[peteeng]

Hi All - Back to the config face off. I added a conventional lifting arm config to the block models. Doing this made me realise that the worse case arrangement for this config is different to the worse case of the others. In the C column config the most compliant position is at the top of the column. These models have Z=250mm with a 100mm nose clearance to the table. The last model had the spindles down so they represented the longest loadpath, which is usually the most compliant. So in this case I lowered the moving column and the gantry so it had the same set up height as the std column design. I also did not include bearings on the column or the XY table so I expect it will be more compliant again then this modelling accounts for.. Envelope please:

1) X dirn - std column 4um mov column 7um and gantry 6um
2) Y dirn - 4um 4um 1um
3) Z dirn - 5um 7um 2um

So if we say the std column is a bit overstiff then the gantry wins again in this cofig.... Peter

some notes -
1) I made the arm on the moving column model 100mm wider vs the last round
2) The std column model weighs 1187 but does not have the XY bearings, saddle etc. The moving column model weighs 1330kg and the gantry weighs 1305kg. Al parts are solid steel...

[peteeng]

Hi All - I added the bearings to the arm on the std column model but the deflections remained the same. Also note that the column is considerably wider then usual. Seems to me the high rail gantry wins in general ... Seems Datron figured this out...Peter

[peteeng]

Evening All - I built a new gantry model which had bearings and features that could be adjusted easily. I then set up 4 freedom cases. I constrained the model at the bearings in 4 cases. So I had a Z axis support, a saddle support , a gantry support and the table support. This tells me how far each component contributes to the total deflection of 12.7um at 1000N or 79N/um in the Y dirn. Can do this for each dirn to establish those as well. These are called structural loops...

z axis 4um
saddle 1.8um
gantry 3um
base 3.9um this equals 12.7um so all good there..

So the Z axis and base need attention. Looking at the video I see the base needs to be thicker. My target will be 100N/um then I start lightweighting... So if I halve the Z axis and base deflection I'll be there. Peter

By the way the rigidity of a member is proportional the the 4th power of the dimensions. So if you make something 20% bigger it becomes 2x as stiff. 1.2^4=2.07 if you increase it 30% its 2.85x and 50% its 5.0x stiffer. Take advantage of geometry whenever possible... So the 220mm wide z axis needs to be 220x1.2=264mm wide to double its stiffness.

Edit- I upped th z axis to 250x250mm and it deflected 11.4um total ie 88N/um
so z axis 3.2um saddle 2.0um gantry 3.1um and base 3.1um so now I strat lightweighting...

[peteeng]

Hi All - Here's the lightweighting history so far. Still cutting away stuff. I only mod one component at a time so I know what's what. Peter

[joeavaerage]

Hi peteeng,
I am familiar with calculating the first moment of inertia of ballscrews and that is to the fourth power of radius. That stiffness should follow the same proportionality should be no surprise.

What that overlooks however that it is to a certain extent geometry dependent.

For instance in your previous M1 studies you took the both the base and the column to be rectangular. There is significant loss of stiffness across the flat or thickness dimension. Had the section been made square the
stiffness would be maximized.

To take it to is logical, if absurd extreme, imagine a column very (infinitely) wide but very thin (infinitely) thin.....the same area but will bend (in one direction) at near zero load and yet be infinitely stiff in the other.

I understand that your gantry design requires a large square base, but that is very much the wrong choice for a column design. Were the same volume of material devoted to your gantry base be devoted to a square base
for the column design I rather suspect the comparison would be closer.

The bottom line is that stiffness is highly dependent (to the fourth power) on outer dimensions but also on the geometry of the material placement within those boundaries.

Craig

[peteeng]

Hi Craig - As I have stated before any config can be made to any stiffness as long as you have the real estate. So yes making the machine base deeper will make it stiffer in the Y dirn but maybe not on the Z direction if you make it narrower. I am not trying to prove which config is "best" that's a pointless exercise. I prefer the high rail design because the walls form a natural bucket, all the mechanics are up out of the muck and the job stays still so I don't have to account for the max payload in the motion calcs so the drivetrains can be optimised easier. if someone's super keen on a std C frame design then go for it. To each their own.

This current exercise is my first run at using block geometry then parring down vs starting somewhere and stiffening up. So far its been a smoother and more understandable approach then the usual upward approach... The bounding geometry has to be the maximum stiffness geometry, anything done within that box has to be less stiff if its the same material. So any framework within that box has to be less stiff.. So its a good way to get a line in the sand above your target stiffness because as you lightweight it you will lose stiffness, unless you gain geometry. That has happened twice now to recover the lost rigidity of removing material. I made the saddle wider and it made a significant improvement. Peter

So Round 6 shows I have dropped in stiffness from 11.8 to 12.8um.. or I'm now at 92% solid stiffness. The weight has dropped to 63% would like to drop that further. Seems the walls are the culprit and I'll change those again. Slow but methodical process...

