Self-Replicating 3-Axis Mill - Senior Design Project

[Jlel12]

Hi all,

I'm a senior at Swarthmore College in SE Pennsylvania, studying mechanical engineering with a concentration in machine design. I've been looking around this site for a while now, but finally had a reason to join up and post - namely, my senior design project.

For this project, I want to design and build a small-scale (benchtop), 3-axis milling machine, with a 6" x 6" x 6" work volume. It should be able to cut aluminum and some steels at a reasonable rate of material removal, while still maintaining good accuracy: my design specification calls for 5 in/min linear feed, with a .25" wide x .1 deep cut in 1030 steel, producing a total tooltip deflection of less than .001". I'd like to keep it as cheap as possible - ideally under $1000, not including CNC components.

Finally, I want it to be self-replicating, in the sense of a RepRap 3D printer - it should be able to make any parts that it needs which you can't buy easily from MSC or McMaster. This is the critical design consideration. It means that all manufactured parts have to fit in a 6" x 6" x 6" box, and can't use any advanced machining techniques (no grinding, no tolerances below what the machine can attain). It also means that any non-machining tasks that are required for assembly must be completable with basic, inexpensive hand tools - so no welding. Consequently, I'm looking at a modular machine - lots of bolts, and lots of pre-fabricated components.

To date, I've done the following:

• Cutting Force Calculations: I've developed a spreadsheet, based on Machinery's Handbook and a McGraw-Hill machine design book, to calculate cutting forces for different milling operations. For the milling operation I specify above (which is the "standard" I'm working to), I calculate a total cutting force of 380 N at the tooltip. This is a pretty high standard - the feed/speed combination I use is the combination recommended by Machinery's, which is probably more appropriate for industrial uses than my purposes.
  
• Frame Design: In order to reduce costs and weight, I'm exploring nontraditional frame designs to maximize the stiffness/cost ratio. To examine the "innate stiffness" of different frame designs, I've created 8 different "mockups" in SolidWorks (made entirely out of 1"x1" members, with the only constraint being that they have enough internal space for my work volume) and tested them in SW Simulation. The best design I've found thus far is a tetrahedral design, similar to that used in hexapod machines - I've attached a photo of it below. Its stiffness/unit volume of frame material (presumably proportional to stiffness/unit cost) is an order of magnitude greater than the other frame designs I've come up with.
  
• Frame Materials: I'm focusing now on picking a frame material for my mill. I'm trying to decide between 80/20-type aluminum extrusions on one hand, and steel tubes/extrusions on the other. I see the relative costs and benefits of each as follows:
  
  -Aluminum extrusions: Pros: Standardized. Widely available. Lots of pre-fabricated parts to join sections together. Purpose-made for bolted-together connections. More easily worked with hand tools and small machinery. Higher-precision (flatness, straightness, etc.) than HR steel sections. Cons: Expensive. Not as stiff as steel. Poor damping characteristics (is this correct?)
  -Steel Sections: Pros: Strong. Stiff. Cheap. Known to work with bolted structural connections. Better damping characteristics. Cons: Harder to work. Not as high-precision. Not pre-fabricated - will be harder to attach linear bearings, etc. to frame.

Anyway, that's where I am - I'll be posting to this page as my work progresses. If anyone has any comments/questions about my project, or answers to the questions I've posed above, please send them on!

JL

[jsheerin]

That stiffness should be okay for cutting steel (about 15N/um per your spec), but just remember that the stiffness of the spindle (the bearings, etc), the linear bearings, and leadscrews, nuts and thrust bearings will also contribute to the total machine stiffness. In other words if you only evaluate your stiffness using a simple fea model like you've posted, the final machine stiffness will be lower than what you've calculated. This is not a critique of your method - it's just something I learned while starting to work on my own design.

5in/min sounds really slow. Really, really slow. I know it's a small work envelope, but you might look into increasing that while balancing chip load, available spindle power, and linear thrust and acceleration. There are calculators available on tooling manufacturers' websites that can help with this beyond what I've seen available in Machinery's Handbook, for example: http://mpwr.iscar.com/machiningpwr/ There's also some good info and tutorials available at cnccookbook.com: CNC Milling Feeds and Speeds Cookbook and Tutorial

Some of the cheapest hand tools would also include files and scrapers which can give very high precision, although if you have a time limit associated with a class you might want to stay away from hand scraping anything... But just for example, you could make a precision straight edge from scratch by using the automatic generation of gages principle along with 3 pieces of steel or cast iron, a hand scraper and some spotting fluid. You could also make a square the same way, all of which could be used in aligning your machine and getting flat surfaces to mount linear motion components to (or to be the linear motion components).

What are you thinking about doing for a spindle?

[wutzu]

Your attached picture shows that the nodes are carrying moments. It would be much more conservative (and probably more realistic) to model the nodes as pinned connections, especially since the connections won't be welded. If they're attached with threaded fasteners, it may be problematic to model the moments they carry; probably more so than just assuming they carry none.

