[nikken]

This is going to be a long a topic on the build of an heavy CNC milling machine for steel with an epoxy granite frame.

Here are the main characteristics of the machine:
- Courses:
- X 950 mm (-150 mm due to ATC position)
- Y 900 mm
- Z 550 mm
- Mechanical:
- Rails: INA RUE 45 on all axis (roller bearing)
- Ballscrews: 40/50 mm diameter with pitch 5/10 mm, double nut or preloaded one C1/C3
- Spindle : BT30 spindle with 2.2Kw Servomotor (will upgrade later the spindle)
- 4-5 Axis: Harmonic Drive SHG 50 and the remaining of gear/worm wheel from a Nikken tilting table
- Electronics/Electrical:
- Servomotors on all axis: Lichuan servomotors with 17 bit absolute encoder
- Electronic: Mesa cards with LinuxCNC
- ATC: 40 tools BT30

Overall design is a modified gantry inspired by Hermle C42 machine, some of the details are inspired by the previous work of Nucky (on cncecke.de) and badhabit to just name a few. For the pictures, I attached the simplifed CAD model currently being used for simulation, the epoxy and granite (Silimix 282) (overall planning to cast 3000kg for the frames/gantry, the electrical cabinet (started to do some testing/development), the rails/carriages/ballscrews.

Most of the elements are already bought/locked what remains is the overall construction, casting of the frames (which will be in August).
I'm still going through the optimization of the frames and rest of mechanical elements.

I will add some dedicated posts for specific topics (assembly and calibration process, electronics, ATC, frame, casting process, etc).

[peteeng]

Hi Nikken - Ambitious project. The high rail design won't disappoint. How are you going to "machine" the rail foundations? so they are flat and parallel? You have 6 cars on the Z axis? I don't think you need 6 but what is your logic on that? Peter

[routalot]

Agreed.Is this a first build?What cam software do you plan to use?

[peteeng]

Hi Nikken - You have the saddle cars on top of each other. This is structurally very good but from a assembly/maintenance and servicing point of view it's not good. (look at commercial machines they do not do this) You can't check bolt tension if needed (without pulling down the machine) and you have to be absolutely sure the buried bolts are done up correctly (and in my experience the bottom cars on the Z axis are the likely ones to come loose) Be liberal with the loctite! Peter I also see a drive nut mount that is awkward. Have you thought assembly through well enough? Peter

[nikken]

Hi Peteeng and Routalot

Not directly a first build, I have build machines, robots and automation before, and smaller cnc/3d printer as well but not something as involved and massive as this one.

For the software, LinuxCNC for machine control (as it allows me to have some flexibility for custom macros, multiple interfaces (Bekchoff PLC, Mesa, ethercat)) and FAO either SolidCam or Fusion 360, we will see how both can handle 5 axis simultaneous when I get there.

Regarding assembly/calibration of the machine,

Hi Nikken - You have the saddle cars on top of each other.

Yes definitely this is just for the simulation at the moment, as I'm designing it, I'm always thinking about how to assemble it, finetuning the design and writing the SOP for assembling as well. So this is a consideration, most industrial CNC are using a saddle which is much thicker, in a mix of welded plates, my current leading solution is to extend the saddle on the Y axis and remove the overlapping of bolts for carriages of the Y and Z axis, so it can be adjusted easily later on.
For the number of carriages on the Z axis, 6 is definitely overkill from an effort and moment perspective (current specs of my carriage is 114000 N dynamic load rating), so I might replace the middles one with the RUDS damping carriage (I have 2) https://medias.schaeffler.us/en/prod...150-d/p/835053 which would help with decreasing vibration for the Z axis. I have 13 RUE45 carriages available at the moment, so I can use 6 on Y or Z axis if I want to, but could keep 2 as spare as well.

For machining the rail foundations (Short answer), on X and Y axis, the rails will sit on epoxy resin filled with steel fillers (DWH310 FL), and be formed with a granite straight edge (DIN876/00 of 1200 mm) and precision level for the alignment. There are more steps also to guarantee planarity between the 3 rails of the X axis for instance that I will detail in upcoming posts.
A shoulder will be created on one side as well with the granite straight edge, and on the other side and be pushed against with alignment pin (see attachment).

