[G-Spot]

A "hobby machine in a flat" refers to a machine that is located inside an apartment or flat, by definition a residential building typically occupied by multiple non-related persons / family groups. Such a situation imposes significant limitations on a machine as to size, noise, weight, vibration, 'operating time/hours' and electrical power consumption.

A machine inside an apartment that draws the displeasure of other residents or the landlord is a recipe for getting the machine owner evicted from the apartment.

Lets get back to the main topic of the thread.

Well surely the correct approach would be to carry out thermal and vibration analysis compare the max deflections to the max deflections caused by machine forces, and if they are less then it wont be an issue if they are more then they need to be addressed. A simple scientific approach where the designer gets to make the call. It's difficult to see how anyone could advise they wont be an issue especially if they have not built a machine of this type before. Strange how the first point of someone's list of the-best-things-to-do, are not to do what the previous poster said, makes it kind of obvious you are not interested in the main topic but discounting the advice of others.

[RaderSidetrack]

... makes it kind of obvious you are not interested in the main topic but discounting the advice of others.

Let me be clear: I'm wearing my 'Moderator' hat here: Drop this issue.

There are a variety of options in my Moderator Toolbox that can be applied here.

[peteeng]

Good Morning PIO & Others - I've had a restless night going over my EG stuff. This morning Thomas published his bending test data on his EG and it came out at 22GPa. He has done very careful work on this if you look up his thread.

Anyhow I built what I call a block model of a small gantry machine. It has a X travel of 700mm a Y travel 700mm and a Z of 350mm. Its this size because I used std parts in my library but I could shrink it now its built. Its solid steel and weighs 1.8tonne. I "pushed" the z axis with 1000N and ran this in fusion and simsolid and they both agree at 20um so its static stiffness is 1000/20=50N/um so if you used concrete it would be way down in stiffness. Looking at the model I need to make the gantry bigger to reduce the deflection. The base is 120mm of steel. The walls are 90mm thick steel.

But this gives you an idea of what it takes to get to 150N/um... I may shrink this a little to see if it perks up...... Peter

[joeavaerage]

Hi peteeng,
1.8 tonne of steel for only 50N/um? I know this was only a rough 'sketch and analyse' for the purposes of discussion, but that is still a swag of steel for that result.

I believe the base and the walls could be made very VERY much lighter and still end up with a good result. Better to concentrate on the gantry and Z axis....that is where the vast majority oy compliance
is coming from. Use heavier steel sections there with much lighter sections in the walls/base.

To OP, were you to make the walls and base out of concrete or grout or whatever, but use the biggest and best steel sections you can get for the gantry and Z axis you might get closer to a practical proposition,
1.8 tonne of steel is not affordable.

Craig

[peteeng]

Hi Craig - The next step would be to shell the parts until they came to reasonable weight with minimum stiffness penalty. It's a trade study and Round 1 so weight is not currently the main concern. Peter

[joeavaerage]

Hi peteeng,
very true.

Your model does rather highlight the gantry as the 'weakest link' and that the gantry is also the least suited to casting in concrete/UHPC or whatever.

Like it or not I think OP is obligated to look at a metal gantry, most likely steel, but possibly aluminum.
A laminated or composite gantry design might also be worth investigating.

Either way it is the gantry and/or Z axis that is by far and away the most critical part of the design. Get that bit right and the rest will be plain sailing.

Craig

[peteeng]

Hi all - I had some time so upped the size of the gantry and the saddle and the Z plate. Got it to just under 11um which is close to 100N/um a good starting point. I think I should have modelled the saddle as a block not the webbed design. Then I could chew it away. Anyways the next step would be to start lightweighting the design using GD or shelling to get the weight down. I won't take it further just wanted to illustrate the process. If I have time I may set up a M1 config block model. Oh yes I made it narrower so its X is 400mm but that didn't seem to make a huge delta. The gantry is only 188wide and 150mm deep, it needs to be much bigger. Peter

edit the smallest VF Haas weighs 11.5T. Even if its half fluff then the frame still is over 6 tonne of cast iron. Its X1000 Y660mm and Z635mm

[joeavaerage]

Hi peteng,
just scanning through the Hass PDF and it looks like a standard (BT40) VF1 is 3200kg while the VF6 breaks 9500kg. Both are still very heavy but not quite 11.5T either.

