Hi Pippin,
I'll do the transverse case for the record and Z. But in any case it would seem the base has to be thicker. Cheers Peter
Printable View
Hi Pippin,
I'll do the transverse case for the record and Z. But in any case it would seem the base has to be thicker. Cheers Peter
Hi Pippin with 1000N applied to the axes at 50mm below the collet...
No5 Mill X 21um Y 8um and Z 17um (Moving column)
No6 Mill X 14um Y 8um and Z 16um (tombstone mill) 10um per 1000N is 100N/um which is very stiff.
Mill 6 the tombstone design as you expressed is stiffer in the transverse (X) direction. the other axes are the same. The delta is much less then I expected so I think the tombstone is not worth progressing. But the machine base needs to be considerably stiffer, shall do... Peter
Hi All - I changed the base thickness from 100mm to 150mm and reran the loads. This improved the figures from No5 to No7. The tombstone is still better in X (column torsion) mode. In the No7 rebuild will try to make the column a bit torsionally stiffer, more rounder. I have to model the 750W servo motor and investigate a 750W inductance motor before I start Round2. R2 will optimise structure and machine logistics may do some generative work on it. R3 will optimise motion....
I'm pretty happy with the moving column, simple and stiff....Cheers for now Keep at what your doin.. Peter
Hi All- Something like this - Peter
Peter,
Width of table looks substantially more than available X travel. Consider whether this is optimal - if the table is not so wide might get away with 2 rails.
Column design: triangle/tapered? Commercial traveling column are often triangular (when viewed from the side)
Hi Pippin - I'm about to adjust dimensions from the spindle backwards. This series, Round 1 was built from the premise of a footprint of 1000x1000mm to see what happens. Some parts have sort of grown out of the original 500x500x500mm envelope spec. I like the three rails on the table and the side drives. Rails and cars are relatively inexpensive and in this case offer better performance vs a thick table. I also like a long table as I and others do long parts. Ultimately I want to mill gantries for full sheet machines so will require a much longer mill. R2 will rationalise a few things. R1 was investigating machine configurations to see how three types compared. I'm pretty happy that R1 did the job and gave a solid direction for the detailing... Thanks for participating Peter
OK all you lot settle down - So to summarise Round 1
R1 objective was to study three machine configs to determine the direction for the actual machine. a) lifting gantry b) moving column c) tombstone
a) was complex with duplicate parts b) was good all round c) was complex The moving column won out for minimal parts and as stiff as the rest.
I've used a round column in No8 and its a little better then No7 all round. The aim is a smallish benchtop machine. Motors were looked at and a 750W servo is the base size. This will drive dimensions of the column. The table will be relatively long and may use three bearing rails for support and two motors to drive it on its ends. This is mainly to remove the screw from the "gutter" a messy place. All steel and bonded stiffness of the machine is in the order of 90N/um which the highest stiffness machine I have designed but steel is 200GPa and the material I'm using will be 40GPa maybe 50 . If 50 then the actual machine is say 50/200=0.25 and 25/2=12N/um. Hmmm maybe more carbon fibre is needed... onwards
R2 is structural optimisation Peter
Hi All - I meant to show how I determined the round column size. The "pad" for the column is 250mm square so a Dia250 column is the current max size. I calculated the second moment of inertia of a Dia250 circle then x0.9 ie 90% of that. I then calculated the inside dia required to achieve 90% of the solid inertia. This worked out to be 55mm thick. I then checked this in CAD see attached. The trick with many of this sort of part is achieving a large section size with the restriction of how do we bolt it together? If flanged you lose real estate. Flanges have to be thick to be stiff as well. Some designs have angled pockets for the bolts. I'll jump those issues in Round 2. Peter
Hi All - I'm still struggling to find static stiffness data on small mills. Seems they are around 1N/um...
stiffness measurements cnc mk3
https://www.glue-it.com/wp/tools/mil...ine-stiffness/
Anyone volunteering to test theirs and publish results would be appreciated... I think I'll aim at 20N/um and see how that goes...Want to know what it takes to mill steel...I realise VMC start at 50 and go to 150N/um Peter
Peter,
I'll test mine if you provide test directions. I don't have force instruments, but I can probably rig up some cables, pulleys and weights and put an indicator wherever you'd like.
Mine is capable of milling steel with 3/8" cutters just fine, but anything over 1/2" chatters. Indexable face mills are not possible in steel unless extremely light finishing cuts are taken, and even then the surface finish is terrible.
Also relevant - using an R8 spindle with TTS holder system. I think this is also a source of deflection/chatter.
Mill is basically a SkyfireCNC. Note the open column in back... I know this negatively affects stiffness. Column is connected to the base with 6/ea 14mm bolts.
For dimensional reference in the photos:
Table 760mm (30") long
Spindle hole in head is 80mm
Rails are 20mm
Last photo is of the underside of the base as that may be of interest when estimating/comparing stiffness.
I'm interested in doing this as a baseline as I intend to disassemble the mill and reinforce it in a few places. Some actual data would be useful to me, and perhaps it can help you.
https://www.cnczone.com/forums/attac...d=451048&stc=1https://www.cnczone.com/forums/attac...d=451050&stc=1https://www.cnczone.com/forums/attac...d=451052&stc=1https://www.cnczone.com/forums/attac...d=451056&stc=1
Hi Spumco - Generally it goes like this. 1) place a large solid bit or round stock in the collet. Some stock is safer as its not sharp. 2) Place a dial guage on one side of the bit (some square stock turned is good so the dial is on a flat). Tie something to the bit, wire or rope and pull on that with a spring scale (you may have a luggage scale not being used anymore or a fishing scale or some weight lifting weights etc). In your case take the string to a pully and hang a weight on it. 3) measure deflection. If possible do this with two or three weights to be able to plot the stiffness. Lets say its 10N/um then to deflect a dial guage say 0.2mm or 200um needs 10x200=200N or 20.39kgf or 45.06lbf Can test in the X or Y dirn or if your keen both. You may have to screw the machine down if its free standing, I don't want you to pull the machine over on top of you! Rough figures are fine better then nothing...5, 10 20 and 30kg may be good? Thanks Peter
In your case you want to use a set up that is repeatable so you can compare post mod. If you use 3 loads you plot loads vs deflection and draw a line through these to get stiffness. It won't go through (0,0) so do not include (0,0). It can be charted in excel to give you its stiffness and best fit line....
