[jsheerin]

I believe I figured out how to simulate something more like the reality of linear bearings in my design. For anyone interested, I used contact elements in ANSYS. I setup my model so that there was a gap between my linear bearing and the rail and then setup contact pairs between the bearing surfaces and the rail using separation only always and selected key options to close the gap before starting the analysis. A refinement of this which I might look into would be specifying the compliance of the coupling. I could use this to simulate the compliance of the THK bearings from catalog data.

Attached is a screen shot of the displacement of my test model. It's a vertical rail constrained in x,y,z at the bottom of the rail. It has a bearing on it, and there's a vertical piece that holds the bearing from the top of the rail (sort of like the screw in my real design). There's a big block attached to the bearing with a cylinder (cutting tool) coming out the bottom, similar to my milling head. There's gravity applied as well as a force in the y direction on the cylinder. You can see the bearing has slid down the rail and that the block has gone down with it and twisted to the side. Harder to see is that the top of the rail has deflected in the y direction in the opposite way. The bearing is transferring the moment applied by the force on the cylinder into the rail and deflecting it in the correct direction.

So now that I know how to model this I can go back to working on my frame design.

[jsheerin]

To be more specific (and just for my own notes), to do contact this way, I'm setting keyopt 5 and 9 to a value of 1 and keyopt 12 to 4.

[ebrewste]

That's some great work there. Really cool to see the displacement showing that it's working. Top notch.

[rowbare]

Have you read this? It is an interesting read for anyone wanting to design a machine.

http://www.mech.utah.edu/~bamberg/re...e%20Design.pdf

bob

[jsheerin]

Thanks for the link. I'll give that a read - looks good.
Here is some more from the same author I found:
http://www.mech.utah.edu/~bamberg/re...astDamper.html
http://www.mech.utah.edu/~me7960/lectures.html - see Topic 14

[jsheerin]

Oops - with my current design the max height of the mill will be ~110". That's right at the height of my ceiling (probably a bit above it actually), so I'm going to need a different approach. The problem is that I was going to mount the Z axis screws on top of the spindles and have the screws move vertically with the spindles. For a 22" Z travel, this adds about 44" of space to the top of the mill. Instead of this, I think I'll have to put the screw to the side of the spindle. The screw will still travel with the spindle, but it will only add 22" to the top of the machine (above the gantry). This would keep my overall height to around 86" or so, but it would only be about 64" should I ever need to move it (with the spindles all the way down). I'll have to think about this some more and sketch something up.

[jsheerin]

Spindle number two arrived the other day - a 7500rpm spindle with a 40 taper from a Haas VF3. Below are some pics along with pics of three of the servos I'm planning to use for this - Fanuc 10S models with incremental encoders. The two big ones have brakes on them. The other shorter one does not. The ones with brakes will be for the z axes. I have another 10S with a brake in a box somewhere, so that will cover everything. However I might decide to go with a 5S for the y axis on the table as that would be the smallest weight to move and could benefit from the smaller motor size. I have a few of those sitting around as well.

[jsheerin]

I'm still bouncing around spindle ideas in my head. Below is a quick sketch of my idea for the Haas spindle to go back to having the ball screw mounted above the spindle and next to the spindle motor. I think this arrangement would be short enough to clear my ceiling on the Haas due to where the motor would be, etc. but I need to measure the spindle and actually draw it up.

I also scored two Yaskawa 5.5kW spindle motors today. It will be a while before they show up.

I've been thinking about ditching the Setco spindle. That would give me more usable travel, simplify construction, etc. and the Haas should be capable of everything I need. I just hate to not use the Setco now that I have it... Maybe put it to the side and use it for horizontal boring? I'd say some kind of 4th axis setup, but I already have a dividing head...

[jsheerin]

I took a break from mill design for a while, but I've still been thinking about it. In the last few days I redid my model to go back to something similar to my first design - a typical gantry design with the y and z on the gantry and a moving table for the x. However this design is with my single Haas spindle. It also incorporates rough approximations of ball screws and sliding contact for the bearings. Additionally, I added some detail to the gantry beam to make it more accurately capture what the structure would be like if I welded rectangular tubes together. In other words, the walls of the tubes that touch in the middle of the beam are not solid - there's a small gap between them and the tubes are just joined at the center and outer edge. Of course I did all this at once so I don't know what the effects of any one element were on the deflection, but my deflection has gone way up. It appears to mostly be in the gantry beam - it's twisting a lot. It needs to get at least 10x stiffer, and preferably more like 50x.

So I have been and will continue running some numbers on different shapes of beams for the gantry. Of simple shapes, round tubes give the best performance for a given mass, as does making the tube a larger diameter. My goal is to find something that is somewhat reasonable, that I can actually build and that will give me the stiffness I want.

[jsheerin]

My gantry beam is having to get a bit bigger than I want to get the deflection I'm shooting for, so I'm thinking about alternate machine structures. Here's a horizontal mill. I came up with this after reading the beginning of Precision Machine Design. He talks about a horizontal mill there with a moving wall for sealing. I haven't thought a lot about it yet, but this could be easier to build. I wouldn't have to get multiple surfaces flat, parallel and square - just get surfaces flat and square. However I'm not sure if it would be any better for deflection. If I had a tombstone fixture on the table (as drawn) and then had the spindle all the way up, those would be some long moment arms... I suppose most of the time I could machine near the bottom of the Z travel which would be more rigid and just ensure the compliance with the Z all the way up stayed reasonable.