Craig

[peteeng]

Thanks Craig - Looks like I read the wrong column (the VF-8 column). I have started shelling out Mill 1. Only lost a couple of um so far. But there's holes in the blocks so the shell is weird and various relationships are failing so it would be time to rebuild the model. Plus Fusions mesher has failed with the odd shapes so I did the last rounds in Simsolid. But I chewed a lot of mass out. With no big consequence only 2um lost so far. But its been the fastest way to get to a low deflection vs working up from the bottom... I also ran X dirn loads and the walls will need ribs in them. The X went to 30um. Better get back to my machine! Peter

[joeavaerage]

Hi peteeng,
I know you are wedded, or at least very much favour, high sidewalls and moving gantry designs whereas I favour C column designs.

What I do believe would be very instructive is to model both to some reasonable stiffness, say 50N/um, and compare mass.
Intuition tells me that column designs would be heavier for the same stiffness, but i don't actually know...ie have the comparative numbers.

Craig

[peteeng]

Hi Craig - I'm not wedded to a particular config. Each config has its place same as each material has its place. Column mill designs are good for small Y axis machines so the table tends to be 2:1 aspect ratio. If the working envelope needs to be big then the gantry kicks in. For instance I'm working with a company that wants a side loading 10x8' machine so it has to have a gantry on columns...

But yes I'd like to do a M1 type block model... But my Lanky machine (8v4 router) and Epoch5 (5 axis machine) need time to finalise and build.. The flanged harmonic drives have simplified many things. I don't want to highjack this thread though... Peter

[peteeng]

Hi Pio - To stop littering your thread I have done a C frame to gantry face off in my Milli thread. I've had a small mill on the list for a couple of years but have been struggling with materials and configs. Maybe this will give me the dirn. Peter

https://www.cnczone.com/forums/cnc-d...ml#post2587398

[piotr07da]

joeavaerage
You are right. I am obligated to test both options.

peteeng
Everything you write is helpful. Thank you.

about my struggle with comparison

I am either doing something wrong or misunderstanding something.

For an experiment, I created a gantry. It is entirely solid. The columns are 320 x 170 mm (12.6 x 6.7").

I conducted some FEA, and I am getting completely different results depending on the size and shape of the object used as a fixed constraint. I attached some pictures showing what I mean.

For example, if an object located on the T-table is:
- 120 x 120 x 30 mm, then I am getting 71 N/µm
- 240 x 240 x 30 mm, then I am getting 127 N/µm
- 120 x 120 x 200 mm, then I am getting 45 N/µm
- 240 x 240 x 200 mm, then I am getting 104 N/µm

This effect is even more magnified if I test at extreme positions of travel - min X, min Y, and max Z. Then, I am getting results from 17 N/µm for a 140 x 150 x 30 mm triangle to 150 N/µm when fixing the entire surface of the table, which completely eliminates the table from the equation.

I also did not find an answer on how this test should be exactly performed in neither ASME B5.54 nor Principles of Rapid Machine Design. There are general rules, but they are not enough to achieve consistent results.

My point is that I can establish some methodology for myself to compare my own designs and stick to it. However, it is impossible to refer my results to any other results and machines. For example, 150 N/µm is meaningless without knowing the exact way it was measured.

So again - I think I am probably far from understanding how I should approach this comparison.

[joeavaerage]

Hi,
my reasoning goes something like this:

With a gantry and the Z axis at its fullest extent and the Z axis central in the gantry is the worst load condition. The gantry is under its worst torsion load condition. The torsion is resisted
by virtue of the rails at either end of the gantry, and thus the torsional stiffness of the gantry is tested in the +Y and -Y directions, with the distance being Y_max/2.

A C column design the column is under torsion, with the torsion being resisted by the base alone. With the Z axis at the top of the column the torsional deflection is at its worse case.
If you have a very low column, ie small Z axis travel, then even a modest column would still be torsionaly stiff. With a large Z axis travel, ie tall column then the torsional stiffness of the column is severely
tested.

I consider one of the advantages of a column design is that you can have a large Z axis travel but the penalty is that the column must increase dimensionally to be adequately stiff.
Gantry designs tend to have smaller Z axis travels, and therefore the moment inducing the torsional deflection of the gantry is least.

peteeng is right in that a gantry design allows large X but also Y travels, but tend to limit Z travel. A column design on the other hand may have large X and Z axis travels but in order to limit the moment
exerted on the column the Y axis travel tends to be smaller.

My guess or intuition is that a column of a column design must be dimensionally bigger, and therefore more massive than the gantry of a gantry design. The balance is that with a column you may choose the Z axis
travel at will (within the exception of making the column longer) but you must limit the Y axis travel to reduce the torsional load on the column. A gantry on the other hand you may choose X and Y travels largely at will
but the Z travel is constrained.