[peteeng]

Hi Nikken - The plan sounds good. I have made steel reinforced epoxy myself and its quite good. Keep it very thin (<1mm) and it will work well. Wax your reference very well (or anything you don't want to get stuck) . If you use paste wax you will need 5-6 coats. Would be a shame to have a stick up on the granite! 6 cars are usually used if the part has a long span and consequent deflections are unacceptable. Simulation will guide you there. Be careful with modelling carriages without their holes and bolts. Its subtle but has caught me out a couple of times. Even the bolt heads can create a clearance problem that unless modelled for visual checking can waste a lot of rework time! Peter

images are my latest router. Mainly laminated aluminium and plywood. For a few years I have gone off the rails in front of the gantry config and have made 90deg saddles with the gantry rails on top not in front. This has opened up the area so access is good for all bolts and has proven to be very stiff. Mori do similar on their mills. The AL saddle is my current std saddle. I'm working with a laser company at the moment to try to make a saddle (then other parts) that does not warp when welded. Laser welding will open up so many possibilities! I have looked at casting saddles in UHPC and have tested a few materials but not happy with its stiffness. Look fwd to your progress.

[joeavaerage]

Hi,
for a machine of this size I think BT30 is too small. BT40, or HSK50 or HSK63 would be more in keeping with the capacity of the machine.

Conversely I think 40mm and 50mm ballscrews are too large. The rotational inertia of a shaft increases as the fourth power of diameter.
I use 32mm diameter 5mm pitch C5 BNFN THK ballscrews, and they are almost over large for my machine. They have a moment of intertia of
about 8.8 x10^-4kg.m². I can accelerate at 0.27g with 750W Delta servos. Note that the rotational inertia dominatnes
the momentum (80% while the servo and load are 20%). If you use 40mm ballscrews (of the same length, 750mm) then the rotational inertia
will be of the order of 21.5 x10^-4kg.m². With 50mm ballscrews its even worse at 52.5 x10^-4kg.m².
These would suggest you would need 3.5kW servos or more just to accelerate the ballscrews? Is this what you had in mind?

With ballscrews it is VERY possible to go oversize and then cause yourself real trouble down the line. Is there a reason that you want such outsize ballscrews?

Craig

[nikken]

Hi Joeaverage,

for a machine of this size I think BT30 is too small. BT40, or HSK50 or HSK63 would be more in keeping with the capacity of the machine.
Craig

Yes BT30 is just to get started for now, as I have some limitation on the power consumption, but yes BT40 is more appropriate for the size of the machine, the ATC is also planned to support BT40 as well.

Hi,
These would suggest you would need 3.5kW servos or more just to accelerate the ballscrews? Is this what you had in mind?

With ballscrews it is VERY possible to go oversize and then cause yourself real trouble down the line. Is there a reason that you want such outsize ballscrews?
Craig

For the ballscrews, I have 2 in 40 mm, one in 50 mm (I have found very affordable occasions for those), they will be paired with 2 1.5kW and 1 2.3kW servomotor respectively, from the computation I made I should be able to run them at 0.25g. Total inertia of 66 x10^-4kg.m² with a max torque of 27Nm.
I agree 32,40 mm would have been slightly better but couldn't find what I needed for the same price. If I find better options, I can still switch and replace before locking the frame/gantry.

[nikken]

Hi Nikken - The plan sounds good. I have made steel reinforced epoxy myself and its quite good. Keep it very thin (<1mm) and it will work well. Wax your reference very well (or anything you don't want to get stuck) . If you use paste wax you will need 5-6 coats. Would be a shame to have a stick up on the granite! 6 cars are usually used if the part has a long span and consequent deflections are unacceptable. Simulation will guide you there. Be careful with modelling carriages without their holes and bolts. Its subtle but has caught me out a couple of times. Even the bolt heads can create a clearance problem that unless modelled for visual checking can waste a lot of rework time! Peter.