[giz]

Have you read this? It is an interesting read for anyone wanting to design a machine.

http://www.mech.utah.edu/~bamberg/re...e%20Design.pdf

bob

Thanks for the link. I'll give that a read - looks good.
Here is some more from the same author I found:
http://www.mech.utah.edu/~bamberg/re...astDamper.html
University of Utah - ME EN 7960 - Precision Machine Design"); if (currentModule == "index") document.write(" - Main"); if (currentModule == "courseInfo") document.write(" - Course Information"); else if (currentModule == "projects") document.write(" - see Topic 14

Weird to see these posted - I took two classes with Bamberg. He taught my Design Methodology and Senior Design courses while I was at the University of Utah. He does a lot of EDM-related research now.

It's too bad that this type of stuff isn't covered until an upper-division graduate level course...

[jsheerin]

Messing around with some quick calculations over the last week, I'm not sure the horizontal design I sketched above would be any better. I think it could be better in some regards but worse in others (torsion vs bending of beams). Since mounting work would be harder (at least to me, who has only used vertical mills), back to the vertical design. In case you can't tell, I am still very much in the design stage even though I already have parts.

Reading Slocum's book the other night, he points out that Abbe errors are made worse by having bearings far from where the force is acting. So that would suggest that I should move my Y bearings off the top of the gantry and make an XY moving table (similar to a previous sketch). That should also make it easier to make a stiff gantry cross beam. I have some more ideas about how to stiffen that up as well, so when I get time to do more modeling and simulating, I'll see how those pan out.

[jsheerin]

Success! Or at least I'm getting closer. I designed and modeled some of my stiffening ideas, and they worked. I also added all the axes in at least some form. I also drew in a Kurt vise on the table. For Z loads, I have a stiffness of about 500,000 lb / inch of deflection. For X loads, I have a stiffness of about 125,000 lb / inch of deflection. My target is 10^6, but these numbers are almost 10x higher than my previous design and there are still some obvious improvements / changes to make on my current frame such as stiffening the x/y axes and tying them into the gantry in a more substantial way, so I think I might be headed in the right direction.

I attached a few screen shots. The first one is a load applied in the X direction - going under the gantry. There's also a reaction load applied to the inside smaller face of the vise. The second two are a load in the Z direction (once again both on the vice and the cutting bit). The vise doesn't move much vertically, but it definitely moves under a load in the x direction. In fact, it looks like the stiffness of the x and y axes in the vertical direction is good enough - it's around 2*10^6 lb/in, so if I had a similar stiffness from the rest of the structure, it would end up right around 10^6 total.

[giz]

I like the X/Y moving table - I definitely think this along with the overhead Z is the way to go. Your design reminds me of an engine stand. Have you come to any conclusions on where to mount your Z ballscrew?

Imagining the size of this makes me feel tiny.

[jsheerin]

I was going to say it's not that big, but currently it 87" tall, 88" wide, and about 77" deep, so that's pretty big... Actually that's good to know as I think I'll have to mount my Z servo off to the side of the Z ballscrew instead of above it. And to answer your question, the Z ballscrew is in the 'normal' place on the z axis. It's in the images above, but it's tough to see due to the resolution.

I did some more fea over the weekend, and a lot of the deflection in the x/y table is coming from the 0.25" wall tubing. The walls are flexing a lot near the bearings but not in other places. This is leading to about as much deflection of the table as the spindle is seeing. So now I'm switching over to a more solid platform for mounting the x and y rails. We'll see how that goes...

[ebrewste]

a lot of the deflection in the x/y table is coming from the 0.25" wall tubing. The walls are flexing a lot near the bearings

That's a great insight! Your time spent in FEA should really pay off - that's a hard problem to fix after the fact. Are the pictures you attached a few posts ago displacement plots. It would be very interesting to see strain plots. While your cutter doesn't see strain in the other parts of the machine, strain is a good indicator of what needs beefing up, next.

[jsheerin]

Yep, those are displacement plots. I convert the model to be in meters before I model it as I find that easier to work in, so the displacements are in meters with 100lbf applied (but it's a half symmetry model, so 200lbf total). It's kind of crazy to see the tubes and rails deforming so much over such a short span, but then again they're only moving 0.0015" or less with those loads. Of course I want them to move about 10x less than that...

I've looked at stress plots a bit but haven't spent a lot of time looking at them so far. I'll try to post one of those on my next modeling run.

[mwmkravchenko]

Hi John

Have you looked into the idea mentioned in the Machine design links of embedding a second tube with a tube? It looks like they used an expanding grout to pre-stress the collumn under test. Looked to be a great solution to getting a very low deflection in a smaller form factor.

Mark

[jsheerin]

Yes and no. I don't see that providing me with much lower deflection (that's what I modeled), but it would probably be useful for adding damping.

[mwmkravchenko]

Having worked on many large machines for quite a few years I have to whole heartedly agree that dampening is very important. Just take for example to quality of cut thay comes off of a cast frame bandsaw versus a welded frame bandsaw. I have used both at length. My shop has a cast frame saw on purpose. They just work better.

I think you could approach that level of dampening and rigidity with the use of stressed members. An iron casting is in itself in a state of stress unless it is baked out to relive that stress. I think you were looking at doing something similar.

As a side note the stress relief methods also include a cryogenic treatment using liquid nitrogen to cool down a part. It's not that much money for the stuff. A little more than the price of milk up here. A proper container is the requisite. Some guys have made an appropriate plywood box caulked and lined with heavy polyethelyne. A machine your size would be a bit of fun. But then again I'm guessing that sections are going to bolted together.

Mark