[jsheerin]

I need a cnc milling machine. While I'd like to pick up a used machining center, the moving and the space in my garage are both problems for me. So I decided to try to build my own. I'm going for something that's massively built but in a small space. I decided on a gantry setup to carry the y and z axes. The x axis will be a moving table. Of course this depends on me coming up with a frame I'm happy with in this configuration. A sketch of this is shown below. I'm shooting for a machining envelope of around 36" x, 20" y, and 14.5" z.

I found a Setco RM5-5 milling spindle on ebay. This is a 385 pound worm gear milling spindle with a NMTB 50 spindle nose. The top speed is limited to 480 rpm, so this is definitely a low speed head. It can handle a 29.7HP input (and put out 27.3HP) at that speed. A pic of me unloading it out of my truck is below. It came on a truck to my work. They have a lower height loading dock that I backed my truck up to and we slid it in using a pallet jack. At home I backed into the garage, slid the engine crane under my truck and picked up the pallet slightly using chains around the milling head. I drove out from under it and then set the head down on the dolly. I might add a high speed spindle on a second z axis to complement this or use a spindle speeder in the Setco. I found a higher speed spindle on ebay (about 6k rpm) that looks okay, but it uses BT35 tool holders which seem rare. I'd rather get something that uses 40 or 50 size holders.

Movement will be via Fanuc AC servo motors run with Granite Device drives and THK HSR25 bearings and rails. Initial calculations show the life expectancy of the bearings should be >10 years when run continuously for 8 hours a day 5 days a week which is way less than I'd actually use it. Ballscrews are on the way for some of the axes and I'm still looking for parts for the Y and second Z.

At the moment I'm working on the design of the frame. I'm getting too much deflection from the z axis structure and am trying a few different ways to reduce it.

[jsheerin]

My rough FEA model is shown below. I'm pushing on the tool in the x and z directions (y and z on the model coordinates - it got turned around a bit). The color shading shows deflection and the scale is in meters. I find it easier to work in metric units in fea but design everything in inches. The deflection is autoscaled which means that it's not to scale. It's exaggerated so you can see where it's moving. So in this design, the tip of the 1" endmill is moving about 0.008" under a 400lbf load and in both the y and -z directions plus the effect of gravity. I should also point out that this model uses symmetry, so while only half the machine is shown, there's a plane of symmetry included in the model so the results are the same as if the entire thing was there. This is done to make the model solve faster (or at all given the limitations of my computer). The limitation is that I can only apply loads that are symmetric or make sense as symmetric loads, so I can't apply a load in the x direction for example.

[jsheerin]

I changed my design - I wasn't getting low enough deflection with the simple z axis design I was using. Now I have two cross beams with the z axis suspended in between them. Then there are 4 z rails, 2 on either side of the z axis. Here's a simulation of this design. This will change a bit as I work out more details, but the general design seems like it's reasonable. Now I need to add the x axis and work out the details for the z axis ball screw support. I'm thinking about putting that in the middle of the z axis. But for this preliminary design, with a 400lbf load in the y, I get about 0.0008" deflection at the spindle nose. That's about 2" / 10^6 lbf. My goal was 1", but this might be about the limit of what I can do without getting too crazy. With a 50 lbf load in both y and z, there's about 0.0002" of deflection of the spindle nose.

[Teyber12]

good call- seemed like with a moving table instead of a moving z/y assembly it wouldn't be much more $ or work to go with your new design and you gain a lot of rigidity

[jsheerin]

This is still a moving z/y assembly and a moving table only in x. The z/y assembly is just supported on both sides. My model is cut in half, so all the support structure would be mirrored. Maybe I should consider a table that moves in x and y though. I did originally but abandoned it because I'd have to have a fairly large overhang of the spindle to get the cutting envelope I wanted. With the weight of the spindle, that didn't seem like a good idea. I guess the problem with a moving x/y if I keep my current z design is that the legs would have to move further away from the center of the machine which would give me more deflection in the gantry beams. I could potentially solve this with more triangulation above the gantry beams. I suppose fixing the z supports to the beams would also increase the torsional rigidity of the gantry assembly... Another disadvantage might be that I would need longer rails for the y... Actually it might work out about the same since the width between bearings under my table is probably similar to the width between bearings of my current y axis. I might have to give that a try. Thanks for the suggestion.