Thanks for the pictures, I have seen some of your design on the other threads here. For the 6 cars, it's mostly due to the large cantilever when the Z axis is at the bottom, up to 550 mm, current simulations are showing a lower rigidity there, not sure how larger machining center are dealing with it, as they have similar or bigger cantilever. My main guess would be they use volumetric compensation to match things in SW. Not sure if anyone has more details there.

For the CAD, I have 2 configuration, one for simulation (simplified geometry, less elements for faster meshing and simulation), one complete with all the details for assembly, designing the molds, all the inserts required, integration of the elements, electrical, pumps, ATC, etc so I can plan everything in advance before casting. Will share more pictures in the upcoming days.

[joeavaerage]

Hi,
do not use max torque but rather rated torque.

For instance the rated torque of my 750W servos is 2.4Nm. It is that value that I used and found 0.27g acceleration. Max torque is 7.1Nm, and that could in turn generate acceleration of 0.75g,
but that is not my expectation in service.

I would anticipate 1.5kW servos have a rated torque of 4.8Nm and a 2.3kW servo of 7.4Nm? Do those numbers sound correct? I have assumed the same 3000 rpm rated speed throughout.

I think it would be wise to post the actual numbers (diameter, length and pitch of each ballscrew, the inertia of the armature of each servo, rated torque of each servo,and the anticipated max axis mass of each axis) and do the calculation.
Once you've poured 3000kg of epoxy its too late to realise that you've made a mistake. Do the calculation now so you have some certainty about reaching your goal at the end.

Craig

[joeavaerage]

Hi,
does this sound right?

an axis equipped with a 40mm ballscrew of 5mm pitch has a total inertia of 66 x10^-4 kg.m² direct connected to a 1.5kW 3000rpm servo of 4.8Nm (rated).

Torque =J_total. dw/dt
dw/dt =4.8 /66 x10^-4
=727 radians/sec²

Linear acceleration = angular acceleration / 2 PI x pitch
=727 /2 x 3.141 x 0.005
=0.5788m/m²
or 0.06g which I think is a little disappointing.

My machine in standard tune has linear acceleration (all axes) of 0.25g, but I find this just scary fast, and my normal tune for all my machining is 0.15g. Sorry to say but I think 0.06g is a little anemic.

Craig

[nikken]

Hi,
does this sound right?

an axis equipped with a 40mm ballscrew of 5mm pitch has a total inertia of 66 x10^-4 kg.m² direct connected to a 1.5kW 3000rpm servo of 4.8Nm (rated).

Torque =J_total. dw/dt
dw/dt =4.8 /66 x10^-4
=727 radians/sec²

Linear acceleration = angular acceleration / 2 PI x pitch
=727 /2 x 3.141 x 0.005
=0.5788m/m²
or 0.06g which I think is a little disappointing.

My machine in standard tune has linear acceleration (all axes) of 0.25g, but I find this just scary fast, and my normal tune for all my machining is 0.15g. Sorry to say but I think 0.06g is a little anemic.

Craig

Hi Craig, thanks for looking at the computation, yes definitely 0.06g would be anemic.
I'm using a 40 mm ballscrew with a 10 mm pitch, connected to 1.5kW servo at 1500rpm (rated torque of 10Nm and max torque of 25Nm, rotor inertia of 1.94 x 10 ^-3 ).

For my computation, in this case I was getting for 0.25g, holding torque 9Nm, and peak torque during acceleration of 15.7Nm. (with a load of 500kg, and axial load of 500N).

I was not directly assuming max torque for my computation, but was accounting to go slightly higher than rated torque in short burst (Similar to S1, S3 rated we can find on some servo). Is that assumption not true ?

[joeavaerage]

Hi,
I think we should have a look at that calculation, it does not sound correct to me.

First ignore th 500N thrust....its absolute peanuts. If you have a 10mm pitch ballscrew (independent of diameter) and apply 1Nm torque the resultant thrust is 625N.....so your 500N thrust requires a fraction
under 1Nm torque by the servo. Forget it....it is just confusing your calculation. There are too many ways to make an error in calculation without injecting some small extraneous term to confuse the hell out of
everyone.