Check out the Gingery book on making your own lathe. The components are all either off the shelf, or cast in sand using forms made with simple hand tools. There's some hand scraping involved, but I think he says he did it for $50 (in 1982). He has later books describing building mills, shapers, etc., most of which require lathe work and casting.

[Jlel12]

Hi all,

Thanks for your replies - I'll respond one at a time.

jsheerin: Thanks for all the links! That cutting force calculator is really good - lets me check my calculator's results (which appear to work, although they overestimate slightly).

I'm aware of the impact of the stiffness of other elements on the machine's total stiffness - these FEA models are just intended to give an idea of what sort of relative stiffness performance I can expect from different frame designs. I'll definitely keep in mind what you said, though - as I continue the design process, I'm going to be evaluating the stiffness of each component, and of the machine as a whole. I'm trying to get a high-horsepower computer set up for me at school to deal with some of the larger simulations, although we'll see how that goes - I haven't had very good luck simulating assemblies in SW in the past...

I'll also keep in mind what you said about the feed rate. 5 in/min is probably on the slow side, but I'm also not trying to compete with a Bridgeport here - I'm more targeting higher-quality bench mills, like my Benchmaster. I'll see what kind of forces I come up with if I increase it, though.

Hand tools - I definitely will include files, less likely scrapers. Again, like the RepRap, I'd like to keep the total amount of sweat time going into this machine to a minimum - it's already going to be hard to machine out/put together, so I don't want to add scraping time to the mix. The idea is to make this machine accessible to makers/hobbyists/college students who don't have the time/attention span to scrape their own ways - once they've gotten the machining bug by building this machine, then they can go on to greater and more complicated things.

Finally, re: the spindle - another dirty secret of this machine, I'm going to be buying the spindle (or at least some spindle components). I'm looking at Sherline (who actually have a set of standalone spindles in their industrial products division), Taig and Foredom (they make jewelry makers' rotary tools, and have some higher-performance spindles - I've seen them used in similar applications). If you've got any other thoughts, though, I'd love to hear them.

Crazy machine you're building there, by the way. Quite impressive.

wutzu: Good point re: the moments at the nodes - I'll keep that in mind. Like I said above, this wasn't intended to be a rigorous simulation of the actual frame's performance - however, I will make sure to take this into account once I know what the joints/frame members are going to look like.

Finally, another question for the forums. I'm beginning to think about linear motion components: because of cost concerns, I'm probably going to be making the systems myself. Currently, I'm planning to use precision-ground hardened rod (at least 1/2" if not larger) and ACME leadscrews, because of the (relative) ease of assembly. What sort of experiences have people had with this type of construction? Other suggestions?

[jsheerin]

For a larger spindle, you might look at a mini-mill head from Little Machine Shop: LittleMachineShop.com - Mini Mill Spindle Box Assembly R8
You can also get the complete head with a motor and controller for $300, but that might be pushing your budget. The R8 taper should make it easy to find tooling. I have one of these that I want to put on my router to use as a drilling head.

Are you thinking about making your own moglice type bearings cast in place on the hardened rods (epoxy + filler materials)? That could work, but it might be worth looking at a square / rectangular rail design as well. That could be simpler to assemble from stock pieces. Typically you don't want unsupported round rod, although I suppose if you made it large enough diameter it would work. You could look at that in fea easily enough. Bolt on linear rails would be the easiest to duplicate option, but you'd be spending money instead of spending time on fabrication.

[awerby]

That frame looks like a sturdy way to hold a motor in place, but that's not really what a mill does. The spindle motor either has to move up and down, or the work has to move to meet it. Most builders go with the first option, since the second is more difficult to achieve. While it seems to perform fairly well on the "stepped on by an elephant" test, that's not too relevant to the sorts of forces that actually operate on a mill. Those include resonance and "chatter", which amplify cutting forces, and the torsion caused by extended axes as they cantilever out to meet the material, none of which are addressed here as yet.

It's also unclear how the X and Y motions would be accomplished. It looks like that frame's triangular spaces would make it difficult to accommodate any sliding parts, not to mention any workpieces that might be mounted on them. Or is the idea that the sides of the triangles are collapsible, somewhat like a hexapod? Or is the workpiece presented to the fixed spindle and moved from underneath, like a hexabot? While those machines have certain advantages, they are much more expensive and difficult to build, and software to program toolpaths for them is rare and costly.

While it's a nice idea to make a mill that "self-replicates", it's more important to come up with a design that actually works, and is at least as functional and economical to build and operate as existing designs.

Andrew Werby
ComputerSculpture.com — Home Page for Discount Hardware & Software

[amishx64]

Or is the idea that the sides of the triangles are collapsible, somewhat like a hexapod? Or is the workpiece presented to the fixed spindle and moved from underneath, like a hexabot? [/url]

I too am confused as to what he is JL is trying to achieve here motion wise. My thought was that it would operate similar to that of a delta robot but with a mill on the end.

"http://www.youtube.com/watch?v=Gv5B63HeF1E"]delta robot