UKC

I want to draw a rather awkwardly shaped 3D object. If I was doing this with pen and paper I'd have eye-balled it safe in the knowledge that people probably wouldn't spot any errors, but I am also using this as a learning exercise with a new CAD package:

So how would you go about measuring a complex curved object (like for instance a DMM Phantom) to give yourself enough information to make a passable stab at drawing it?

I've had a quick think about this and have come up with a few ways of doing it ranging from massively inaccurate (lots of calliper measurements which will be hard to reference to each other) to really time consuming (setting up a dial gauge on an X-Y stage and taking lots of measurements). I have a sneaking suspicion the answer is either get someone else to do it or buy a 3D scanner but I was wondering how a proper draughtsman would do it?

In reply to Reach>Talent:

With something complexly curved like that, like you say it would be very difficult to accurately measure and reference those onto your cad image. I don't do 'CAD' as in engineering models but I do a lot of general 3D modelling for visualisation purposes (where absolute accuracy isn't required). So if I was doing this I'd take a side on photograph of it with minimal perspective, do this by taking it from far away but zoomed in (you might need a tripod) Then load this image into your 3D software and use that as a reference to set out all your curves and things. it may not be super accurate but at least everything should be more or less in proportion.

Then supplement this method with a few additional measurements taken with calipers to make sure everything is more or less spot on and for the depth of things. Also, just do half the model then flip it over and join it together to get both sides.

In reply to Reach>Talent:

2d or 3d, and how accurate does it really need to be?

Assuming we're talking about something like a krab in 2d you could do a pretty respectable job just by clamping it up in a bit of a jig (or even fixing it to graph paper as a guide) then using calipers and the jig as your reference. Tracing a well set up photograph will probably get you within a few of 10ths of a mm. Add a height gauge or DTI to your jig and calipers plus a lot of patience you could do a pretty respectable job in 3d but at that point you'd probably be better as you say setting it up in a mill or on an XY table.

I don't know much about 3d scanning but I guess that's the more highly automated option that may still leave you with a big job to do turning the surface data into a usable drawing.

jk

In reply to jkarran:

3D and the accuracy is a very good question, which I probably should have answered first! I suspect the answer is "it depends": I was having idle thoughts of making a rather overcomplicated clip-stick just for the hell of it, so I suspect it rather depends on exactly how I decide to make it and what materials I use. I was also interested in how to solve more difficult high precision problems as I occasionally get interesting challenges at work (like "how can we machine and then assess the quality of a complex multifaceted point on an object smaller than a grain of rice?").

In reply to Reach>Talent:

As it's cast you're not working to high tolerance, so any means of getting a decent image which you can take co-ordinates from, you may need 2 or 3 images from 2 planes. You can then use these to build your 2 d drawing or 3 d model. Hi precision reverse engineering or measurement used in quality control will use tools like optical projectors and/ or coordinate measurement machines to do this, and if the software has been developed sufficiently the measurement and cad can be interfaced. At that level you can work to 3 microns or 3/1000 mm tolerance.

In reply to Reach>Talent:

Thinking about it it's probably not cast but drop forged, with tighter precision without looking at it closely enough I'd be guessing but I'd hazard a guess of 25 microns, but all the other principles apply.

In reply to Reach>Talent:

You've got it about right, its either accurate or a bodge. If you want it accurate, like properly accurate, you need to either invest in 30k's worth of laser CMM which will scan the shape to +-0.005mm or you need to scan (in a household scanner) the object and then put that into your cad as an image on a sketch, scale it using salient dimensions and trace around it. As for actually getting it made, well give it to someone as an IGES file with again salient measurements such as overall size and toleranced dimesnsions and they'll just use that info. For programming a surfaced shape, that infor is far more useful than a fully dimensioned 2d drawing - they simply won't use it as doing 3d surfacing, you'll be using the surfaces of the model rather than a paper drawing! Let me know if you need some help as I run full modelling and prototyping services for exactly this sort of thing...

In reply to mike kann:

Many thanks, when it occurred to me that the getting it made part applies a reasonably well defined specification for how accurate it needs to be as there isn't much point working to much higher resolution than the 3D printer that I was planning on testing it on (which isn't brilliant). So that removes some of the hassle which is good as FreeCAD is driving me round the bend

In reply to Reach>Talent:

Haha - that's like using a broad sword to do open heart surgery - bit of a blunt heavy instrument for delicate work. But yeah - you're right, the reoslution on most cheap 3d printers (and quite expensive ones) is fairly rubbish compared to a CNC mill which is what we use. I suppose for what you need, a fair bit of bodgit and scarper can be applied...

In reply to Reach>Talent:
The reason programs like free cad will drive you round the bend is they're crap if you're doing any cad work for anything other than a hobby, you're better getting an old licence for solidworks, even 2005 on xp is going to be miles better than any open source program

In reply to Reach>Talent:


To be 3d printed? Mine is barely krab shaped and it works fine, I'd scribble round the krab on graph paper or import a photo reference layer at most for a job like that.


I'd speak to a gemologist/watchmaker for ideas, unless of course it's *much* smaller than a grain of rice.

jk

In reply to Reach>Talent:

A 3D Scanner would give you an enormous amount of data, which you would then have to clean up afterwards - a job that could well be more time consuming than creating it from scratch.

Case study, If I wanted to create a CAD model of a DMM Phantom, I'd do the following -

1. Pop it on a flatbed scanner and generate a bitmap file
2. Trace the bitmap file to a vector format using Illustrator or similar
3. Import the vector file into your CAD package and extrude to a 3D object
4. Use the usual CAD tools to add the finishing touches.

In reply to jkarran:

I did say over-complicated, I was having an idle doodle earlier of a mechanism for opening the gate using an rc servo

In reply to Reach>Talent:

It does not take long to measure things accurately with calipers, and the effort in making an accurate 3D drawing with a CAD package is worthwhile. One can enter accurate xyz points and reference lines into decent CAD packages and then snap your 3D model onto these. One can make curves that interpolate through these points and then loft these into 3D.

There is a long learning curve with all the CAD packages, but the cheaper ones will drive you nuts.