We built it within 0.01 tolerances.

https://fxtwitter.com/ahmedshubber25/status/1826040036607025662

50 comments

This guy is fucking incredible. "Founder" in bio ass motherfucker. In another thread recently he suggested there is no need for what he refers to as "2d's" (engineering prints) and that everyone should just work off of 3D models. This guy supposedly runs an engineering company and has never set foot on a shop floor, engineered anything, or presumably even turned a screwdriver in his life. Every engineer and machinist on Twitter is shitting on him and he's doing that "they hated him because he told the truth" thing. Then you have the handful of dumbasses one-upping him, who think CAM software can be replaced with ChatGPT.
A case study in why Americans will never return to the Moon.
- Imagine being able to have a computer that runs a 3D model constantly that doesn't get shit all over it within the first couple of hours. The reason you have prints is so you can, you know, print it out!
  
  Naa. That can be done. I use a computer on the shop floor. I do CAD/CAM on the shop floor. Yeah, the computer gets fouled up with oil and shit, but fuck it. If it dies, buy a new one. Operating expense. We also have tablets at each workbench. Nothing is stopping us from having an app for viewing 3D models on there. Some people in that thread were like "oh it wouldn't make sense to buy a CAD seat for every machinist." No, that's not it either. Just pirate that shit. Even if it were free, it would be a stupid idea.
  A print is nice, because it can be marked up. I can pull out a pencil and add notes. I can highlight dimensions on it. I can walk it over to an engineer and POINT at a feature (or even, when proper characteristic maps are made, just reference a feature number in a phone call or an email).
  Prints have a design language to them which allows you to express fully constrained geometric designs on a napkin if you need to. Dimensions, radii, diameters, angles, datums, positions, projections, sections, GD&T. None of this is obvious in a 3D model. You don't know what the driving dimensions are, what can be inferred from other dimensions, if it is a coincidence or a requirement that two features line up, etc.
  There are rules to engineering prints. You never dimension a feature when it can be inferred from other dimensioned features. Dimensions are added until nothing is ambiguous, without constraining any feature more than once. It is formulaic. This is fundamental to how manufacturing processes are developed and how inspection plans are generated. It is the fundamental way of proving the geometry of a part has been fully considered. And then once you have it, you can send it to any shop in the WORLD and they will understand exactly what you want. They will be able to inspect parts and know exactly what to accept / reject.
  I do a lot of one-off shit in the shop, making fixtures and custom inspection gages. A lot of times the programmers will skip making a blueprint for things like this and just send me a STEP file, which is FINE. I can select the geometry I need from the STEP file, turn it into a toolpath, and cut it, but it tells me NOTHING about what tolerances I'm aiming for, if it is better to go over/under size, which dimensions are the most important, or what datums I should be referencing when I fixture the part in the machine.
  
  Prints have a design language to them which allows you to express fully constrained geometric designs on a napkin if you need to. Dimensions, radii, diameters, angles, datums, positions, projections, sections, GD&T. None of this is obvious in a 3D model. You don't know what the driving dimensions are, what can be inferred from other dimensions, if it is a coincidence or a requirement that two features line up, etc.
  This is so critical. In architecture and structural engineering, you can add to this that you don't actually know a lot of the real dimensions - you're laying out the important ones from the structural grid or from survey points, and whatever is left doesn't matter.
  Even markup, at least in a design environment, can be done in 3D (or at least on a computer), but the communication of constraints, that is, what dimensions are important and which are irrelevant or unknowable has not yet been developed in 3D models, and I suspect it will be some time before any useful language for that purpose stabilizes.
  
  Even markup, at least in a design environment, can be done in 3D (or at least on a computer), but the communication of constraints, that is, what dimensions are important and which are irrelevant or unknowable has not yet been developed in 3D models, and I suspect it will be some time before any useful language for that purpose stabilizes.
  This is a thing which exists in a lot of CAD packages, and it is reasonable for an engineering firm and a contractor to use the same CAD package as part of a contract. Some of the more fancy (expensive) stuff can even check for interference in assemblies throughout tolerance ranges. But still, if you cannot turn your design into an ASME / ISO drawing I have severe doubts in your competence.
  