[peteeng]

Evening All - The gist of the matter is I have hovered around 80N/um for this design. I have reduced the weight from 2720kg to 1706kg so that's a loss of 40% would like a little more. But its a design exercise not a real machine. The trends are as I reduce mass I have to make things bigger to maintain the target N/um. Next step is to make the ribs thinner, the base thicker and the gantry bigger. But I have learnt enough for now. Got to get back to my real machines!! This process is much better than small to large process. Grow the solid geometry until its past the machine target geometry then hollow things out to attain target weight if that's the target... Onward Peter

[peteeng]

Evening All - A little bit more work at Round 10. Couldn't help myself. Wanted to get to 50% weight ie 1360kg but got to 58% at 1578kg. Scored 95N/um would like 100!! But that's enough I can see it has to gain BIGGNESS to get there. That's OK but enough time has been spent chasing this Grail... Peter

[joeavaerage]

Hi peteeng,
good work. It seems that 1500kg might be about right for that machine. 100 N/um is pretty damned good and would be a very good match with various industrial machines on the market.
Is it a surprise that such a machine might weigh 1500kg? I think not.

At the start of this thread it seemed that your goals were in the region of 20 N/um and maybe an all-up weight of 500kg.
This line of work you are now pursuing is a very much more capable machine, and to be honest that you can get that sort of stiffness for 'only 1500kg' is I think
quite a feat.

Craig

[peteeng]

Hi Craig & et al - I ran another couple of rounds. I'm at the point I think where to gain stiffness I have to gain weight. So Round 12 got to 107N/um at 1684kg. I made the base deeper and played with the webs but the stiffness stayed the same. So I reverted that approach. Then I made the saddle wider and that really helped with no major weight gain. Now it would seem I need to do the other axes in loops to see what they are doing. But the global stiffness is X=107N/um. I'm rerunning the model so will get back here soon. Peter

The history of the model is that the original parts are based on using 30mm rail with D16 screws. This gave me the minimum separation distance between rails. So the starting geometry for parts was around the 200x200mm solid mark. The Z axis grew to 350mm wide x 250mm deep and 35mm thick. The gantry is 220mm square by 35mm thick and the walls are 175mm wide by 500mm high with 20mm thick ribbing. The base is 200mm thick with isogrid type ribs 20mm thick. Its footprint is 1200mm wide by 800mm deep.. Envelope is X450 Y250 Z 350mm.

If it was made in aluminium it would be around 33N/um but only weigh 582kg. I will thin the steel one out to 1/3 its thickness and see how its stiffness drops. The Y dirn stiffness is 71N/um and I can see that the gantry has to be bigger to improve that number. The Z is failing, not sure why, it has solved every other time. Just solved and Z is 172N/um... I ran the GD on the gantry and it created a truss where all the hard pints are connected with a straight pipe. It created other solutions but that one was light and stiff. This is typically how you do it manually, like building roll cages and motorbike frames... I have reached a reasonable solution in 12 goes. That's far away better then my initial efforts that may have taken 50 goes to get somewhere...

Taking the gantry at 220x35mm the solid rigidity is 3.904E10 mm4.GPa. The hollow is 3.061E10 mm4.GPa so is 78% of the solid rigidity... Milli is meant to be a benchtop machine but this one is way over that category. So I'll have to look at what I've learned and try to back it down to a benchtop scale... or renovate under my shed so it could live there? But then it floods every 10 years or so and I wouldn't want a machine to go submarining.... like lots of things did 2 years ago here... Peter

[joeavaerage]

Hi,
have designed a spindle monitoring board that can interlock the VFD from operating or at need send an Estop command to the CNC controller.

The board monitors the three Tool-in-Place switches within the spindle, the PTC temp sensor in the spindle, it monitors the NTC thermistor that is detecting the presence of coolant flow (in the coolant return line),
the Tool Change Request switch and monitors if Mach is Enabled and finally it monitors the VFD and can thus tell if the spindle is still powered and spinning.

I have used a small AVR microprocessor for logic and monitoring purposes.

The board can interrupt the VFD ON/OFF signal, can signal an over-temp event and Estop Mach should it be required. It also controls the Tool Change air solenoid, the Seal Air solenoid and the Taper Blow air solenoid.

Fits on a 80mm x 100mm board. It is powered by a 30VA transformer, 230V input 18VAC output or 24VDC when rectified and filtered.