The 500kg axis mass cannot be ignored but whats the bet its less than 10% of the overall momentum, and could at a pinch be ignored without undue error. Don't believe me? Then lets do the whole calculation properly.

What is the length of your 40mm ballscrew?Also please confirm it is 10mm pitch?

What is the total mass of the axis plus workholding plus part?

The total inertia is found by:

J_total=J_ballscrew +J_servo+J_linear

J_servo=19.4 x 10^-4kg.m²

J_linear is the moment of inertia of the linearly moving axis but in terms of rotation of the ballscrew.
J_linear = M p²/40 where M is a total linearly accelerating mass, p its the pitch (in meters) and 40 Is near as dammit (2 x PI)²

So with your numbers of 500kg and 5mm pitch:
J_linear=500 x (0.01)²/40
=12.5 x10^-4kg.m²

So as you can see even a 500 kg mass actually has very little momentum relative to both the servo armature and the ballscrew.

Lets assume the 40mm ballscrew is 1.5m long:

mass = (0.04/2)² x PI x 1.5 x8000
=15.5kg

J_ballscrew= 0.5 x mass x r²
=30x10^-4kg.m²

J_total= (30 + 19.4 +12.5) x 10^-4kg.m²


=61.9 x 10^-4kg.m²


Just so you can be clear 48% of all momentum is in the ballscrew alone, while 31% is in the armature of the servo while only 20% is the linear axis. This calculation surprises a lot of people, the rotating parts
(servo armature and ballscrew) totally DOMINATE (79%) the momentum equation. You could double the axis mass to 1000 kg and still not change the acceleration that much. The truth is that the ballscrew and armature are
rotating at very VERY high speeds while the linear axis is a slow as a wet week. Sure its much heavier, but speed wins over mass any day.

dw/dt= 10/61.9 x10^-4
=1615.5 radian/sec²

for a linear accel of 2.57m/sec² or 0.257g. As I posted earlier I find 0.15g entirely enough.

Craig

[peteeng]

Hi Nikken and others - The issue with using overdrive torque values is that now you are entering thermal overload land. If you use the maximum continuous torque value this means that the motor will come into thermal equilibrium and not burn out at that output. Using a value over that torque means the motor generates more heat then it can get rid of and ultimately will overheat. If you look at the overdrive thermal cycle time spikes they are quite small.

The linear bearings have a friction co-efficient of around 0.004 so the 500kg weight is only 2kg (for X and Y axes. Z axis is being lifted so that's a bit different unless you counterbalance the axis). The seal friction is about 1kg as well. So in terms of direct forces the motor only needs to push 2 or 3kgf. But its the inertial force that is big especially rotational inertias. So to accelerate 500kg at 0.25g means it weighs 123kgf... so its the inertia that needs to be considered... and as Craig says mostly its the rotational inertia that dominates high accel drives...

These high torque conditions may never occur in practice. These calcs are dimensioning calcs to check your in the ballpark or to make reasonable decisions on the size of screws and motors etc. In some things like I used to be involved in, say a production line robot that did exactly the same thing every 2 mins, 12 hrs a day you can optimise the motors and screws to do exactly that job and can squeeze a little more out of a motor if it has enough time to cool down in the work cycle. But with a general purpose machine some headroom in the system has to be left for that day or week when it maybe needed! Peter

[joeavaerage]

Hi,
as a comparison lets do the same calculation with the same axis but a 32mm ballscrew (10mm pitch) rather than a 40mm ballscrew.

Given that the rotational inertia of a rotating cylinder varies as the fourth power of diameter:

J_32mm= J_40mm x (32/40)⁴
=J_40mm x0.4096
=12.3 x10^-4kg.m²

This is a substantial reduction given a modest reduction in diameter. Following through with the other factors:

J_total=J₃₂+J_servo+J_linear
=(12.3 + 19.4 + 12.5) X10^-4kg.m²
=44.2 x10^-4kg.m²

With the same servo the angular acceleration is:

dw/dt= 10/44 x10_-4
=2262.5 radian/sec²

Which results in a linear acceleration of 3.6 m/sec² or 0.36g.

