[G-Spot]

Here is the drawing I gave it to optimise
this machine.....made out of 500mm thick columns.....is tenfold better than a lot of top machines.

This has a max deflection of 2.89 micrometres!

[peteeng]

Hi G - I'm always a little skeptical about twin gantries with a central Z axis. This is because then you have a parallel load paths which means one path (one side) will be dominant in practice (vs the perfect theoretic model). Then you have less geometry to do the job as only one side is doing the work & you have structural hysteresis when the load changes direction. I'd prefer to have a really large gantry that the load has to go through every time and all dirns. Having the CoG central does not help, the structure has to be UBER stiff and the CoG position does not matter if its stiff enough. Many apartment sized mills do not do the split gantry thing and they work fine. But there are split gantries out there so if it works its fine. Peter

Hi G - Please keep in mind that forums such as these are shark tanks. I can only point out possible negative issues as i and others see fit. I'll occasionally say hey that's great (if it warrants it) but most observers here only spend time on the negative stuff to be efficient. If your coping a lot of flak take it complementary as people don't spend their time on trivial stuff.... Keep at it, for instance in the linear world you will not see biased loadpaths as the solver assumes linearity. You need to include rails and cars and sliding then run a non linear model and you will see it develop a bias if its not stiff enough....

[G-Spot]

Here's a beefed up optimisation.

Thanks for that info Peter, I did this with a large skylight on wooden beams and had to replace the linear bearings with v-groove wheels running on the same tracks. Always thought is was a wood issue that you would not get with steel.

So does the rule of not having more than 2 bearings supporting the same load on the same track apply here too?

Better to just go up a bearing size than add an extra car to the track ?

[peteeng]

Hi G - You can have as many bearings as you like on a rail. They often do this on mill tables to keep the span down. But often more than one is redundant (in fact the problem then is described as in-determinate) as square cars support moments, so the load does not get past the first bearing in the load path. But if you model the rails and cars with sliding then the load distribution will be correctly calculated. You then have to include the drive screw as this then becomes part of the elastic system. Or you have to fix one bearing to allow the others to slide. Bigger bearings are stiffer plus flanged bearings allow the use of bigger bolts (stiffer). Bigger bearings often match the stack height of the drive screw better as well. For your length screws they will need to be very big to prevent critical vibrations. You will have to establish the speed profile required for welding and for machining so you can match the screw pitch. I suspect you will need to go to R&P on this project. The cost of large screws you will need a second mortgage.... Peter

my view of the twin gantry is that it is taking up space that a bigger stiffer gantry could be in. Since the twins are not connected very well they are not the same stiffness as twice their geometry. When people use vertical twin gantries they quickly realise they have to connect the two together with some sort of brace structure this can't be done with horizontal twins...

[G-Spot]

Back with a single beam.

The optimisation tell me to lose the back face of the beam and use 4 legs.

Starting to look like an elephant. That's the solution make a mold of an elephant and attach your rails to it.

Read this really good article on stiffness. So in-depth understandable and only 2 pages.

https://personalpages.manchester.ac....2-concepts.pdf

I think the article and the FEA are making me think 4 legs, one directly to each bearing is the best solution.

Funnily enough the local stadium, built for the World cup, here is constructed hanging off a single beam with 4 legs. One of the most awesome buildings I have ever seen.
The whole thing hangs from a a single beam the outer shells are almost transparent and lean outwards, when you are up close below you cant see the beam, so they look like they are defying gravity. Very Roman Colosseum inspired in the way they use the shade cloth.

