[PedroCG]

jsheerin,

Are you aplying a force on the tool and a force on the vise (in the opposite direction) or just a force on the tool?
I ask this because a ballscrew of 50mm diameter 500mm distance between nut and suport bearing has a stiffeness of 0,8N/um. I dont know what's the definition of effective stiffness of tool/work piece interface but 10N/um would only be possible if the table was fixed to the machine.
I've never designed a machine but i've designed a FSAE chassis and when people talk about torsional stiffness they often seem to forget that theres no point in having huge stiffness numbers if you then use "soft" springs in the suspension.

[PedroCG]

I just made a fool out of myself by a factor of 1000... lol. that screw would have a stiffness of 800N/um. Anyway are you applying the force on both places? also could you post the strain energy plot i talked about? thanks.

[ad_bfl]

Zach -

Thanks for finding that number.

JSheerin - you may want to look at linear roller bearings given the amount of steel in your design. When you get to this level of "stiffness" those may be useful, but sit down when you ask the price.

https://tech.thk.com/en/products/thk...in.php?id=313#

CO preload has a big impact on rigidity.

Best Regards
Al

[jsheerin]

Thanks for all the info / suggestions guys. It really helps.

Al, I have a box of 100 or so new in the plastic HSR bearings of various sizes and some HSR rail, so I am pretty committed to using those. I've read that the roller bearings are much better in several regards for mills such as not brinelling the rails when you crash the machine due to a line load instead of a point load, but I'm trying to keep the budget as low as possible while still achieving good results.

Pedro, below is a plot of strain energy. I don't see any big difference between this and the Von Mises stress plots I've been posting as far as showing the distribution of stress, so I'll probably just stick with those. Would you expect there to be a difference? I'm constraining the vise face to not move and applying a force on the spindle tool taper. No worries on the misplaced decimal point - I catch myself doing that all the time.

John

[ebrewste]

I have a box of 100 or so new in the plastic HSR bearings

100 lm bearing blocks!? If you are dissatisfied with the stiffness of the lm guides (reasonable, given your goals), have you considered using lots of bearing blocks? It would be typical to use two bearing blocks per rail, but there is no reason why you couldn't use more. 2x the bearing blocks, 2x the stiffness...

You might consider running more rails, too, understanding that you don't have such a large supply of those...

[jsheerin]

Yeah, I have actually been planning to run at least 6 bearings per rail. I just haven't added them to the model. It's a pain to add them - well, not to add them but to have them in there. I have to find the area numbers for each surface I want to apply loads or constraints on and type that into my script. There are three surfaces per bearing, and they're hard to find because they tend to be buried under other stuff. So adding three or six more hasn't made it into the model yet.

You made me go actually count my bearings - I have 30 HSR25's and 15 HSR35's plus some 20's and 15's, so ~50. I bought them all from a Sony plant that was closing. They were spare parts for the plant. I paid a few dollars each for them, but I sold some other stuff I bought there so they essentially turned out to be free. I just wish they'd had some rail...

[PedroCG]

I'm used to spot these kinds of ooops before posting them online

Strain energy has a closer relation with displacement (in a spring E=1/2 K x^2), and trying to maximize stiffnes is essencialy trying to minimize displacement.
When you look at the displacement plot of your machine you are looking at total displacement, but strain energy gives you an idea of relative displacement in each component, one could say stress is proporcional to strain but strain is displacement per lenght and therefore does not give a clear idea of the relative displacement of a structural member.
The strain energy plot you posted sugest that rezising the top two diagonal members could have a possitive effect in stiffness.
If you think about a simple mass-spring system the natural frequency will increase as the stiffness to weight ratio increases. A more complex structure has more than one vibration mode so i guess in some cases odd things may occur but not to often...
Try apllying the same force you put on the spindle to the faces of the vise, that is what happens in reality.

[jsheerin]

Like I said, I'm constraining the vise face. This effectively does the same thing you're suggesting, as I am not constraining the model anywhere else. When I was constraining the base, I was applying an opposite force on the vise, but I decided to not constrain the base as that assumes the concrete slab is infinitely stiff, which it's not.

[PedroCG]

ah ok i missunderstood that part soz. i've never used ansys so i'm not familiar with it's capabilities, i normaly use cosmos M (it's a bit hard to generate the mesh), cosmos has spring elements and those can be used to model the stiffness of a support fairly well, dont know if ansys lets you use something like that.

[jsheerin]

I've been taking a break from mill design recently, but THK did send me rigidity (stiffness) data for the HSR series bearings (attached below for anyone who's interested). They didn't comment on whether or not I could change the preload of the blocks myself though.

I've also been thinking about different frame designs. At the moment I'm using mostly standard rectangular tubing sizes in my frame, but realistically I'd have to build the main gantry beams from flat stock. So if I'm going to do that, I suppose I could do something more like what's shown for the frame in these Mori Seiki videos:
[nomedia="http://www.youtube.com/watch?v=h0nTvEHVQiU&NR=1"]YouTube - The NV6000 DCG offers the world's best accuracy[/nomedia]

[nomedia="http://www.youtube.com/watch?v=Io563Baq11Q&feature=relmfu"]YouTube - The NVD1500 DCG offers the world's best quality for dies and molds, and precision parts[/nomedia]

I do have two ball screws that I could use for the Z axis if I wanted to move the drive point closer to the center of gravity of the spindle... This could also make life easier on my linear bearings in the Z axis.