For a modest reduction in ballscrew diameter (20%), with all other factors the same, you can increase the linear acceleration by 40% (0.36 / 0.257).
This illustrates how sensitive the momentum (and thereby acceleration) equation is influenced by ballscrew diameter.

In your case with 40mm ballscrews and a 10Nm servo gives you an entirely adequate acceleration with rated torque of 0.257g. This is about the same as my machine and yet I dial it back to 0.15g.
Unless your machine is intended for production then I would guess you'll find 0.257g any amount enough. For a production machine accelerations of 1g and higher are the norm.

Craig

[nikken]

In your case with 40mm ballscrews and a 10Nm servo gives you an entirely adequate acceleration with rated torque of 0.257g. This is about the same as my machine and yet I dial it back to 0.15g.
Unless your machine is intended for production then I would guess you'll find 0.257g any amount enough. For a production machine accelerations of 1g and higher are the norm.

Craig

Thanks Craig, yes I have a spreadsheet with all the computation for the ballscrews including the different inertia term (so I can simulate the scenarios for the different axis, and optimize a little bit the different parameters based on auctions I found), I'm well aware of the main contribution coming from ballscrew inertia itself, overall I think the numbers aligned in term of overall inertia terms and impact to acceleration, I added a few terms because I wanted to model the different situations as well (to make sure this will not be a problem), not just the main contributions. I mostly followed the details computation of NSK documents.

For the 50 mm I will see if I can find a suitable replacement before locking the design, to get a slightly better acceleration. and I'm definitely not looking for production machine acceleration for now.

On the thermal overload, I know where from a physics perspective it's coming from (thanks peteeng for the comment), but would like to understand where the numbers are coming from, and how much intermittent we can have those burst of torque, as long as we can dissipate the thermal energy generated. Delta documentation seems to describe some of the overload parameters, how long we can be at specific overload torque. To give an example you can run some of the delta A3 servomotor at 200% for 8s. https://deltaacdrives.com/Delta-ASDA...ser-Manual.pdf page 1260

I will probably add this to my list of optimization to look at once the machine is assembled and running.

[nikken]

On a different topic, for the frames I'm planning to have different pieces casted (less than 600-800 kg and less than 1400 mm in one axis because of a windows) of moving them around (I have a pallet jack, an engine crane up to 3 tons), and those will be bolted together with inserts planned in the different elements as well as position for anchor bolts. I already got a steel table 1200 by 1200 by 160 mm (see pictures)

For assembling the ballscrew support, what would you recommend, using this technique with adhesive and alignment everything in place with shim, gage? Those pictures are taken from another build thread from a german forum who built a very impressive VMC as well using EG frame.

[peteeng]

Hi Nikken - Fixturing the parts is the big debate in concrete machines there are several philosophies, I'll summarise and comment here:
1) Concrete will stick to steel or aluminium initially but over time it will uncouple from the surface. So surfaces have to be very rough to ensure mechanical keying not bonding. There are primers for metals/concrete worth researching those...
2) Steel inserts can be cast into the parts in-situ then machined to spec. This is probably one of the best routes but then you have to get massive parts to a machinist who has a suitable machine. Steel is used because it has the same expansion co-efficient concrete.
3) I feel that pre-casting in "pockets" for steel or aluminium inserts is the way to go. Then epoxy the inserts into the pockets. If you have suitable gear these can be shimmed or supported so that they are cast in the correct place and then your done
4) Otherwise bond in billets and get that to a machinist
5) German companies that make concrete machines machine the concrete so the surfaces are very accurate. I had a stonemason nearby and planned to have him machine some parts but he sold up and now I don't know another stonemason...
6) Other ways like scraping in, casting surfaces against reference surfaces are all doable if taken slowly slowly and will be as accurate as your set up and your measuring instruments allow. Good Luck Peter

[joeavaerage]

Hi,
I've always calculated using rated torque. The truth is that rated torque is rather imprecise as it is a thermal limit. As Peter suggests were you to use your servo at rated output continuously then
it would get warm, very warm I would suspect, but the manufacturer has determined that temperature is OK. Minor alterations in the servos operating environment could have a large impact on the thermal equilibrium
and therefore equilibrium temperature. rendering 'rated' output imprecise.

