[peteeng]

Hi Jag - Yes I have had the same issues with convergence and with this model as well. For instance if I use a forced deflection I get a certain machine stiffness, if I use a force then convert that to a N/um I get a different stiffness. Maybe 10% delta which is not good. Does not matter how many adaptions are used. I have sent examples and questions to Altair and they can't exactly answer the question. Welds have all sorts of spikes in them etc etc. So its a fast handy "design" tool but not yet right for confident detailed analysis that other FE systems give. But in time they will sort it and this is the future of FE. Peter

[peteeng]

Hi All out there - I've started adding features so the parts can bolt together. I want to use a build "bottom up" approach if possible so no bolts are underneath or behind things. Peter

[klaas123]

Milli 16 is a z axis bolted on to a saddle, I think we need a nr 17 Peter.

[spumco]

Peter,

I know you're still in the early stages, but if you're using flanged trucks you can get rid of the column flanges (for bolt access) and bolt from below. All the fussy features on the bottom of the column can go away.

Looks like you've got clearance under the trucks for screws and a stubby allen key. If not, a couple of access holes through the saddle would do the trick.

-Ralph

[peteeng]

Hi Klass - yes it should go to the next version... I usually jump a version when its time to do some more FE work as my FE file is organised by the version serialiation...

Hi Ralph - I like flanged bearings but the screws are smaller and they have to be done up from below. I'm not sure I can do them up as tight. My first router had something like that and I was not 100% happy. So a 20mm car uses M6 screws from the top so M6 bolts are stiffer then M5 bolts. M6 can be done up to 11kN preload. M5 from below is 8kN so much less total preload 73% so 17% less friction in the connection. We shall see next round when I get this one loaded up. I also don't like square cars with blind holes that you have to be careful with the bolt length so it does not bottom. Pros and cons.

On my first router I had a part with blind holes and as the machine settled I kept torqueing things down and the bolt had bottomed. Took me a while to figure out what was going on. I was chasing a small movement in the Z axis assembly.... you now make me want to use flanged cars bolted from the top....

Then there is interference from the drive screw and nuts that will be in there somewhere.....outside ones are easy from the bottom the inside ones are a pain....Klass maybe call this one 18? Peter

[peteeng]

Hi all you critics out there - hope your comfy in your armchairs So its Milli no17 and I added bolts to the column foundation to see how that goes. Bolts are fine. Stiffness is still up so that's great. So the stats are in N/um (0.001mm = 1um) X 29 Y 63 and Z 107 huge!! M6 bolts done up to 20kN. The foundation plate is 6061-T6 aluminium 16mm thick and will be screwed and bonded to the CSA column. I have to think through this a bit. Do I have parts machined then glue them on then have them finish machined? Do I build jigs and bond and set all at once? Do I embed parts then machine? Hmmmm? But the Z 107N/um is interesting so I'll cut the depth back on the column a bit. I made this column a little bigger then No16 to compensate for the bolting. I could make the foundation plate out of steel. It will be stiffer and easier to bond. Details for later. Next is to bolt the arm and I may flip the square cars to flanged.... No17 also has a lightened base...

[peteeng]

Evening all - been working on getting the parts bolted together. Milli is coming together. I've adjusted the width of the parts and included the car holes so I know they work. For instance I've checked the insertion height of the bolts. They are tight but 25mm should be fine vs the 30mm bolts I have at the moment. The bed for the table rails is now really big and I have to decide if I make this all one piece which would be stiffer, Perhaps I use aircrete for the base and glue 10mm Aluminium each side to make it really stiff? I'll get the parts bolted together maybe even the rails bolted on then start FE again. Then I can get back to the motion parts. Ballscrews arrived for YaG today from Shanghai, going out to check them now even though its late.

I wanted to decide between using 25mm aluminium or 16mm steel for the subplates. 25mm al is 30% stiffer and 54% the weight, But threads in aluminium have a habit of stripping, full depth threads into the alloy should be fine.. Peter

[ardenum]

The one thing you should not make weak is the bed, why aircrete?

[jaguar36]

So the stats are in N/um (0.001mm = 1um) X 29 Y 63 and Z 107 huge!! .

Do you have any thing to compare that too? I've looked around and I thought I found some analysis on a Haas minimill, but I can't find it now. My current design is at about 50N/um but I have no idea if that's good or not yet.

Also how are you handling the stiffness of the linear guides? I've found a couple charts, but nothing comprehensive enough to really use.