[jsheerin]

Here's a better pic of the business end of the Setco milling head. There's a 1" endmill holder with a 6" extension mounted in the spindle. Next are two ball screws from a Kitamura machining center. Travels are 22" and 34.75" and they're a bit over 1.5" in diameter. My size 10.5 shoe is included for some scale. These will run the Z and X axes. I got these off ebay and they feel good - they spin freely and don't have any noticeable backlash. They definitely need some cleaning up though. I was worried they'd have a bunch of play, but they seem fine.

[sergizmo]

Wow, this is a huge build.

Do you have any estimates of what the total weight will be?

480 RPM is S L O W. Real slow. The only time RPM would be optimal is with a 4"+ diameter face-mill. What can you get it up to with a speeder? Man that is going to be a biiig spindle, 50 taper with a speeder attached.

Could you use 35mm rails on the Z (one pair) instead of two pairs of 25's? Aligning 4 rails on an axis could be tricky.

Good luck with the build! Great work so far with FEA analysis and design revisions. I look forward to any further updates.

Serge

[jsheerin]

Thanks.

I'd estimate the weight right now to end up around 2500 pounds, but a lot may change.

Yeah, I know the Setco is slow. I have two ideas in my head that sort of justify this. One is that I'll mount a second higher speed spindle on a second z axis. Then I'll mount a large face mill or boring head in the Setco and a smaller tool in the other spindle (as I probably won't do an automatic tool changer). The other idea is that I don't care (that much) if this thing cuts slowly. Since it will be under computer control, I can leave it running while I do other things.

Alternatively, the spindle speeders I've seen increase speed in the 1:5 or 1:7 range, so up to about 3400rpm.

I could use 35mm rails - I do have a bunch of HSR35 bearings. However the HSR25 rails seem like they'll be plenty strong even with just two. It's the supporting structure that's showing deflection. Granted my fea setup is not 100% accurate (there should be sliding contact between the bearings and rails where as I have them fixed together), but still, balancing the load instead of cantilevering it should reduce deflection.

[jsheerin]

I did some quick calculations today and if I changed to a movable x/y table, my y travel could decrease from 48" to 37", and my linear rails could similarly decrease 10". This would make it a lot easier to find a suitable ball screw. So that's all for the good. The disadvantages would be that my gantry supports would have to move out from 21" to 38", so I would have to use larger beams to get the same deflection (or more bracing). However, that might be easier to do since the z axis doesn't have to move across the gantry. Additionally I can fix the z axis structure to the gantry beams, so the beams will be stiffened by that. I can also box in the gantry beam where before I had that area open for the y axis movement which should stiffen things considerably. The gantry will also have to be a bit taller to accommodate the y axis being below it and maintain the same z clearance. So likely some more bracing there... But it looks promising so I'll work on a model of that scenario.

[Khalid]

Jshreen
Nice work you are doing.. I am also working on a similar setup but this time conversion of large Radial Drill machine with Arm length 2000mm.. I have to install the Moving bed and the length of the Rails will be 4000mm.. I am using 65mm Rails for my bed (Total 4 with 12 Blocks).. The bed will be able to carry 5000Kg of Load..

Seeing your progress is very motivating..:

1- What ball screw you will use?
2-What Linear Rails You will use?
3-What type of Servo/Stepper you intend to use on all axes?
4- Do you use VFD?
5- Will you drive with Mach3?
6- Can you show us all the forces i.e. Cutting forces, reaction forces, torsional forces and what are the basis of the forces?.. I have to do the similar FEA on Algor.
What are the individual component costs?
I have made an excel sheet for all the components i require for machine along with the prices..Do you have such a Excel sheet for sharing with us?

[Drassk]

Nice work so far, it looks like you're putting some good brainpower into it!

I'm looking forward to watching this one progress, it looks like you're converging on a good design for a nice shop-built machine. I assume you'll be filling all those steel members with epoxy and sand or concrete for damping? That'll move the mass up!

[jsheerin]

Thanks guys.