Will do. I'll report back when done.
EDIT: I'll borrow a fishing scale tomorrow. Any pully system I cobble up will have deflection in the pulley mount as well and I'm not sure I could accurately compensate for that... unless a second DTI on the pulley shaft is ok?
Hi Spumco - you don't need to compensate for the force side. As long as you have as an accurate as possible force number. The important thing is to measure the deflection as accurate as possible as its small deflections so a dial guage (or gage) is best. Do the best you can and it will be good enough. If you do it accurately you will find it "strain stiffens" ie small loads will move it more then big loads initially. This is the bearings, loose bolts and connections etc tightening up as the load comes on. So apply a small load and measure and release, then a bigger load and release and see if there is some "preload" in the machine. Then apply an increasing series of loads and measure deflection. If you can also measure it as its unloades that's good as well as this will indicate if anything is loose as the unload chart will be different to the loading chart. The load/unload deflections should be the same... (but often they are not and then it has to be sorted) Cheers Peter
GooooD Evening - I've started No8 Round2. I'm going to approach this different to usual. Usually I build many parts and test and review and review etc.I suppose that's because of the sheet metal medium. With this design I'm going to start with large block geometry and whittle it away. I'll also do some generative design in which the computer will figure out what "ideal" shapes are vs me guessing. I'm aiming at a 20N/um static stiffness in all directions. So I built a block version of No7. The blocks use the extant geometry available from version No7. Using 1000N as the design load means we are looking for a deflection of 1000/20x0.001 <0.050mm
Because I'm using composites I looked back at some of my test figures for high stiffness fibreglass laminates and found I could do 35GPs which is half the stiffness of aluminium. If I use CF I'll get around 60GPa using a simple laminate and 80GPa using a bending optimal laminate. My plan is to use glass thick skins then aluminium grit or steel fibre to fill. I'm setting up two tests to determine these materials modulii at the moment.
So I gave the blocks the modulus of 30GPa to see how they go.
X axis 0.070mm
Y axis 0.034mm and Z axis 0.041mm so the 20N/um is achievable bar the X axis. I need more stiffness in torsion to achieve this. Use geometry or material. Perhaps this is the case for an aluminum tube?? Al 70GPa... or a tube 1.2x bigger is twice as stiff so maybe enlarge the 250x250 column to 300mm.
generative design means the computer uses an algorithm and removes low stress material from the block. It then reruns the solution and removes more material. You can set say a 50% removal rate. This is also called an adaptive solution. So I think I'm in the ballpark and need to set up the first generative run in Inventor. I boolean the blocks together, export from Alibre as a step file, import into Inventor and run the generative solver.... be back soon... Peter
the wonders of simple models - I made the column 300x300 and the deflection dropped from 70 to 44um. all good time to pack up... Peter
Hi All - I did some generative work this morning. I haven't read up on the background to how Inventor and Fusion do this solutionbut I have read a paper by the solvers author before autodesk bought it. They are generally stress or strain driven which means the answer may not be as stiff as I want it for machine design. Just because its strong does not make it stiff. So I ran the first one as steel and then as aluminium. Both answers were the same so this means the solver is modulus independant which points at a stress domain calculation. So in X (column torsion) it left the column intact which is what I did last night by making it bigger. The Y axis (push pull) it chopped away quite a bit of the top of the column. The Z axis it chopped out the middle of the column which I thought was odd as then the column had no shear path but then I realised I had it starting as hollow so it couldn't leave a web anyway so maybe fill in the hole and run again.... The mass reductions are always about 40% which must be a solver setting. The screen shot mass reduction should be expressed differently. None of the cases touched the outreach much. So I think we are on the right track. I may make the outreach (does anyone know the correct term for the this part? overarm or arm?) bigger. So I'll look at it closer tonight. There are some optimisation tools that will start from the loads and restraints with nothing in the middle sort of additive optimisation vs subractive. I suppose you could start with a huge block and see what happens that maybe interesting... I want to build new rails and cars that have less features for this Round. The supplier models have all the small fillets and rads in them. These take time to solve and make the model more complex for setting up and checking. Cheers Peter
Hi Peter, why not a fixed gantry, moving table solution, if you are worried about flex, that seems the way to go.
Hi Klaas - That requires a saddle and a tall Z axis, both have issues that the moving column fixes. Every configuration has pros and cons. A tall fixed gantry with a high Z has lots of flex as well. Cheers Peter
With saddle you mean the tight planar spacing of two axis in regards to screws and rails as gantry mills (X and Z) and common vertical mills (Y and X) have, yes?Quote:
Originally Posted by peteeng
There was a gantry design where the Z axis was running through the gantry like a plunger, getting rid of the saddle, but mounting sounded complicated..
Hi JTS - The saddle on a machine is the part that allows two axis to move at once 90deg to each other. On a gantry machine this is the part that allows the X and Z to move. It means the Z axis is pushed out from the gantry increasing the tool leverage. Having 8 bearings in a tight space creates design difficulties if your trying to make things compact, With a twin gantry having the Z down the middle you still have a saddle maybe two. All designs have +s and -s the designer has to wade through these and decide the "best" path at the time....Peter