My own experience with servos is that 99.99% of the time they loaf around doing next to nothing and then in brief, VERY brief moments draw a swag of current. I used the analogue output of my Delta servos
to monitor the load. 99.99% of the time you could scarcely measure the power consumed, less than 50W, but then the power would peak at 750W or even higher, but for vanishingly small lengths of time.
I, in short order, realised that me monitoring servo load was a waste of time. At no time, under any circumstance, have my servos ever gotten a few degrees above ambient.

Given that our servos are cold then I suspect you could for at least several minutes go to 130% to 150% before any overheat condition was reached. For genuine overload, say 300% or more however then
the heat generation is in the windings, very localised and very intense. I suspect then that the manufacturers recommendation is correct, 8s, without risking a local failure within the servo.

Provided your servos are adequately specified for the job at hand, and I would guess your 1.5kW and 2.3kW devices are very much in the 'well specified' bracket, then you'll never have to worry about them ever again,
at least for a hobbyist machine under hobbyist machining conditions. My own 750W servos are almost outsized for my machine. Indeed when I was designing it I had in mind 400W servos. When I went to buy however I found
that I could have 750W units for only an extra $40, less than 10% extra....so I got 750W servos. In many respects I'm glad I did because like you I saw and jumped on some very good 32mm diameter ballscrews in the secondhand market.
I did not know or realise just how much the larger ballscrews would change things. As it turns out with 750W servos my somewhat oversize ballscrews are no trouble, but I probably would have been disappointed with 400W servos.
Since that time I'm much more careful about conducting the momentum calculation. Its no so much that I want or need to predict exactly the demands in service, but rather to make sure that my idea is 'in the ballpark'. If it is the servos
natural tendency to really 'give its utmost' will ensure that they do the job without complaint.

Craig

[nikken]

Hi Nikken - Fixturing the parts is the big debate in concrete machines there are several philosophies, I'll summarise and comment here:
1) Concrete will stick to steel or aluminium initially but over time it will uncouple from the surface. So surfaces have to be very rough to ensure mechanical keying not bonding. There are primers for metals/concrete worth researching those...
2) Steel inserts can be cast into the parts in-situ then machined to spec. This is probably one of the best routes but then you have to get massive parts to a machinist who has a suitable machine. Steel is used because it has the same expansion co-efficient concrete.
3) I feel that pre-casting in "pockets" for steel or aluminium inserts is the way to go. Then epoxy the inserts into the pockets. If you have suitable gear these can be shimmed or supported so that they are cast in the correct place and then your done
4) Otherwise bond in billets and get that to a machinist
5) German companies that make concrete machines machine the concrete so the surfaces are very accurate. I had a stonemason nearby and planned to have him machine some parts but he sold up and now I don't know another stonemason...
6) Other ways like scraping in, casting surfaces against reference surfaces are all doable if taken slowly slowly and will be as accurate as your set up and your measuring instruments allow. Good Luck Peter

Thanks for the long summary, so I'm planning
1) for all small inserts for rails, misc, everything is M12, or M6, All M12 inserts are based on concrete rebar, and M6 based on hexagonal spacer + screws. (see pictures attached)
3) will be used for ballscrew support
6) Casting surface with DWH 310 for rail foundations.

Also for surface, I will be adding specific foam when casting to make sure it keeps some roughness compared to the wood mold (see picture attached from the German forum). I also attached some example tests I have run to make sure inserts behave as expected, tolerances are respected for rail assembly, and finetune epoxy ratio (targeting 9% epoxy in mass, but will add some details in another post). Same with vibration, I have 2 motors to vibrate the larger pieces, I will hook up some temperature sensors and accelerometers to make sure I hit 2g (hopefully), that's what has been recommended in the few papers/thesis available online.

I plan to scrap/handlap some of the surfaces that will be in contact between the frame elements.

I will share my current process to align everything, and current metrology I have available as well later on.