Western Railroad Discussion > Motive power reliability. .

How do the various locomotives on the BNSF stack up for reliability? Here's a random day in January 2002 vs. a similar day this year. The measurement is mean time between failure (MBTF) measured in days for locomotives in the same class. Note that there are multiple classes of SD70MACs. First number is 2003, second number is 2002. All figures are rounded.

AC4400. . . 180, 105
SD70MAC. . .98, 108
SD70MAC (B).87, 91
SD60M. . .91, 107
SD60. . . 89, 93
SD75M. . .78, 63
SD40-2. . .85, 88
C44-9. . . 86, 100
C40-8. . .56, 71
GP60. . . 94, 73
GP60B. . .72, 62
GP60M. . .114, 71
B40-8W. . 62, 71
B40-8. . .43, 41

What does it all mean? Beats the heck out of me. Some classes of locomotives are doing much better than a year ago--the SD75M's, for example, habitual problem children. The big class of C44's has slipped 14% in mean-time failures, while the GP60M's are doing much better than a year ago. Both maintenance classes of SD70MACS are down somewhat. The small (3 unit) class of AC4400's is up significantly, too, no doubt as they settle into a routine pattern of service. The GP60s are a bit more reliable, too--perhaps because they are found more in regional secondary and local service than they were a year ago.

The system average for the fleet was around 87 days.

A possible thought on the GP60M and SD75M is that most of each fleet has been overhauled during the past 2 years. Also all 23 of the GP60B units were overhauled during 2002.

Ron

dickpalmer wrote:

> SD40-2. . .85, 88
> C44-9. . . 86, 100


Interesting indeed. Not much difference between these two groups, in 2002 at least. One dominated the rails in the 1980s, the other in the late 1990s and early 2000s. Is one good because it fails only as often as a group of engines that's 15 years younger, or is it bad because that group is probably in less-demanding service? Hmmmmmmm.

BobE

actually, bob, the first date is 2003--so the reliability as a whole of the C44 fleet has dipped while the SD40-2's remain pretty constant. for a while the SD40-2's were restricted to lower-priority trains where they'd be utilized in higher throttle settings more of the time, for fuel conservation. . . interesting that that hadn't effected their time between failures that much, eh?

If the first figures are indeed for 2003 then I would think they are not of much use. That is only about 45 days verses a years worth for 2002. Also they have not gone through a hot summer yet in 2003 and if failures of traction motors or cooling systems or whatever occur more often in hot weather that is is not yet reflected in the 2003 figures.

I came up the hill the other day running for 35 minutes in the 5 minute red zone. It was 8 degrees outside. The B40-8 finally cut back just as we topped over due to overheated TMs. In summer they would have been cooked much sooner and then worked longer when the cut back occurred sooner though at a lower power. Or we would have stalled.

Weather has a lot to do with failures. Extreme cold in the -20 or lower ranges can be as bad as heat. Most of the cold this year has been in the Northeast and south not the north. So failure rate can vary wildly year to year due to differences in climate from one year to the next.

dickpalmer wrote:

>Here's a random day in January 2002 vs. a >similar day this year."

AAK wrote:

> If the first figures are indeed for 2003 then I would think
> they are not of much use. That is only about 45 days verses a
> years worth for 2002.

Just thought I would re-add that he chose one day out of 03 and a day out of 02 not the whole year.

Chris M.
Southern Pacific Golden State Route Milepost 419.8
http://www.geocities.com/bnsffreak
Home of the Mighty Sampson of the Cimarron

Ah yes, I missed that "Random day in January".
This brings up a question. Just what are these figures? How are they derived?

If loco model X has a MTBF of 100 days that means that half of the model X locos that have failed have ran more than 100 days since their previous failure and half have ran less than 100 days since their previous failure? If that is correct, and I am not sure it is... I am getting confused here, it is too late at night for my brain to work like this. Are they counting only the locos that actually have failed or all of the locos of that model? Some of which I assume have never failed since built and thus currently have an infinite TBF. If they are counting ONLY those locos that have failed, then how do you do that? Surely the same unit is not failing every 100 days? I am missing something here. It seems it would have to be a running MTBF of the model as a class but over what period of time?