In any case I'll be taking a break from this for a while to work on redesigning some parts of my cnc router using all the modeling techniques I figured out here. My goal is to get the design work done on that so I can rebuild it this summer and use it in the fall and winter. I'll come back to mill design after I finish that.

[jsheerin]

Here's a graph of the stiffness of the HSR25A/B/R bearings for no preload and C0 preload. I also put in my initial estimate of no preload that I used in my model. It was reasonable, but a bit optimistic for the reverse radial loads. I'm not sure how I could model that easily other than to set the stiffness to the lower level. I could add the different side load stiffness without any trouble.

[jsheerin]

My (hopeful) spindle motors showed up yesterday. They're 3.7kW continuous, 5.5kW/30min motors from a Matsuura MC500V. They're rated to go up to 6000 rpm. They were $30 each - score! Honestly this was their biggest selling point. Of course shipping was a bit more... I'll have to test them and make sure they work. For that price it was not guaranteed that they did, but I don't have a reason to think they don't. They were off a machine that was being parted out. The rest of the machine ended up getting sold for scrap because the owner ran out of time to part it out. He had just bought it for one drive or axis motor to fix a broken machine. Now that I'm going single spindle, hopefully I can use the second motor for a cnc lathe (an even further off project).

I'll be slowing down on the mill for a while. I have a few other projects in front of it - redesigning and rebuilding the z axis (at least) on my router and enclosing it and then building a few projects with it. I'll probably still do things on this as I feel like it, but it will probably be a while before I focus on it again and then start building it.

[jsheerin]

More stiffness data - the HSR35's are not massively stiffer than the HSR25's in direct loading, but they do seem to be much stiffer under an applied moment (~2x). The first comparison plot of the 25's and 35's is with no preload.

[ebrewste]

Is this measured data from you or from the data sheets?

[jsheerin]

This is taking deflection vs load data from the THK data sheet I posted above and converting it to stiffness (since that's what I need for my fea). The data in there is calculated, not measured. I have been measuring stiffness of my router, but it's about 100x lower than these values. I don't think I could measure stiffness on these things very easily.

[jsheerin]

Alright - I lied. I've been playing with a new model in a few minutes of free time when I'm not rebuilding the sprocket carrier on my motorcycle (got to get it ready for warm weather!). It's basically like the Mori Seiki DCG machines. I left the frame solid, no x/y table and just worked on the z axis. I put the dual ball screws and the linear bearings on the plane of the center of the spindle, so the static load on the bearings should be negligible, with only loads being applied during acceleration and cutting. This assumes I can mount the spindle drive motor directly above the spindle. I could couple it directly to the spindle but I'd have to get creative about how to setup a draw bar. Or I could couple the motor to the spindle through a secondary shaft off to the side to go around the draw bar. I also figured out how to get a better mesh in my model - use the smartsize command to have a large mesh size in areas where the geometry is big and in the middle of sections and a smaller mesh size in areas where the geometry gets small. This makes my models run a lot faster and in some cases makes it possible to model things period. Y and Z stiffness is great - 429k in y and 4.5M in z. These would come down with the x/y table added obviously and with the frame shelled out to be something reasonable. Right now it's ~74k lbs of steel. With HSR35 bearings with no preload, the X stiffness is only 92k, so given that value will decrease as well, that's too low. With C0 preload, the X stiffness goes up to around 240k, but given that things will get more flexible that still makes me a bit nervous. Most of the deflection is coming from the bearings at the moment.

Because of this and reading Slocum, I've been thinking a bit about hydrostatic bearings. They use pressurized fluid pushed into the space between a bearing pad and rail. Slocum gives stiffness values for these bearings of around 9000N/um, or about 90x stiffer than the THK bearings. If I could make my own bearings, this could also reduce cost... It would also remove the unknown of trying to preload the THK bearings by replacing balls and the cost associated with that. They're also supposed to last a really long time as long as the fluid pump doesn't die and stop supplying fluid while the bearing continues to move. I threw that bearing stiffness into my model and the X stiffness goes up to 561k. If I space the bearings apart another 6" it goes to 607k and now the z axis structure is obviously bending (but I can work with that). So things to think about... If anyone has any input on the bearings, I'd be interested to hear it.

[rowbare]

Check out the base these guys use. It is a bridge design with a moving head and table. It looks pretty rigid. Of course it is an 11,000 lb machine with a fairly small cutting area...

bob

[ebrewste]

Check out the base these guys use. It is a bridge design with a moving head and table. It looks pretty rigid. Of course it is an 11,000 lb machine with a fairly small cutting area...

bob

I think you are missing a link...

[rowbare]

I think you are missing a link...

Oops

High Speed Cutting

bob

[ebrewste]

Since this is the only thread I know of with quantitative stiffness and resonant frequency analysis, I thought I would add a fun youtube video from Dixi (Mori Seiki) that has a mention of 120 N/um and 80 Hz resonances in their jig boring machining centers.

[nomedia="http://www.youtube.com/watch?v=Eo33ATkGqeg"]YouTube - Mori Seiki's Dixi brand Uber-Precision & Jig Boring Machining Centers[/nomedia]