Khalid, your first three questions are already answered in previous posts with the exception that I don't have all my ball screws yet, so I don't what all of them will be. I'll likely use some Mitsubishi VFD's I already have to control spindle motors, and yes, I'll use Mach 3. I already described the forces I'm applying in my model above as well. I'm shooting for a stiffness around 10^6 lbf / 1 in of deflection at the spindle nose / tool tip. By applying forces in various directions at the tool tip, all reaction and torsional forces, etc. in the frame are taken into account. As mentioned I also apply gravity. Sorry, no spread sheet.

5000kg is huge! You could put two trucks on there... Do you have a build thread?

Drassk, yes, I've thought about filling the steel beams with something. Sand would probably be the cheapest... I suppose I might also think about / model in FEA applying some constrained layer damping to the tubes. At the least I'll do a modal analysis and see what I can see.

[jsheerin]

I spent some time finishing up an initial model for a moveable x-y table yesterday. It has slightly more static deflection but with smaller beams than I was using before. I ran a few more iterations with some additional stiffening which improved things slightly. A screen shot showing deflection under a 50lbf load plus gravity is shown below. Once again this is a symmetric model so there would be another spindle next to the one shown and actually loads applied to this spindle would also be applied to the other spindle. I think I'm at a pretty good place right now. A lot of my deflection is in the vertical direction, but it's just deflection of the structure under gravity (about 0.0005"). But under a vertical load, the compliance looks to be about 0.045" per 10^6 lbf of applied load once symmetry is accounted for which is about 20x better than my target of 1"/10^6 lbf. Parallel to the cross bar, compliance is about 0.4"/10^6 lbf, and perpendicular to the cross bar it's about 0.63"/10^6 lbf but those two numbers would probably be a bit different due to symmetry.

Next I'll add the rough design of the x and y axes to the model. I might also change my model a bit to try to more accurately model how I'd weld these tubes together. Then I need to design the support structure for the z axis screw.

[jsheerin]

I found a long ball screw last week which gives me all the screws I need for this machine. However in thinking about it some more, I realized it might work better to change how the dual spindles are oriented and place the long ball screw differently. See the sketch below. The spindles are now placed next to each other in the x direction. This will let me make the gantry narrower which should improve stiffness and make it easier to package the ball screws into the mill.

Now my y axis screw will be 10" longer than it needs to be though. It already has the end blocks, so I don't really want to shorten it. I'm pretty sure I can package it as is to get my desired travel in the minimum size package, but I'm contemplating just giving myself an extra 10" of y travel. This would give me a working envelope of ~31"x34"x22" instead of 31"x24"x22".

[Khalid]

I really want to see this in finished form.. The efforts you put in 3D modeling and stress analysis is awesome and plausible..
Regards

[jsheerin]

Thanks Khalid. I'm sure it will be quite a while before it's finished, but hopefully I'll get there eventually.

I received my last ball screw yesterday. Below is a picture of the x axis ballscrew (the long one) and the second z axis screw. The z axis screw is from a Mori Seiki MV45/40 mill and includes one of the bearing blocks and triple bearings at the other end. The x axis ball screw will give me 49" of travel and the z axis 22". I must be getting pretty good at estimating screw travel from pictures with rulers in them, because this is almost exactly what I estimated these screws would do (ebay sellers didn't give the actual travel measurement).

The x screw appears brand new and unused. I'll have to find bearings for this one and fabricate bearing blocks, but that shouldn't be too bad.

I've also been spending some time Machine Tool Reconditioning. It's incredibly dry, but it's got all kinds of suggestions for how to check alignment of machinery which is giving me lots of good ideas of how to build this thing. I've also been reading Michael Morgan's scraping book and watching his dvd trying to get a handle on scraping. It's good stuff - http://www.machinerepair.com/. I picked up a scraper and some Dykem hi spot and will be giving that a try soon. I've also been thinking about how I'm going to stress relieve the finished machine frame after welding. I've started doing the calculations for how much power a stress relief furnace would take and it seems reasonable to maintain the temperature in a large enough space. Now I have to figure out how much power it will take to get up to that temperature (so I can figure out what kind of heating elements I'd need). I've built a small propane fired forge before out of stacked fire bricks (see pic below) and was thinking of a similar but much larger setup for this - something that I'd build around the machine frame, use once and then disassemble. I was thinking of electric heating elements instead of gas though. I've also been reading about vibration stress relief, for example: http://www.stressreliefengr.com/ and http://www.meta-lax.com/. I could easily build one of these units - a 3 phase 1/2HP motor driving an off center mass run by a vfd (as I have plenty of vfd's sitting around). But opinions seem mixed on whether it's a real effect or snake oil.

