks).

World class component manufacturing strives for 6.0 sigma, Cpk 2.0, defect rate 3.4 parts per million.
Cpk 1.33 or greater (0.0032% defect rate) is generally required to qualify a process as capable for a feature.

Engines and other large assemblies are a little different but manufacturers strive to attain the 6-sigma quality levels in order to drive overall quality.

So let's assume 3.4 failures per million on a low-ish part count engine, say 500 parts. And a moderate production run of 100,000 engines. 0.0000034 * 500 parts * 100,000 engines = 170 defective engines. And that's just engines alone, not even the whole powertrain, let alone the whole car. And it ignores failures from unlucky tolerance stack-up. So it's an insane understatement, likely 1-2 orders of magnitude lower than the defect rate for whole cars -- even assuming every part in the car is made by a process that achieves the optimistic defect rate you mentioned.

If we take the 0.0032% part defect rate, we end up with 1600 defective engines. Then you layer on the aforementioned adjustments...

Is that all roughly correct?

Yes, I like where you're going with this, but honestly i didn't check your math. I'll trust you here.
But you lost me at 'unlucky tolerance stackup'.
There really should be no such beast if the design engineer is worth his salt and did a proper stackup analysis.

We also need to define 'defective'.
Does that mean my XJ 0331 cylinder head cracked and needs replaced?
Or does that mean a coolant passage is undersized and slightly restricts flow with no other material impact?

This is a question of defects versus defective, and these are important differences.
A product may have many defects – imperfections. But a product is not defective unless the defects prevent the product from functioning. If a product is not usable, it is considered defective.
So engines with 1600 defects definitely does NOT mean 1600 defective engines.

Other than that, well done!


Thanks. Makes sense.
cheers3.gif


By the way, I suspect my post may have come across as having an accusatory/confrontational tone. If so, my apologies -- it wasn't my intention. I was really just trying to illustrate the point that even with very low defect rates, it's still possible to generate more than enough problematic assemblies to stick in people's memories, such that it'd be completely understandable if some people ended up convinced that the whole product line was a sham.
 
Not in the least dood. I don’t understand the six sigma process much as it applies to the auto industry.

With the high influence of automation and robotics, parts and assembly have come a long way from the days when a guy would spend twenty years on the line installing bumpers. The Monday and Friday issues are much less now. It’s a fascinating subject and as already mentioned, today’s vehicles are far better in terms of reliability than the products of the 70’s and 80’s.
 
I think there is only so much testing the automakers can do. It is hard to have the latest technology in your production vehicle if you need to test it for 5 years first. Constantly changing requirements for emissions and mpg also influence this.
 
Originally Posted By: Imp4
As always, analysis can only be as good as the data set so, YMMV.


Sure. Sufficient, statistically valid, clean data set is a must. This data set seems to be doing alright, and the analysis is aligned with the field experience with taxi fleets in my past.
For reasons unknown to end user, Grand Marquis was consistently trumping Crown Victoria:

http://dashboard-light.com/vehicles/Mercury_Grand_Marquis.html
 
Status
Not open for further replies.

Similar threads

L
Oil Viscosity Trends and Hybrid Reliability: Is Spec Sheet Performance Misleading Real-World Use?
Replies
15
Views
1K
mikeduf
C
VRP Testing for the Long Haul- Cadillac 3.6
Replies
10
Views
964
Hohn
J
High Mileage Mazda 3, Engine clean, oil for long term
2 3
Replies
52
Views
2K
twX
wwillson
  • Sticky
Pennzoil - Low-speed pre-ignition (LSPI) Q&A - Answers
Replies
2
Views
1K
SubieRubyRoo
A
2007 Toyota Corolla with P0741
Replies
7
Views
1K
AKhan87
Back
Top Bottom