Craig

[bravernoe]

Hey Peter,

That sounds like a solid upgrade! Moving the drives to the outside and extending the table should definitely help with debris management and improve your setup. Placing the drives up on the walls to allow fluid drainage is a smart move, and having bearings directly under the setup point will certainly enhance stability and accuracy.

[phomann]

Hi,
have designed a spindle monitoring board that can interlock the VFD from operating or at need send an Estop command to the CNC controller.

The board monitors the three Tool-in-Place switches within the spindle, the PTC temp sensor in the spindle, it monitors the NTC thermistor that is detecting the presence of coolant flow (in the coolant return line),
the Tool Change Request switch and monitors if Mach is Enabled and finally it monitors the VFD and can thus tell if the spindle is still powered and spinning.

I have used a small AVR microprocessor for logic and monitoring purposes.

The board can interrupt the VFD ON/OFF signal, can signal an over-temp event and Estop Mach should it be required. It also controls the Tool Change air solenoid, the Seal Air solenoid and the Taper Blow air solenoid.

Fits on a 80mm x 100mm board. It is powered by a 30VA transformer, 230V input 18VAC output or 24VDC when rectified and filtered.

Craig

Nice! F you want to reduce the parts count, use a ULN2003 instead of discrete tranies, diodes, etc. they interface to logic really well and have flywheel diodes built in.

Cheers

Peter


Sent from my iPhone using Tapatalk Pro

[joeavaerage]

Hi Peter,
the solenoids are typically 6W and therefore I'm expecting 250mA per solenoid circuit, probably a bit on the high side for a ULN2003.
Additionally I already have these Darlington transistors in stock, I use them in power supplies regularly in automotive instrument boards, so they have
come to be a 'goto' part.


I have made a few changes to that board. Firstly I corrected a mistake with the LM713 wiring that I am using as a constant current source for the PTC monitoring. I have also
used another output to control the coolant pump VSR. I have included a LED to indicate the status of the Interlock Relay, it is when all said and done the most important function
of this board. Everything HAS to line up (temp, coolant flow and Tool-in-Place) before the interlock is released.

Craig

[peteeng]

Morning all and sundry - Been interested in another thread on a small mill. It has brought back memories of an exoframe design I wanted to look at. So I built a model this morning and its very good. All axes over 100N/um and did it in 2 goes. Plus it weighs only 796kg. My high wall design weighs 1300kg. The theory goes a machine needs an enclosure so why not use that geometry as structure. Seems to be a good plan. I can see I can loose weight and gain stiffness. Its footprint is 1000x800mm and its 400x400x300 envelope. Its bolted steel. From this rough model I can see how to consolidate the plates better. But for my composites I'm worried about the mechanics in the muck.... Will think it through... Peter

X 103N.um Y 123N/um and Z 143N/um steel

[pippin88]

Problem is access for the operator to the work area / work piece.

[peteeng]

Hi Pippin - Same problem once you put an enclosure on so make access same as having an enclosure. The "box" can be taller so the hole is bigger giving good access. Maybe the front of the "box" is not needed, its not in the NEO. Next build I'll make it separate parts so I can figure out what bits are doing what... Peter

edit - its down to 743kg with the deflections under 10um so its better then 100N/um all round...

[peteeng]

some more machine fun - Peter

https://youtu.be/q3kw75XCj_4

Double Column Machining Center - HD Video produced by ???? S&J Corp - YouTube

[joeavaerage]

Hi,
Have finished my Spindle Monitor/Interlock board.
Quite pleased. I found one PCB error, which I was able to correct by soldering a link wire....no re-make of the PCB required. The microprocessor is programmed to monitor the spindle temp, Tool-in-Place switches
and the VFD, and all seems to be working correctly, scarcely a hundred lines of C code.

Especially pleased how the NTC idea worked out. The NTC (negative temperature coefficient thermistor) gets warm (approx 60C in still ambient air) by self heating with some 70mA passing through it.
When it is immersed in water, or cooled by return coolant flow as will be in my machine, the temperature drops and that changing resistance is detected by the monitor, in this case the microprocessor.
It responds to the application of cooling flow within a couple of seconds, and deactivates in absence of cooling flow within about five seconds. Simple and effective and cheap, if memory serves the NTC itself was about
50cents when I bought it!

Craig

[peteeng]

Evening all - Some more stuff on HP concrete. I am contemplating 3D printing a part "skin" in plastic then inserting the inserts then filling with grout as a test part. Hmmm. Peter

Fiber Reinforced Concrete | Synthetic Fiber Reinforcement | BarChip


https://www.mmsonline.com/articles/i...ol-moonshot(2)