Part numbering conventions - what's the best practice in aerospace?

[chrebmo]

Hi everyone,

I was having a discussion with an aerospace partner recently about part numbering conventions in PDM systems, and it got me thinking - how do different organizations approach this challenge?
One approach I've seen is a hierarchical naming scheme that embeds mission/project context directly in the filename structure, something like:MISSION_PARTNO_NAME_REV.SUBREV
For example: PROJ-A_4301-01-00_ReactionWheels_01 or PROJ-A_32010200_PCBBottom_01.01

Where the 8-digit part numbers create a hierarchy (4000s = platform, 3200s = power, etc.) and COTS parts get handled separately.

It seems like aerospace has some unique challenges compared to other engineering sectors - the mission-critical nature, long product lifecycles, and traceability requirements. I'm wondering if this drives different approaches to part numbering than you'd see in automotive or consumer products.

What approaches have you found work best for your organizations?

Some follow-up thoughts: Do you embed project identifiers directly in part numbers, or rely on the PLM system to manage those relationships? How do you handle COTS integration? Any thoughts on hierarchical vs flat numbering schemes?

Would love to hear about your experiences, especially if you've been through PLM migrations or evolved your naming conventions over time.

Thanks!

 

[Ng2020]

"Do you project identifiers directly in part numbers" yes but as a 1-3 alphanumeric char code.
Some manufacturers use a WBS based part numbering scheme, sothe part number gives some indication of function and locatio . that scheme is my preference.

"handle COTS integration" a whole (dull) textbook could be written on this. assign cots a part number only where specific acceptance tests/inspections are applied (e.g to ascertain configuration and quality). if the cots manufacturer holds the appropriate design approvals for the item (eg tso) then this is not required
Managment of obsolescence is another issue to watch for.

 

[TugboatEng]

Not a fan or alphanumeric. It only introduces additional sources for errors. Stick. To numbers only, the most universal language. Don't do anything sequentially. A fat fingered character should produce a wildly incorrect part. Go ahead with identifiers at the beginning but in the instance of a nut and bolt the bolt should not be ***-****-1000 while the nut is ***-****-1001. No similar part should be 1000 or 1001. Use random number generators.

 

[verymadmac]

Is it for STC's, mods or new airframes who's the end user. Depending on just what approval system you are operating under will likely change how you system will likely function.

I worked for a helicopter mod shop who's STC packages defined drawing revisions had approvals in multiply countrys. Which meant the drawing stacks were frozen for any thing except a really good reason (some thing fell off or we can't get all the bits to make new ones etc). For common parts they would use PN's unique to each package, they would build one batch of parts and then reinspect them to change them to the part number required to kit out the appropriate STC kits, this however could get really complicated really quick if you had to make changes to a part.

I have worked for an GA OEM that only used an Airframe specific part numbering, if you wanted to find a part that should be a standard part you would either go for a walk around the assembly hall to sight the offending part or text search several IPC's. It meant things like a duct bulk head fitting didn't have logical PN's the 1" duct fitting had a 71 ATA based PN while the 2" had 23 ATA based PN etc. When I was working for an Airline designings repairs, finding parts that will do the jobs for AOG was always interesting as the stores system wasn't really text searchable (they rarely used the same naming convention from part to part) and with half a dozen OEM's meant often the part you wanted was often unfindable because it was hidden behind a numeric number where as alteast the Boeing parts yielded to a BACXXX search. I would note not to fold to suggestions by stores about not making part numbers too long because they don't want to type the extra part numbers, you always seem to go though drawing numbers faster than expected.

Personally my favorite is unique airfame parts and assembly's get ATA based number roughly looking something like AAB.CCCC.XXX-YY
AA is the OEM designator, B model, CCCC the ata number, XXX unique number, YY part configurations. while for common parts (AKA standard parts) found in numerous locations and multiply airframe models etc (ducting parts, shear clips, etc etc) I would rip off the Boeing hardware designation system which runs something like this: BACR15CE5D8 is a shear head rivet, the number breaks down as BAC company designation, R is the first letter of descriptor the 15 makes it unique rivets, the CE is the specfic rivet type and the remaining numbers ID the specfic part via the table drawing (table drawings seem soo under rated these days but save so much time / offer significant flexibility) .

I hope that all makes sense.

 

[chrebmo]

Thanks everyone for the detailed responses - really fascinating to see how different approaches have evolved!

