Tolerances On Models
Tolerancing is one of those practices that is applied very unevenly across the field of engineered parts. Some people are meticulous about applying tolerances, and figure stackups, reject rates, possibly rework, or put together over-tolerance parts with under-tolerance parts to achieve an assembly that is within tolerance. Tolerances are used for estimating manufacturing costs, estimating scrap, estimating repeatability, failure rates, and I’m sure a lot of other things I’m not familiar with.
To people who design consumer products, plastic, or things that are difficult to measure, tolerances take on a different meaning. With plastic parts, you often use a few reference dimensions to help qualify/dial in a mold/process. You can get tricky with tolerancing a profile against it’s digital equivalent to be within X of the manufactured part at any point, but the problem with this is that the drawing and the tolerances become kind of a legal contract between a designer and the manufacturer on how to determine if a part is acceptable or not. Honestly, on a complex part, there are so many things that can possibly go wrong that might prevent it from actually functioning, that its almost impossible to completely account for everything.
Look at all of the ways we have of tolerancing parts: traditional +/- tolerances, GD&T methods, statistical methods, computerized methods in conjunction with CMM, tolerances on models (Model Based Design). Sometimes the tolerances are on actual geometrical measurements, sometimes on properties like mass, electrical resistance, translucence, friction, color, etc.
I’ll bet readers are all over the map with tolerancing, from using a standardized tolerance block on 90% of drawings, to going all Monte Carlo on part variations. I think the fact is that most of our experience with tolerances comes from having made a lot of good and bad parts. Machinists have a good idea of what type of cutters/machines/process they have to use to achieve +/-0.001″ as opposed to 0.010″.
I think one of the biggest missed opportunities in product development today is in how value engineering and general design practice crosses manufacturing tolerances. So often our tolerances are far tighter than they need to be for simple assembly. If we’re honest, there are sections of parts that could conceivably be +/-0.25″ and the product wouldn’t be any worse off. Sometimes it’s hard to imagine stuff like that, and we just ignore it or apply +/-0.005″ out of habit.
How do you deal with tolerances? and with your answer, please mention the type of product or part or process you apply those tolerances to.
Neil,
The thing is that we don’t always manufacture everything with CNC equipment. Sometimes we use chop saws, etc. or maybe you manually glue something together, or processes like welding or soldering where it’s going to vary a lot. You know, the whole mark with chalk cut with axe bit. Or maybe you’re at the end of a piece of stock, and don’t have enough material to hold onto, do you scrap the part, or do you just allow it to be bigger? Will it make a difference or not?
I was in a shop last week where the stock they received could vary as much as 1/8″, and the product wouldn’t be hurt by even visible variations from design dimensions.
You have to look at how many parts you’re scrapping, and why they are being scrapped. Do you really scrap parts for a dimension that was designed too tight?
Not sure what you are asking here. If you have decent drop saw for example and take a bit of care you can cut a length of RHS to 0.25mm manually. Even really large CNC stuff is usually better than 0.1mm/m. Precision is readily achievable these days and a selling point or indicator of a quality product, rightly or wrongly. Yes you could make some things to 6mm but wouldn’t you have to be deliberately doing poor work or using subpar tools to produce that these days? Automation with temperature control and statistical monitoring of cutting tool wear and so on consistently churn out very near identical parts. Would you as a consumer be happy to buy a car where the doors fit only to the nearest 3mm even though it has not functionally different from 0.5mm? Robotics, CMM and CNC probing have enabled far more accurate and consistent parts to be made – often with the lights out – why not make use of that capability? -as long as you aren’t expecting ‘tenths’ absolutely everywhere why not specify tight tolerances…
Matt:
Great article on an incredibly important topic. Putting the appropriate tolerances on your design is not a CAD function, it’s a design function. This not taught well in the university engineering curriculum and is certainly not practiced well in industry.
I remember my first job where a brilliant machinist asked me: “do you want me to make this to print or do you want me to make it right?” I walked to the shop floor and asked him to educate me. The big takeaway: every zero after a decimal place (tolerance) adds a zero to in front of the decimal place (cost)!
John McEleney