The beam is made of steel all the interesting details here.

https://www.idc-online.com/technical...da_Stadium.pdf

[peteeng]

Hi G - Your starting to get along the road. The next issue is a saddle. A small deceptively simple part of the design but turns out to be a many sleepless night affair. You have to fit 8 bearings into a potentially small space plus get access to all bolts, plus drive it... I have done this several ways and now I place the gantry rails on the top of the gantry and have a 90deg saddle. This displaces the two sets of bearings so bolt access is easy. Also triangular gantries show up around the place. They are a bit stiffer then a squarish one. Peter

The article is interesting and points at designs being 3D trusses. In CNC world that maybe possible but complex, when 3D printing gets cheaper it maybe a good path. We currently tend to go down the monocoque path ie stressed skin approach as the load has to move around... There is a lot of high shear loading in machine parts and this tends to go counter to truss designs into thick webs and monocoque especially with a torsion component.

What is your optimisation criteria?

[G-Spot]

Hi G - Your starting to get along the road. The next issue is a saddle.

Yes I like that design, you save at least 70mm that way.

The criteria is to save mass using the deformation results. You can watch a movie of it optimising, it removes mass, then reruns the deformation test, then either keeps it or tries something different. I like it because it refines what you give it and you can select any face and exclude it from the optimisation.

Once you get something you want to work with you spend many hours/days making a smooth mesh then you will get accurate results, you then take those results into CAD and smooth it into something that is easily machinable/makeable, you then take those CAD drawings and run the simulations on them to confirm you get the deformations you expected. If you add all the details like rails and bearings and bolts etc, then the simulations take 10 minutes each and you quickly exceed the limit for the student version. Far better to quickly work through the basic theory of your design with simulations that take less than 30 seconds if you have a fast graphics card.

I noticed how in the optimisations it always left like 50mm of solid steel behind the rails because traveling along the steel in that plane is the quickest path to the legs,not going over the top or underneath.

Also noticed how when I used 10mm or less thickness the weakness is in buckling of the plate the rails are attached to.

Me next simulations will be with the beam rails attached to a 20mm thick plate then the beam.Then I am going to cruise the scrapyards looking for 20mm or thicker plate.

Already did the economics and I could just afford to build with new steel with a frame thickness of 10mm.

I get the feeling from all the simulations I have run that a 20mm thick angle Iron beam will be the best solution. But you weld triangle shaped stiffeners behind it.

I am going to see if I can use the FEA to place stiffeners in the correct place, I get the feeling if you lay an arch bridge with a suspended road deck on it's side, that is what the stiffeners would look like.

[G-Spot]

Tempted to go the fixed gantry route.

Fastest way to get milling. you can attatch a simple z axis to the bottom and use it to finish the rest of the machine.

Biggest problem is they take up double the space of the machine and the z axis is double the steel again. But it is the easiest way to get machining.

Moving gantry isn't going to be useful or safe until its finished.

Also they are much stiffer and just about every commercial machine for the last 50 years is fixed gantry and only the huge monsters are moving gantry.

Modern CNC milling machines the same size working area that I want.....nearly all fixed gantry

[G-Spot]

OK I am definitely going to run with this design.

This version is 10mm thick 300mm x 300mm cross sections. The weight is 650kg for the frame. Which is within my budget of 1000kg for steel.
The pics below are the last before I redesign from scratch, I'm posting a lot of stuff here because it is an excellent diary that I can use to go back to get numbers and stuff that I delete by mistake.

If you look at the numbers as closely as the colours, you will see that the frame is better than the HAAS machine specs the problem is with the spindle post design.
With the same force and direction applied to the both sides of the frame you can see that on the base side the entire frame is better than the HAAS specs, the red area on the base is the same spec as the HAAS. But there is just one area of the frame in the middle that I want to add some stiffeners to. Then I think the whole frame will be deep blue and far exceeding the HAAS specs.

Going to work my way from the frame to the tool tip until I have all my metal thicknesses and stiffeners etc correct, then I will move onto engineering drawings.

[G-Spot]

This is the frame with 50,0000N of force pushing down on the base. 380 MPa is the maximum stress.
Which must be close to the yield strength. Only 5mm deflection. Wow!
Well I will have the 50 ton press I always dreamed of if worse comes to the worst

[peteeng]

Hi G - I don't think you have done enough research to claim that for the last 50 years "all" machines are fixed gantry they are clearly not. At the size machine you are designing a high rail design is probably more optimal (stiffer) than a moving table/bridge design. See mill attached. Since you like concrete, the walls and gantry could be cast in concrete. That makes a lot of sense to me.