[spumco]

Peter,

Coming along nicely, thanks for the continuing updates.

Just noticed something now that you're adding details to the motion bits... I don't see any reference surfaces for the rails or trucks. All the 'grown-up' rail/truck installations I've seen, including on my mill, have a precision edge somewhere to align the components. In addition there's usually a method of tightening the component against that edge.

Otherwise you're relying on friction only between the rail/truck and the mounting surface to keep everything aligned.

Mine uses the set screw method for the X/Y/Z trucks, and the rails are in a precision channel.

I've attached a page from the HiWin catalog for examples, as well as a closeup of my saddle to base truck (Y-axis). Note the grub screw at the left of the photo - that is what clamps the truck against the opposite surface on the saddle before the truck screws are torqued down.

(ignore the filth)

-Ralph

[peteeng]

Hi all out there - we must be having fun due to the responses!

Hi Ard - you say strength but these sort of machine elements rarely have strength issues. Strength means it could break somehow (static or fatigue) I think you mean stiffness or rigidity. Stiffness is a material property in eng speak and rigidity is a structural property. Aircrete is interesting because it is 20x cheaper then epoxy and 4x stiffer. If you put 10mm of al sheet each side of a 80mm thick aircrete slab its 1.44x more rigid and 34% lighter then a 100mm thick CSA slab. When it comes to lifting a 210kg base I look for lighter options. Milli is intended as a benchtop machine but currently a stand would be appropriate. I'm thinking of cutting it back again to achieve this.

I would make the aircrete out of CSA a bit of an experiment...

Hi Jag - Modern VMCs are better then 100N/um per axis. The attached diagram is quite old. Benchtop mills in the UKCNC group have been measured at 1-2N/um and they have stiffened them up. There are people in the forum who have measured their machines at 8-10N/um. My first router was 2N/um and cut timber foam plastic but struggled or failed with very hard timber and aluminium. Usually the X axis is the tough one as this places the Z axis in bending and combined torsion.

Bearing manufacturers publish the bearing stiffness. See attached Hiwin example. As a starter I make the cars aluminium, thats about the right stiffness to simulate them.. Also unless you model connections including friction and bolt preloads a "bonded" FE model could be up to 50% efficient. So your 50N/um maybe 25N/um in reality. a 50N/um model is very stiff. Well done. Is this in its worse geometric condition? If you search the forum for bearing stiffness you will find my discussions on how to do this in FE.

Hi Ralph - I develop machines in layers. This layer is just the structural one. I've just started figuring out how to bolt it together. I also have to add gutters for the ballscrews and bearings etc etc. I've found that if you add all of those things initially you will be changing them regularly and all those details take a lot of time. So once I get the bones of the machine right the rest will come. I also have to be careful with cost. This is an entry level Maker grade machine and as you step up the milling and accuracy details the cost escalates. This is a production machine not a one off so every cost down that does not impact intended performance must be considered.

Thanks for participating... everyone. Peter

[ardenum]

Here's an idea, instead of having a precision edge or surface on the machine itself, have the precision machining done on the mold part, align the rails there and then just cast the column around it and the anchors. You'd still be left with some adjustment options on the rails due to bolts being slightly smaller then the holes in the rail. starts @1:57

[peteeng]

Hi Ard - Yes that is the object of casting. Do it once, cast it many times. However there is a trade off to the cost of the mould. Many details more cost. I've been involved with moulds from a few $$$ to 100,000's of $$$ and I could easy see spending $10,000 or more on a fully detailed Milli base mould. $$$ I don't want to punt on selling one or 2 machines. So unless I get a queue of presales I'll have to wait to decide how far to go with details in the mould once I get there....Regards Peter

[jaguar36]

Hi Jag - Modern VMCs are better then 100N/um per axis. The attached diagram is quite old. Benchtop mills in the UKCNC group have been measured at 1-2N/um and they have stiffened them up. There are people in the forum who have measured their machines at 8-10N/um. My first router was 2N/um and cut timber foam plastic but struggled or failed with very hard timber and aluminium. Usually the X axis is the tough one as this places the Z axis in bending and combined torsion.

Bearing manufacturers publish the bearing stiffness. See attached Hiwin example. As a starter I make the cars aluminium, thats about the right stiffness to simulate them.. Also unless you model connections including friction and bolt preloads a "bonded" FE model could be up to 50% efficient. So your 50N/um maybe 25N/um in reality. a 50N/um model is very stiff. Well done. Is this in its worse geometric condition? If you search the forum for bearing stiffness you will find my discussions on how to do this in FE.

