I don't know about specific locomotive numbers, but any locomotive I own (or most), they'll get the number series correct for that railroad. The Erie Lackawanna RS3's for instance, are numbered in the 1000 series; the GP9's were numbered in the 1200 series, and those numbers were generally correct for the numbering series.
As for cars, that's a different story, as many of the older models are incorrect for the details vs. the number series. I've checked the numbers out on various railroad picture websites.
Back when prototype photos and information were not as easily had as they are today, often manufacturers would use the number of the photos of a car or locomotive pictured in the Car Builder's Cyclopedia or Locomotive Builder's Cyclopedia. Modelers of that time used the same resource. Not just numbers but paint schemes: I recall some modelers and manufacturers were led astray by a photo of a certain steam locomotive that had been specially painted up just for its official "birthday" photo at the shop and was repainted before delivery to the actual customer.
Dave Nelson
As mentioned, some models are based on real engines with specific details. For example, Athearn offered the SDP40F with road numbers used by the Amtrak San Francisco Zephyr which had ice breakers on the roof. Numbers were chosen from two groups, probably from photo's.
Athearn also used road numbers for D&RGW GP40-2's which had some details like the exhaust silencer. For the D&RGW SD45's, they used road numbers consistant with small and large herald paint schemes and even match the type of brake cylinders mounted on the trucks for that road number (low mount or high mount). However, they also chose a road number from the recent run of D&RGW GP40-2's that duplicates a road number Atlas offered on their D&RGW GP40-2's. Which means I'll have to renumber one of them, probably the Atlas model. I bought a decal sheet recently that should help in that matter - all road numbers for Atlas, Athearn and ScaleTrains.
OTOH, When Proto 2000 did GP30's they chose not the best road numbers on their Rio Grande phase 1 models. For example #3005 was written off in a wreck early in it's career and #3010 didn't have orange paint above the windshields but the model did.
Rio Grande. The Action Road - Focus 1977-1983
While I certainly can't speak for all manufacturers, I have been part of a few projects with the NHRHTA. When we get questions from manufacturers on engine numbers, it'll be like, "Are there any numbers we should not do?" Those are most important. For example, the NH had 20 engines of a particular type, but only 4 of them had radios. They want to make sure they don't make a radio one. Another example is that the NH had 10 engines but three of them were damaged and were stored out of service for years before being fixed up. They don't want to make those numbers.
Another factor is when a company makes a new version of an older model made by someone else. They generally don't want to duplicate these numbers, so they'll ask us for all the road numbers that have already been made by everyone else.
After that, your guess is as good as mine why they pick the numbers they do. It's not always based on the pictures they have, either. We've supplied photos of engines to manufacturers and they don't pick any of those numbers. Instead they use some other numbers entirely. Maybe they use a random number generator?
NittanyLion ...followed by "did we produce this number in a previous run, if yes, pick a different number."
...followed by "did we produce this number in a previous run, if yes, pick a different number."
And that is sometimes where they get it wrong. Case in point: Bachmann's "Russian Decapod". The Santa Fe had three of them, all inherited from the Kansas City, Mexico, and Orient. The Santa Fe road numbers were 2554-2556 ("2554 class"). In the first run, Bachmann used a correct number, 2554. For the second run, they still would have had the choice between 2555 and 2556, but for some mysterious reason they chose 2552 (Santa Fe 2552 is a 2-8-0 Consolidation, not a 2-10-0 Decapod). For the current run, they got it even "wronger"; it now carries a 24XX number.
JW
I know one manufacturer chooses to run the number of the prototype locomotive they have the most pictures of. That way they can be pretty sure that all the details are right for that specific locomotive.
Other than that, not sure.
I always wondered how Life-Like chose the numbers 703, 707, and 722 for their Zebra-Striped Santa Fe Geeps.
-Kevin
Living the dream.
Aside from specific units that are required to have a specific number because of the paint scheme, there's probably no universal answer. I'd wager it goes down this order and not much further: random selection from research photos followed by "did we produce this number in a previous run, if yes, pick a different number."
I have to ask, how do model railway manufactures determine what road numbers should be done/included for a run of locomotives. Do they decide by the popularity of the number by railfans, or another method?I personally like locomotives with matching numbers, or having two same numbers, numbers thath have aparticular menaing to me, or with simular or rythming such as CN #3034, a Intermountain HO scale CN Tier 4 gevo I own.