In the machines I have made the high rail design has proven to be strong and very stiff, even in plywood... I would not go back to column machines or portal machines unless the machine was very small...

The design you have done, some call it a double column mill, or a portal mill or a fixed gantry. Having worked in the crane area for a while a gantry is always moving. There is no such thing as a "fixed" gantry. But people call things different things all over. Keep at it. Although you appear to be "better" than HAAS, HAAS is a real number, you have cars, a saddle and various bits to go yet, each one dilutes the stiffness. Keep at it. Peter

Heres a bridge cast build - very good

Y-axis from UHPC Tegno concrete - YouTube

[G-Spot]

Spent the weekend chasing micrometres.
Have it down to 20.

Found that putting your top rails above the centre line of your beam means you can never hope for better than 50.
The top of beam always twist first because the base of the beam is more supported than the top.
Even worse is having a top rail above the centre line of the beam and the bottom rail below because the top twists the one way the bottom the other so you double whatever deflection you get at the bottom of the spindle post.

You can over come this weakness by making your beam ridiculously tall or wide but you could also overcome by making the whole machine out of a solid block of steel.
The article I read on stiffness I posted helped tremendously to get this "theoretical" machine down to 20. What was interesting in the simulations was that many of the bad designs were because of a small deflections elsewhere on the machine and the laws of leverage amplifying them at the spindle.

Anyways I went with creating the shortest path for loads, as the article said you should do. And that's the bottom of the beam, so that's where I mounted my rails. I also took inspiration from that super stadium beam but in reverse, 1 beam expanding to 2. And it worked,see attached picture.

Only using 10mm thick steel and a limit of 300mm beams then this is the only way to do it. There are no cantilevers in nature because it is the most inefficient use of materials so just hanging something of a wall without additional supports is anathema to me.

Talking of stuff that does not exist, that would be your friendly steel supplier saying. "Just 2m of that 300mm x 300mm and 1.5m of 250mm x 250mm, no problem sir", we are talking 6m,9m,13m lengths costing more than $600 a length. So the designs I am working on so far have to be able to utlilise a full length.

Using the Student version of FEA is working more for me than the full versions I have used in the past. It is forcing me to continually simplify the model as I go and the speed of the calculations makes me dive deeper into the science of what is going on. I think using welded blocks instead of bearings is definitely the way to go, as it removes the stuff like bearing design which is out of my control at this stage.
I also see the more complicated your parts arrangement the harder it is to isolate the reasons for a deflection. I found that just increasing plate thicknesses in certain areas where there was a weakness did not help much, that better relocation of parts was a better strategy.

At the end of it all I came to a conclusion, you cannot hope for super stiffness if you don't design your own beam, using an off the shelf beam which in my neck of the woods is limited to 300 x 300.

I also discovered that structural beams have an allowable twist of a few mm per metre which makes them only suitable for making the frame and not something rails can run on.

Can anyone confirm that. I checked the specs of the beams here and they all specifically mentioned allowable deviations to straightness in mm.

Anyways I think this machine I am going to build will be made from steel plate for the gantry, and that would be "gantry" as in the structure used to hold wooden barrels up used by the French that was not moving. Maybe they put wheels on them in NZ because of all the earthquakes you have there, Peter

[G-Spot]

There is no such thing as a "fixed" gantry.

Y-axis from UHPC Tegno concrete - YouTube

See attatched

[G-Spot]

Need help selecting bearings.

Was going to go with 25mm HIWIN EGR25 for all my axis

What are the different variations in block widths and heights for? they all seem about the same price so I presume it is going to be depending on use.

[jackjr-123]

Longer carriage: more balls, more load capacity, higher stiffness.
Flanged carriage: less tall, higher lateral stiffness.

