| BMW M3 (E46 M3) Forum | Post Reply | Preferences | Search | New Archives | Old Archives | Boards FAQ | Notify the Admins |
|
| Author: |
| ||||
| Subject: |
| ||||
| Message: | With the discussion over the past several days and the pending updates of 2002 production failures (http://forums.roadfly.org/bmw/forums/e46m3/forum.php?postid=633884&page=2), as a 4/02 owner, I became motivated to examine the thus far healthy S54's posted on roadfly to determine if the healthy engines could be used to predict what production dates or mileage thresholds exist for further failures. (Sorry Allen, i couldn't wait :-) All of the charts discarded the singlet data, but these were entered as an unplotted aggregate in the tables. (SD=standard deviation, CV=coefficient of variation, n= number of cars). Linked URLs are at the bottom but they're here too: The chart is at http://home.attbi.com/~biodan/workingChartE.htm raw data is at http://home.attbi.com/~biodan/workingEnginesT.htm I also plotted Jason's data: http://home.attbi.com/~biodan/failedChartE.htm Several observations arose: 1. Obviously there is an uneven sampling in both datasets. One would hope that reporting of failed engines would improve with time as publicity on the net becomes more well known. 2. For both datasets, where n is small, the coefficient of variation is large (duh) 3. The mean and median (~8000miles) of the healthy 2001 engines matches roughly the corresponding values of the failed engines (but we only have healthy 2001 engines for the Sep, Nov, & Dec months where n > 1). However, as Allen Schiano suggested earlier, there may be a multi-modal distribution of the failed engines. This can be seen by comparing the aggregate of 12 engines from Mar-01 to Sep-01 (mean/median=8020/8800) with the Oct, Nov & Dec values (medians=5350,6335,7400 where n=9,27,12) which are lower but _ascending_ and 'approaching' the mean/median of the Oct-Dec engines. (i know the differences are within the SD ranges). Since the range of the Oct data is large (CV=.78) and the median substantially lower than the mean (5350 vs 7972), the lower 'failed chart' plots the data with the 24K mile outlier discarded. 2 other datapoints were discarded (500 miles & 18K because they were >1 SD from the mean) and the resulting plot looks 'cleaner' but the SAME pattern of the top chart emerges; i.e. the Oct-Dec failures happened at lower mileages than the Mar-Sept 01 failures esp if one compares the aggregate Mar-Sept data to the individual Oct, Nov & Dec means & medians (8700/9300 vs 5969/5275, 6370/6267 and 7310/7400 respectively). Healthy engines:  Failed engines: (lower graph)  4. The chart of the healthy engines suggests that the 2nd failure mode (Oct-Dec 01) should have been reported at similar frequencies by now for the Jan and soon Feb 02 production dates (mean/median= 6944/7138 and 6284/6300) since these values are within the cursed months' range. It would appear that this failure mode has been remedied since 02 cars from Jan to Apr 02 have values generally >5K yet we haven't seen the same frequency of 'low' mileage failures AKA a lethal post-natal 'birth defect'. Jason's pending 5 reports of 02 production failures will be important to tell if mode 2 failures have really been remedied. However, for mode 1 failures, we don't have enough healthy cars with high enough mileage- Jan 02 has a mean/median of 6944/7138 - well below the 8-9K of the mode 1 engines. The bottom line is that we need another month of drivers reports when the Jan values will have matched or exceeded these levels of the first failure mode engines (we seem to be averaging 1k miles/month). Of course, if we see a rash of relatively low mileage failures, then mode 2 failures haven't been remedied (of the pending 5 reports, one had 6K and the other ~7.5K miles; jason@nyroc.rr.com & the german fellow at the Gladbeck racecourse). Its been proposed on this board that mode 1 failures represent a 'genetic pre-disposition' to failure since they occured before the bearing 'fix' months _and_ they tend to occur at higher mileages (esp the 18K & 24K 'outliers'; the exception being the 500 mile in the 4/01 set). Oddlycalm posted yesterday that the S54 design might be susceptible to defacto oil starvation since most failed engines show heat damage, e.g. dirt clogging a too narrow passage. We should also recall that M3 production increased almost 3x around Oct 01 (maybe it was Nov?). So the mode 1 failures would represent a larger fraction of the monthly production - yet they generally occured at higher mileages. The only problem is that we don't know if the reporting frequency of the mode 1 failures is the same as the mode 2 failures. working S54's failed S54's table of working S54s ----- With ~6k miles on my 4/02, i'm going to keep this car for less than the warranty period unless BMW extends the warranty. I really don't care if BMW is known to have continued replacing the nikasil V8's after warranty- there is no guarantee that this will continue or be extended to the S54. Further, since i suspect the mode 1 mechanism is unresolved, i will change the oil and air filter every 3-4k miles (or before any DE, whichever comes first) to prevent any genetic predisposition from expressing itself. Even so, i have to think hard about whether to put more after-market mods into a car that i might sell soon esp if the failures continue at any rate. Despite all this, i love driving my M3 and i'm glad that BMW brought over the real deal (vs the e36M3). My hope is that BMW will not dumb-down the next version (M4) because of our legitimate concerns. These concerns also need to be addressed in a forthright and fair manner. | ||||
© 1997 -
2010
Jump Internet Inc. All rights reserved.