Well the 50N/um is a bit of an exaggeration as I'm not accounting for my Z axis, nor the bearing stiffness. I'm glad to see that I'm in the realm of something reasonable though. I wish that chart called out what specific machines they were referring to, quite the range on it. I'm still playing around with my design some, debating how much benefit I'll get increasing the cost/weight beyond what I've got.

I've seen similar charts to those from Hiwin, but they lack information on the rotational stiffness. Obviously this isn't the primary direction at perhaps it wouldn't matter with a ideal design, but its going to play at least a small part in mine.

I've

[mactec54]

Here's an idea, instead of having a precision edge or surface on the machine itself, have the precision machining done on the mold part, align the rails there and then just cast the column around it and the anchors. You'd still be left with some adjustment options on the rails due to bolts being slightly smaller then the holes in the rail.

This is a Grinder and does not use linear rails, this is a different concept and does not apply to linear rail mounting which need a precision machined edge or a pocket machined to straighten the rail when mounting

[ardenum]

This is a Grinder and does not use linear rails, this is a different concept and does not apply to linear rail mounting which need a precision machined edge or a pocket machined to straighten the rail when mounting

I'm really interested in your response. So if I cast in the anchors, the anchors aren't enough to straighten the rail? The anchors have a thread inside them to screw the rail bolts in.

in other words, I can't straighten the rails prior to casting them into the CSA/EG ? They wont stay straight after curing?

[peteeng]

Hi Jag - You do not need the rotational stiffness of the bearing. And in any case if you did you could not enter in a rotational stiffness into the program (unless you are using some king of analytical connection?). If you want to model the car accurately you need to model the "top" of the car (where the bolts go as a chunk of geometry and model the sides of the car (where the balls and sliding happen ) as another chunk of geometry. ie the car will be a three piece assembly. You then adjust the side piece materials E (youngs modulus) to get the transverse stiffness to match the charts stiffness. You then adjust the top pieces modulus until the vertical stiffness matches the charted value. Then the rotational and other stiffnesses will be close enough for modelling the global compliance of the car. This is called calibration.

Hi Ard - I think it would be a bad plan to cast rails into a machine. They do need adjustment occasionally and once cast they are there for good? Just cast the lands, gutters, pockets or features you want.... Peter

[peteeng]

Morning ALL! - Some people have thought Milli was a BIG machine so I'm showing a drawing here so its now in perspective. Its only 840mm high and 800mm wide. Last night I shrunk some parts a bit more, will see if it loses stiffness much. Main aim is to get it bolted together so I can move onto placing the motion parts. Then it will grow a bit again....Peter

addition - run No17 in simsolid. In N/um X 16 Y65 Z 30 so dropped in the X a bit. But it will need to be wider to get bearings in there so it will gain some geometry back soon.... Peter

[jaguar36]

Hi Jag - You do not need the rotational stiffness of the bearing. And in any case if you did you could not enter in a rotational stiffness into the program (unless you are using some king of analytical connection?). If you want to model the car accurately you need to model the "top" of the car (where the bolts go as a chunk of geometry and model the sides of the car (where the balls and sliding happen ) as another chunk of geometry. ie the car will be a three piece assembly. You then adjust the side piece materials E (youngs modulus) to get the transverse stiffness to match the charts stiffness. You then adjust the top pieces modulus until the vertical stiffness matches the charted value. Then the rotational and other stiffnesses will be close enough for modelling the global compliance of the car. This is called calibration.

While that method will give you accurate stiffness in the translational degrees of freedom, the rotational degrees are just going to be random derivative of that. Assuming your X axis is along the rail, the rotational stiffness in the Y and Z axis is probably pretty close, but the rotation about the X axis is not going to be very accurate. That DOF is also the one I am most concerned about. I'd prefer to just use a set of springs instead. I now have my linear guides in hand, so I'll probably just measure them. It's also pretty irrelevant, since I am using it to compare different designs, as long as its somewhere in the ballpark it should be fine. I do apologize for hijacking your build thread though.

[peteeng]

Hi Jag - No hijack, good info for other peoples. In my case I can allow the car to slide on the rail so the material stiffness and geometry give it correct rotational stiffness . Using material as a spring or a spring element is the same once calibrated. As you say it will be good enough, there's always more then one way to skin the cat. Peter