So for machine tools you usually want to go with the longest flanged carriage, with the highest preload.

[peteeng]

Hi G - Agree with Jack, plus flanged bearings use larger dia bolts so the connection is stiffer. Std steel sections as you have found are made to quite slack tolerances when it comes to machine building. Usually extra material is welded on to a section so you have material to clean up and true the part in a mill. This welding introduces internal stress (which is internally in equilibrium post welding) hence the reason for stress relief prior to finishing. If you don't stress relieve, the part can (will) change shape as this stress then is not in equilibrium with the geometry. This change in shape means the milled part will not be the geometry you want it to be. 10mm plate is relatively thin for machine design. You can make a beam globally stiff but in a CNC frame the sections are relatively short. Look up short beam theory vs long beam theory. In short beams shear effects can be dominant. ie the sectional shape is not stiff enough to stay in shape (due to shear flow) hence becomes locally and globally inefficient. Especially corner stiffness in square sections. Rails should be mounted on deep sections or webs not on "air" as the 10mm steel will act as a membrane and needs to deflect before it stiffens like a hammock does. As you have found every small deflection gets amplified.

Sorry I'm in Oz but we do have the occasional shake. Only a week chasing um's I've been doing that for years... welcome to the chase. I agree building own beam is the best path then you can put metal where you need it... You need to get onto your saddle, it is not trivial and will further soften your structure. Same as addition of bearings into the model. Many Devils are in the details. Peter

Since your heavily into FE see if you have a large deflection solver or a non-linear geometry solver. If so run it and compare the NL answer to the linear answer. The result may surprise you as the NL will include the membrane effects vs the linear which won't. Plus check that the elements you are using include shear effects. If they are simple linear elements they will not include shear effects, but then you have to use many more simple elements to capture the shear effects. For instance you need 4 or 5 elements in the thru thickness direction, or more. If you only have one or two simple elements in thickness you will not capture some of the local deflections... another layer of knowledge needed... plus if you are "bonding" blocks of geometry together, analytical connections tend to be over stiff. So when you get to modelling connections more accurately you maybe 5-10% stiffer than reality... so what I'm saying is move along in your design there's a long way to go before you have a detailed manufacturing model for review and build.

[peteeng]

Hi G - I can see the attraction of the twin gantry - symmetry has advantages... Peter

https://youtu.be/cKVgugQSMfE

[G-Spot]

What's the stiffest beam with a maximum cross section of 300 x 300 that you can weld or bolt together(air powered 90deg tool) with only 10mm thick steel plate

Ladies and gentleman I give you the OctaBeam ....or beast. This is a 2.8m span with 100 tons on top a the midpoint. Only at 300Mpa on the beam, ignoring the test pads.

As stiff as you can get, it is super stiff against twist with only 2.6micrometers with that industry 5000N test.

It is stiffer against twist than bearing loads. I designed it by mistake, set out to make a hexagon, ended up with an octagon.

Think I am going to do the hexagon next....I am trying to perfect my FE bolting and welding simulations, but I think a hexagon will be simpler and just as stiff for the 1.5m span I am looking at for the machine I intend to build.

The best thing about these polygons are the faces are very resistant to twisting forces, so I might abandon the twin beam design....again...or maybe 2 hexagon beams.

[peteeng]

Good luck building that thing. There's lots of shapes you can make in the CAD that you can't achieve in reality. If you could get the triangular sections I suppose you could glue it together. I think Einstein said make it a simple as possible but not simpler. Simple is best. Peter

Plus with that amount of analytical bonding I expect its over stiff. make a side by side monolithic version and compare the stiffness....

may as well buy a D300 heavy wall tube and weld flats to it. Has been done before....

[G-Spot]

Longer carriage: more balls, more load capacity, higher stiffness.
Flanged carriage: less tall, higher lateral stiffness.

So for machine tools you usually want to go with the longest flanged carriage, with the highest preload.

Thanks. That's hours worth of research summed up perfectly.