[jsheerin]

I've been doing some thinking in my free time about evolving my design, and I realized I've been modeling my frame somewhat incorrectly. I have the z axis fixed to the gantry through the linear bearings in x, y and z directions. When I first started out, this was probably somewhat realistic. While the bearings obviously would not be supporting the z axis in the z direction, the ball screw would be mounted fairly closely to the bearings so the vertical load would still be applied to the gantry in roughly the same spot. However on my newer designs such as the sketch in post #14, the ball screw will be attached to another frame on top of the gantry which will distribute the vertical load to the outer portions of the gantry instead of the middle (assuming this works well when I model it). This means the vertical forces that I'm currently modeling as being transferred through the z axis bearings are not accurate at all and are possibly causing me to over design my gantry. So now I need to figure out how to model and constrain the z axis so it's close enough to reality to be a useful design tool. Looks like I'll be learning some new ANSYS techniques...

[ebrewste]

That's some monster you are designing there Exciting stuff, especially since you are actually doing some analysis and not just guessing.

In addition to remembering that the bearings are free to move in one direction, don't forget that they also have compliance. They HSRs are mercifully the same stiffness in four directions, but they also have another stiffness in torque. You may want to model this if you think you are getting any significant cantilevering effect, though I can't see much in your drawings.

Also, don't forget ball screw deflection. They can be modeled as a spring on a spring (ball screw column axial deflection + ball nut axial deflection).

Not sure how much of the above you have already considered. Ignore if you've already gone down that road.

Awesome stuff -- I can't wait to see your progress.

In Moore's "Foundations of Mechanical Accuracy", he seems to really rail against vibration as stress relief, but I can only state this as a believing reader, not someone with any real experience in a controlled experiment. Fantastic read, if you can find a copy somewhere.

Erik

[ebrewste]

Regarding the annealing / stress relief: You may want to see if there is a race car industry nearby. I couldn't tell where you live. Race car roll cages should be annealed / heat treated, so there is likely someone with a big furnace that is used to heat treating tube assemblies, likely for cheap.

If you just want heat treat yourself, please take pictures, I and others will be fascinated!

[jsheerin]

Thanks for the comments and compliments.

Because I'm doing solid modeling for my fea, the deflection of the screw itself under load would be accounted for, but the compliance of the ball nut would not. Similarly, I know the HSR's have compliance but I haven't modeled that yet. I have a spreadsheet where I've done some initial calculations on what loads would be and the life expectancy of the bearings based on that, so I could probably add that in there. I guess my position on this is the deflection of the ball screws and bearings will be whatever they are, and I've tried to pick ones that will be appropriate for the load ranges I'll use them in. In fea I'm looking more at just what the frame does and I'm trying to make as many simplifying assumptions as possible so I can get a reasonable design. I usually tend to over complicate things and I'm trying to avoid that as much as possible here (although I'm sure some would argue I'm already doing that), but I do want to have a reasonable idea that what I build will have good performance without a lot of unnecessary expense and mass in the frame.

I have heard of the Moore book somewhere - I think I have it on my Amazon wishlist at the moment. I'm currently slogging through Machine Tool Reconditioning (it's great for falling asleep with). Maybe Moore's book will be my next one.

I think I might heat treat myself. My interest in doing it myself is that I don't really have the equipment to haul my completed frame somewhere to get it heat treated. So my options are to buy / build / borrow a trailer and appropriate crane or come up with a way to do heat treating in my garage. The heat treating would seem to be more interesting and give me more capabilities in the future, so that's why I'm leaning that way. Whatever I do, I'll definitely take pictures although I bet it'll be a while before I get there.