Testarossa dreaded differential issues and questions

Seven pages of nil

 

Hmmm wasn't that the lesson from the first couple posts in this thread?

 

Your correct

 

+1

+1

+1

1+1+1=3.

 

Right, so why waste everyones time with nonsense. You're . . . aw forget it.

 

But here 1+1+1=111

 

New math....

 

no
1 + 1 + 1 = 1
better is not possible

 

Guy's, does this mileage distribution look right for all existing TR's? I have set the MEAN around 25-30k miles with positive skew tapering down to almost nothing >100k miles. Note that although it's not obvious, this distribution will consider the garage and trailer queens with factory delivery miles.
Image Unavailable, Please Login

 

Impossible to confirm....

My '90 has seen over 9 years of storage over two time periods while I was working overseas - my mileage is currently 18,711. I consider this car to be an extremely low mileage example.

As I recall (?) the car had something like 16k on it when purchased in '97.

I am surprised to see the curve you depict. These cars need to be evaluated by their condition - too many cheaters out there. We've seen it here on fChat!

Ciao,
Vincenzo

 

Vince, according to the distribution, there are about 500-600 cars out there with your mileage. This does not imply you are average but instead what the Monte Carlo simulator pumped out based on the inputs and chosen distribution. I would agree, you are amongst the lower milage vehicles.

For clarity, the chart represents the amount of cars (frequency) TODAY at each mileage interval. Nothing to do with when those miles were racked up. As for cheaters, yeh I'm sure there are plenty of those but for now, I'm assuming the accuracy with 95% confidence intervals so there's always risk of being wrong.

 

Do you have the mileage of all the failures?

 

Yes but only with the following assumptions:

-Failures occur between 20,000 and 60,000 miles
-Normally distributed
-MEAN centered (40,000 miles)
-300 total diffs have been sold between all known suppliers
-50% were preventative maintenance.

Based on these assumptions, I ran another Monte Carlo simulation with 3 sigma limits each way. The distribution below is the result.

If my assumptions are wrong, then please correct me. Again, I'm not saying the failure is mileage dependent but I believe it's a good variable to start with while we learn more about the true root cause.

Next is to run a Censored Reliability Analysis to determine the probability of failure at each 10,000 mile interval.
Image Unavailable, Please Login

 

Did you include Ferrari as a supplier of new diffs. They were doing them before the others got involved. I remember Brian Crall speaking about the Daytonas having the same problem and he replacing some when he worked for Ferrari.

 

When you say failure ....does that mean actual failure 20-60? Or does failure include broken but caught it before full failure ? Is there a way to estimate ?

 

No I didn't. It's the first i hear about that. If you have any numbers, fire them over and I'll include it in.

I doubt anyone has ever pulled a cracked diff and not replaced it so we can safely assume that all diffs found to have cracks are considered failures and have already been replaced. What is difficult to assess are the amount of diffs that "could be" cracked without confirmation. If we can catch it before it grenades, that would be gold.

Also, the 20-60 is the approximate mileage (x1000) of confirmed failures

Anyway, the analysis is far from completion but the more numbers we get, the more precise our output will be and the closer we may get to the root cause especially if we discover that the higher mileage vehicles (>75k) have the same probability of failure as those with <10k i.e. extremely low probability. This must be explained somehow.

 

Btw I'm digging where this thread is headed now. To statistics!

 

MC generally spits out excellent results! Even seemingly "trash-in" still miraculously yields good results. I am a firm believer in the methodology!

Still though - disconnected odometers don't count.

Rgds,
Vincenzo

 

Also, just a thought. Were these carriers hand made originally, or by robots? The reason I ask is, could something like the day of the week come in to play in the carrier quality? Meaning, Monday= bad because people were hung over, Friday=bad cuz people were looking for the weekend, etc? It may sound crazy, but I remember reading in this very forum that someone went to the f factory to see his boxer getting made and the guys there talked about how much influence the guys wives and girlfriends had on the welding process, ie if they were fighting the dude didn't want to do a good job, and conversely, if he was getting some, he did a good job.

 

Im gonna say hand made. Could be wrong but back in 84 it was prolly a guy with a lathe etc. So the Wednesday perfect car would apply.

 

when I replaced my diff about 5 years ago you could also purchase an 'M' diff directly from ZF and when priced it came in around %30 higher than the easily available options. I am sure a diff from the ferrari dealer would have been closer to twice the price. Not sure if these are still available, but at that stage ZF was still doing runs of diffs.

 

The human factor is always something to consider, maybe not as much today via all the technology. But you bring up a good point. Which btw , looks like jay gt/4 was also correct. Their is product insurance available when manufacturing , or distributing parts.

Product Liability
Image Unavailable, Please Login

 

Anyone want to buy my original carrier out of a 1984 Boxer. It looks perfect, make me an offer I can't refuse.

 

When you say 300 diffs sold by know suppliers you are leaving one out ......Ferrari.

 

The diffs were made by ZF in Germany. I doubt very much they were hand made. CNC and automated welding has been around for a long time. The few I saw photos of, all the welds look sound. I still believe it is the stress reliving after?