If I have an average of three units on my trains every day should I expect a failure roughly every 30 days based on a MTBF of 100 days? If so I must be mighty lucky because I am not experiencing that.

I find it hard to believe that every C44 statistically will be expected to fail every 100 days. Or I also find it hard to believe that a single C44 that has failed also failed approximately 100 days ago.

So I am missing something at this late hour. How are these MTBF figures derived?

Al,
these figures are for a day in January, 2003 vs. a day during the same week of January, 2002, so they should be pretty good comparisons.

AAK wrote:
> I find it hard to believe that every C44 statistically will
> be expected to fail every 100 days. Or I also find it hard to
> believe that a single C44 that has failed also failed
> approximately 100 days ago.
>
> So I am missing something at this late hour. How are these
> MTBF figures derived?
>

I'm not privy to the methodology, but I believe the figures are taken based on models and fleets as a whole--some of those C44's will fail three times in 100 days, some once in 300 days. I imagine the numbers are averaged performance of the fleet and classes within the fleet. I intentionally left out the models on the railroad with a small number of locomotives (with the exception of the AC4400s because)with one or a dozen locmotives two failures will significantly skew the MTBF numbers.

I don't understand the MTBF of locos. It can't be a daily thing. With 1400 C44s, one day 20 C44s might fail and the next 1 C44 fails. On a -40 degree day perhaps 200 will fail. It has to be a running total of some period. For GP60Ms of which there are less than 100, thank god, the class must go for days or even weeks without a failure. It seems to me that the period of measurement DOES matter.

I understand the TBF of a single unit. I understand that MTBF for a group of units means half fail sooner and half later. But WHAT UNITS? All 1400 of the C44s or only those that failed within the period? If you consider all 1400 C44s must all of them have failed? Lets see. If 1000 have never failed in their life, if 200 failed every 50 days, if 200 failed every 150 days, what is the MTBF? I doubt that is the way it is. If so I'd junk those 400 C44s :-)

I find it amazing that on average each and every C44 fails every 100 days? Maybe who ever wrote the program to keep track of this inadvertently counts the days out of service for the 92 day inspections :-)

Admittedly my pickup truck is not operated like a loco but I use it almost every day and it is 27 years old. The only "failures" I can recall in those 27 years are an alternator, a front 4WD locking hub, and an ignition module. You'd think they could build a loco that will run for at least a year?

I admit that statistics are beyond me. I have heard it said that if you buy two lottery tickets instead of one you have exactly the same chance of winning. I don't believe it. You have twice as good a chance of winning assuming you don't pick the same number on both tickets. Carried to extreme I could buy 64 million tickets (or whatever the possible number combos happens to be) and I would have a 100% chance of winning. If the pot is above $64M I come out ahead IF no one else has a winning number.

They say you have a 1:xxxx chance of being killed in a car wreck and a 1:yyyy chance of dieing in an airplane crash and it doesn't matter how often you drive or fly. Bull! If I never get on an airplane then I have a 0% chance of being killed in an air crash.
I find it hard to believe that if I drive every day I don't have a greater chance of being in an accident then if I only drove once a month. OK, they mean on each particular trip you have xx chance of being killed but it still seems to me their would be cummulative effects.

Like I said, I don't understand statistics and "means" are especially wierd. A mean is not an average and it is hard to comprehend just what mean means.

OK, as I understand this, if we have 1400 C44s and the MTBF is approx 100 days that means about 14 fail every day? I do not recall a single C44 failure on my trains since the C44s came out. Guess I am very lucky? I have had a few dead ones in my consists but none have failed while I was running them.

Of course I'd define a failure as one that will no longer load or will no longer load to its full potential due to some mechanical/electrical problem not just wheel slipping. They'd all fail every day if that were so :-)

But I wonder what BNSF calls a failure? If both ditch lights burn out or the toilet plugs up and the unit has to be switched to trail position is that a failure? If it quits loading and the train stalls but rebooting the computer brings it back on line 100% is that a failure? Or is a failure only when a loco has to be taken out of service and shopped?