A few themes are jumping out at me that has me needing to dig deeper into,

Hierarchical vs Flat Trade-offs:
@TugboatEng - Your point about avoiding sequential numbering (bolt shouldn't be 1000, nut 1001) makes sense for error prevention. But @Ng2020 and @verymadmac seem to favor more structured approaches (WBS-based, ATA-based). How do you weigh the benefits of logical structure for findability against the risk of human error from similar numbers? Is it inevitable that the system is as strong as it's weakest (manual) link in the data toolchain?

Searchability Optimization:
@verymadmac - Your example of needing to "walk around the assembly hall" to find parts really highlights the search problem. It sounds like teams often optimize their numbering schemes to compensate for poor search capabilities in their tools. Do you find that part numbers end up carrying more information than they should because the systems can't search descriptions or metadata effectively?

Custom vs Standard Parts:
@Ng2020 - The distinction you made about COTS parts with appropriate approvals (like TSO) is interesting. How do you handle the boundary between custom and standard parts in your numbering? Do you find teams treating functionally identical parts differently just because one is custom and one is COTS?

Obsolescence Management:
You mentioned this as "another issue to watch for" - how do different numbering approaches help or hinder when parts become obsolete? Do you find that hierarchical systems make it easier to identify replacement candidates, or does it create more complexity?

I'm particularly curious whether these challenges are driving teams to work around their tools rather than with them. Are the numbering conventions a response to system limitations, or do they reflect genuine engineering logic?

Maybe there are other problematics in data integrity in terms of part numbering, i.e. descriptions of parts also needing to have protocols. Or simply the compliance aspect of needing to cross reference conventions across different standards.

What really is the core pain when it comes to the question of part numbering, and otherwise what would be the dream? Caveat that I look at this from a software perspective, I am not myself an aerospace engineer (probably evident without having to say ) but have been building software for engineers for over a decade. I like to geek out on these technical questions, and connect the dots of what software can otherwise do.

Once again thanks for your responses.

 

[ewh]

https://www.eng-tips.com/threads/drawing-part-number-help.136668/ is an older thread with some good references.

edit: typo

 

[CWB1]

5-6 alphanumeric, random digits is the norm. "Smart" part numbers are among the dumbest, most costly ideas pushed by MBAs in recent decades.

 

[verymadmac]

chrebmo How do you weigh the benefits of logical structure for findability against the risk of human error from similar number​

Its a non starter, the alternative is a meaningless number. Just look at our existing MS, NAS parts, sure there may be mistakes in the parts call outs, but because it uses a standard number nomenclature we all can pretty quickly pick up if its the wrong alloy, size, head type by a quick glance at the drawing.

Do you find that part numbers end up carrying more information than they should because the systems can't search descriptions or metadata effectively

This is a controlled data verse uncontrolled data problem. Meta data and descriptions need to be utterly consistent to be reliable which is a man power problem. For the small team with limited resources this becomes a why do three things well when doing 2 things well works with less labour, its a simplicate and add lightness solution. But as ways rarely does a single solution suit all situations.

As a note of COTS parts I would suggest considering placing your own part number on to the COTS part. Why, well n the case of an agricultural aircraft manufacture was they were buying plastic chopping boards at the local large chain store for canopy slides and someone used the last one in stock followed by a bit of panic about just which ones were they buying (they had brought up several years supply at once and the details of just what they were buy was lost).

 

[MintJulep]

If there was a single magic "best practice" that worked for everything, then everyone would use it. There isn't, and that's ok. Do what works for your unique situation.

There are plenty of cases where a degree of "smartness" in the part number makes sense and works fine. E.g. bolts, where a "dash number" identifies the length, hydraulic fittings where the dash number carries the nominal diameter.

In general, the more rules you try to apply to a numbering system the more likely a future case breaking the system becomes. So, favor flexibility over a rigid set of rules.

If you can easily foresee a condition that would break a "rule", then it's a bad rule and should not be used.

Embedding project? Instantly breaks if you get a follow-on project, or a project for the same product but a different customer. Instantly breaks as soon as a part designed for project A would work just fine for project D.

Embedding hierarchy? Instantly breaks as soon as a part would work just fine in more than one location in the hierarchy.

For searchability, drawing titles are more important than drawing numbers in many cases. If you know "exactly" what part you're interested in, then search by drawing number. If you're looking for "parts similar to something", then searching by title should get you there.