  Yeah, I know some of that exists; it never showed up in my world. We never modeled buildings with enough detail to make it really useful. Occasionally someone would get excited about Tekla and we'd spend some time trying to do shop drawing reviews in 3D and then go back to PDF. What I meant was that it isn't yet a standard thing that is understood by any technician in the industry, it's proprietary software that is subject to change with every release.
  
  I'm fairly sure some things(like impeller or fan blade geometry) can't be meaningfully communicated on a drawing and only through a model. I expect some of that to be more common as AM allows manufacture of internal features designed for/around fluid flow.
  
  I never thought about any of that before but it totally makes sense.
  When learning skills that are depicted visually, I've found it's better to get a book that's old enough to have illustrations or diagrams if possible. 2d line drawings in black and white, or with judicious use of 1 or 2 other colors. A skilled illustrator can show exactly what you need to see but photographs necessarily have various other stuff in view. Illustrations can clearly depict motion, hidden parts, or changes over time in ways that photos can't hope to.
  And that's without the benefit of assuming a lot of knowledge on the part of the reader the way you can do within a profession like you are describing.
  
  Can you come to my job and explain how tolerances work to these engineers?
  
  You can say 'it's an operating expense' but that isn't the problem. The problem is that the shit breaks when you need it to work, especially if you are welding or sizing or just have machinists or welders being fucking idiots on the floor as they often are (because it is hard to not be a fucking idiot in a place that you spend 40-60 hours a week). Let alone fumes, grease, and dust. It's an idea that sounds good but just doesn't work imo. Paper just works better, and I'm not even a geezer.
  You can have a computer on the shop floor (I recommend at least three within walking distance for layout design), but having a computer (or even screen) in every booth to run CAD through is dumb as hell. I knew this from working in a shop and I know it from experience when I was overruled by management because they got it into their heads from some networking guru that every booth needed a screen, half of which now no longer work after three months and aren't going to be replaced. I still haven't had the balls to say I told you so, but it's annoying as shit to see stuff coming from a mile away. Maybe it works for the Germans and Finns, but our shops just aren't nearly as clean as theirs are.
  Sometimes it is nice to see the 3-D, but the print should tell you everything you need to know, and if it doesn't then you have to talk to an engineer about it because someone fucked up. And if they aren't making prints, it is totally up to the machinist's discretion. Idk, I can't even imagine not making a fully speced print for a design, it's such an anathema to the kinds of products I work with. Even a small run has to have tolerances or the welders will literally have to feel out good parts from bad parts, which they do not appreciate.
  
  You never dimension a feature when it can be inferred from other dimensioned features. Dimensions are added until nothing is ambiguous, without constraining any feature more than once.
  Tbf engineering 3d models should also be this. They're usually not, but
  
  constraining a CAD model and communicating tolerances like datum, or runout vs cylindricity are very different things.
  
  I recently fixed up a colleagues cad model and he left everything blue :( it sucked when one dimension needed changing.
- The guys in the shop know way more about what they're doing than I do. I feel like I'm doing nothing in solidworks, but shuttling back and forth between the workshop and computer. Apparently insurance liked to have drawings though
- Dumb question, but can't you go from a CAD file direct to gcode in fusion for CNCs?
  
  A Solid Model (CAD) file (while very useful) doesn't contain all the information you need to convey what a good part is. There are an innumerable amount of things you want to convey to the manufacturer what you are looking for, from masking parts of a surface, to how parts need to be packaged, to which of the dimensions are the most important and how you should measure them (GDT etc).
  There are services out there that you just upload a solidpart file and they give you a price. They are highly automated, cheap and quick. The parts are not high quality. Often their standard tool sets cant machine a feature. They communicate that and you have to make other arrangements, either alter your design, or plan on doing some post processing. I was actually supposed to place an order with one today, but had some problems with capability so I am altering the design to accommodate those limitations.
  Its sad, but good machinists are a dying breed.
  
  They are out there, but they are usually stuck running basic processes that are essential to products but can't be trusted with incoming hires to maintain quality. Ironically, incoming hires usually work on more bespoke projects because they aren't as essential to the business being successful (unless the process has been highly automated, and then no one gets any real machining experience). Shops are weird like that.
  
  I ask this because if I need something machined, I just send an onshape link to a friend with a shaepoko or run a fabworks order, neither of which need 2ds. I can't do half depth through fabworks because it's just lasered, although I've had significantly fewer issues with fabworks than with shaepoko guy.

50 comments