Incremental Sheet Forming (ISF) is a method of creating thin shell-like objects out of metal. It involves slowly pushing a round blunt tool against the surface of a piece of sheet metal suspended in a jig to trace out a 3-dimensional form. This is different to most other sheet metal forming techniques as it requires less set-up costs such as the production of expensive dies. However it's slow speed and comparably lower accuracy hasn't encouraged wide adoption. For small scale batch productions of objects with lower tolderance requirements it is a viable production method.
From the amount of research I've done on this process I've found a lot of work surrounding the engineering side of things such as the mathematics and scripting behind being able to 3d model the process and the architects using robots to be able to create a much wider scope of geometry. My work with ISF is more directed towards the readily manufacturing of parts as quick as possible and without any specialist set-ups or tools.
I use a 3-axis CNC router over a robotic arm because: A - I don't know much about robotics, B - I know CNC and the router I use can apply more force along the Z-axis than most small-scale robot arms. This machine restricts the geometry that can be formed somewhat however once I've refined the process I might look at introducing a varible jig to allow more variation.
I don't write my own toolpaths (yet) as I've been able to tweak those in an existing CAM package to do the job well enough.
Step 1: Tooling and Toolpathing for ISF
A quick web search of tools used for ISF brings up a number of bespokely created titanium ball bearings that are also heated by a current running through them. I'm sure these work well but are cost-prohibitive, especially when just starting out. I used a 12mm dapping tool that came as part of a larger set of 10-20. It's definietly not fabricated as accurately as those used by others however it seems to work well for the pieces I've created.
I do all my design work in Rhino/Grasshopper and use Rhinocam for all my toolpathing. A large part of what CAM programs help you with is simulating what a finished object will look like after it has been milled. I knew I wouldn't be able to rely on this as I'm not removing material from an existing stock, rather, I'm stretching it from an existing flat state. So, throwing this portion out the window I created a simple tool and created a Horizontal Finishing toolpath.
Every CNC machine will differ however there are some commonalities you should be aiming for when changing the toolpath parameters.
First, you want your spindle speed set to 0
Engage, Cut & Retract speeds set pretty low, 500-800mm/min
A small stepdown increment, I've been using 0.5mm
And your Cut Levels Ordering set to Level First. This means that it you're trying to form the shape of two rabbit ears, the CNC will form them both at the same time, instead of forming one before the other. I chose this way as I wanted to minimise any possibility for failure. Watch the embedded video to see what I mean.
I'm still refining these settings and looking at using different 3-Axis toolpaths like Spiral Machining to get one continuous line.
Step 2: CNC-ing Attempt 1
I created a jig out of 19mm plywood to hold the 0.6mm sheet of aluminium I was intending to form. This was then clamped down to the CNC vacuum bed too stop it moving.
Everything was going well until I realised that due to the peak in the middle of my geometry, I needed a small buck to form over. Without it I could only create negative-like curvatures with flat bottoms (still cool though).
Step 3: CNC-ing Attempt 2
I chopped the middle portion of my surface off, made it a solid and printed it out. I also made an MDF jig to sit in the plywood jig which would hold it at the right height.
Things went much more according to plan and the aluminum stretched right over it! I sprayed WD-40 across the surface every now and again which seemed to help a little. The good thing with sheet forming is that the tool is always at the lowest point the so the WD-40 follows it and self-lubricates.
Step 4: Final Results
The final object came out looking pretty nice. There's an issue with the whole sheet warping after it's been formed which will be looked at in future. For a first proof-of-concept that negative and positive curvature can be formed on a 3-axis CNC with a cheap dapping tool it's pretty good.
I've started to push the geometry a little more in terms of variation and slope and will eventually move up in scale. The CNC I use has a bed size of 2440mm x 1200mm x 200mm and as I can get sheet aluminum in similar dimensions I don't see why I can't shoot the moon whith this. Maybe beforehand I'll aim towards working it into the design of a furniture piece to test it in an environment where accuracy will matter.
I hope this has served as a brief introduction to the overall process. It really isn't as tricky as it looks and requires significantly less tech than advertised online!