MIL-STD-100 Chpt 300 offers pretty good guidance on drawing titles.

A negative example of bad drawing titles is a well known Canadian manufacturer where any given project will have 700 drawings titled "bent plate".

 

[TugboatEng]

In the modern days of internet, part number databases are instantly accessible at all times from anywhere. There is absolutely zero reason to bake information into the part number itself. It simply needs to link itself to a document that has the information you need.

Numbers are sufficient. Letters only add complexity. Try ordering port adapters from Parker. They love using O's in otherwise numeric chains that you would assume are 0's. Very frustrating.

Caterpillar had forever used 1x-1111 format for parts where 1 is a number and x is a letter. They have moved to a 111-1111 format. It's a little harder to memorize part numbers but who needs to do that?

And I still find myself doing a lot of business over the phone. B's and V's or N's and M's are no fun to work with.

 

[IRstuff]

We used strictly 8-digit numbering with barcodes on the physical product; anything that removes the human element from part identification is a good thing.

 

[RoarkS]

part numbers with intelligence fail.
I started using a "born on date" part number. For instance this post for me 080120250055_Part number memo_rev A released

 

[MintJulep]

born on date" part number

Doesn't ensure a unique number if there's more than 1 creator.

 

[drawoh]

chrebmo,​

I am a firm believer in dumb numbering systems. Take the next number off a list or out of a database. Any intelligence in your system should be a database field, which you can edit when your requirements change.

Think hard about part naming conventions.

SCR CAP BTN HD HX SCK STL ZN PL M6X1X20

In the PDM system I have set up, I can sort our parts library by description. If I have a bunch of fasteners named like the one above, they sort by screw, by type of screw, head of screw, socket/actuation and material.

One of my DFMA rules is that any part called BRACKET probably can be removed from your design, with their functionality incorporated into existing parts. I worked for a Canadian manufacturer that systematically named just about each and everyone of their parts BRKT... Oh well.

 

[RoarkS]

Doesn't ensure a unique number if there's more than 1 creator.

LOL. I wrote in the manual that repeated numbers are deconflicted by creation date of the first part feature in the design tree that goes down to seconds. I figured the probability of that was high enough it was acceptable.

I just find having to keep a part number master list to be an unneeded step. That was the simplest universal way I came up with. But to my own argument there is "intelligence" to that part number and of course there is a flaw in it!

 

[rb1957]

"We used strictly 8-digit numbering with barcodes on the physical product; anything that removes the human element from part identification is a good thing."

who decides what numbers to use ? a random number generator ?? That sounds ... awful ?

I'm with MJ (post 9) ... there is no best ... use a system, any system. IDK your business but in mine (aircraft modification) we had a system of drawing numbers (so 520xxx was a structural component of some sort) then we went to ATA paragraphs (so a 531000xxx is a structural part in the fwd fuselage) ... all well and good until you reuse a fwd fuse part in the rear fuse ... so then you revert to 530000xxx and lose some information but gain some flexiblity at the cost of some information.

"Doesn't ensure a unique number if there's more than 1 creator." ... more than one person born on the same day ? add universal time (to the second, 1/10th of a second) ?? smile, just being a jerk !

 

[CWB1]

who decides what numbers to use ? a random number generator ?? That sounds ... awful ?

That's the default for how most PLM/PDM systems assign p/ns.

"Smart" p/ns cause a ton of issues on the shop floor. They're long, and once you get beyond 5-6 digits humans struggle to accurately read, write, and remember them, or even distinguish between similar p/ns. That extra length causes the wrong parts to be ordered, misplaced in the wrong warehouse/location, etc. "Smart" numbers also cause mistakes bc rather than being forced to use a database to identify parts, workers tend to rely on memory of what the "smart" bits mean (and guess wrong).

If you dont have PLM/PDM or ERP, a simple Access/other database can be setup to manage p/ns, search feature data, and feed non-engineering users.

Coding hierarchy into p/ns assumes 1. a single use of the part, 2. that both the company and product's hierarchy is consistent product-product and over time, and 3. that clear boundaries exist between systems. Over time those assumptions fail, sometimes quickly, sometimes it takes a few years but ultimately you're stuck with cobbled complexity that never had a practical use.

COTS parts without an internal p/n are uncontrolled - you're allowing use of any manufacturer's vaguely similar part, which is why internal p/ns are the norm.