Lets say the X travel is out 'yardstick'. If the machining volume is the same for the two designs (gantry vs column) then:

Column: X travel =l, Y travel =l/2, Z travel =l, machining volume = l³/2
Gantry: X travel = l, Y travel = l, Z travel =l/2, machining volume =l³/2

My intuition tells me that for both machines to be as rigid as each other would require more mass in the column design, with that extra mass being largely concentrated in the column.

Craig

[peteeng]

Hi Pio - Having been at this same question for a few years here's my thoughts:
1) Seems industry is not really concerned about the static stiffness of a machine for comparison purposes, otherwise they would have it sorted and spec'ed on their marketing data sheets
2) For purposes of design individuals or companies will have their own methodology
3) For the purposes of decision making across configs or materials or innovations just keep it as close to apples to apples as possible and use the static stiffness as a guide for the decision
4) I think you have a good understanding of the issue as you have explained it very well
5) Welcome to the chase of the static Grail
6) Wait till you get to dynamic stiffness thats Grails within Grails
7) You will make a very good machine as you have the tools and approach to get there

The forum has many answers and resources to help you achieve a great machine keep asking the questions.... Peter

Looking at your models I would use a finer mesh. The general guide for mesh transitions is 10% by size. Some of your areas where small features meet bigger features can't achieve this. Depending on your element type (linear, parabolic etc) this may lead to strain hardening or strain softening in those areas especially with analytical connections vs contiguous meshing. In your models you have contiguous mesh I think? Yes as you have made the model as a "solid" rereading your txt. Technically for purposes of deflections nearly any mesh should work but tetragonal meshes being triangles which are in themselves stiff tend to be stiffer then quad meshes. There is an issue with this sort of meshing called "mesh locking" but haven't come across this in some time. Recent FE systems tend to go tri meshes as they transition better then quad meshes and cope with complex geometry better than quad meshes. These sort of issues can be checked by running a fine mesh model and a coarse mesh model to see if they deflect differently. Then you can progress the work in the knowledge the mesh is OK.

Q - can your FE plot rotations? I find this is sometimes more useful then plotting global deflection as it will show up areas that are rotating locally easier then looking at global deflections....Unfortunately Fusion and SS does not do this, Strand7 does.

[joeavaerage]

Hi,
one thing that may influence your design is the use of concrete as a primary material. The most critical structure of a C column design is the column itself and the base. Both lend themselves to
the use of concrete or EG. The dimensions of both may be swelled to meet any target stiffness.

A gantry design on the other hand the critical components are the gantry and Z axis, neither lend themselves to the use of your primary material. These will like as not end up being made of steel
being highly stiff and cost effective. Concrete (or EG) could well be used in the base and sidewalls, but they are far from critical. No matter how big or stiff you make the base and sidewalls the gantry
will still almost certainly be the biggest contributor to compliance.

Craig

[piotr07da]

Hey,

joeavaerage

I was expecting exactly what you are describing as the worst load condition. That was my intuition. And this was the case when I fixed the base, measuring only the gantry part of the machine. However, interestingly, when I modeled the whole system, having a fixed constraint on the surface located on the table, then the stiffness of the table and base of the machine became the weakest link.

Having Z at full extent down gave 30 N/µm (thanks to the good stiffness of the gantry design and low torsion on the base).
Having Z at the top of the travel gave only 17 N/µm (because, although no torsion is applied to the gantry, large torsions are applied to the base and the table).
The above results are shown in the attachments (although in second picture max deflection is is at the top, I used only the spindle nose deflection so the comparison should be valid).

So, my conclusion from the FEA tests is that in the case of the gantry design, the stiffness of the base and the table is as important as the rest of the system. Or it is an even bigger contributor to compliance than the gantry, although the gantry as an isolated part seems to be rigid.

I expect the whole system would behave slightly better in the case of a movable gantry on top of the tall walls, as they could stiffen the base, as in peteeng's model. At least in bending in the Y direction. But without performing any tests for the tall walls design, I don't want to claim that because my intuition has failed me many times lately.

Nevertheless, I agree with your comparison between the gantry and the C. It makes sense.

Initially, I planned to compare column and gantry designs using the same travels. But now it appears to be an unfair comparison. To be fair, I should swap the distances of Y and Z travels.

And here is the thing - I have started to question whether I need this much in the Z axis. And maybe it is a good idea to swap travel distances between Y and Z so it is more flat than vertical. For the gantry design, it would result in even better stiffness.

I initially used 380 mm (14.96") for Z because I read somewhere that more Z is always needed than expected due to the tool holder, tool height, vise, or fixture, or both. Of course, I understand that I cannot expect you guys to tell me how much Z I will need while you don't know what parts I will be milling, especially when even I don't know that. I imagine them to be rather flatter than taller. But I will appreciate it if you will share your own observations and experiences.

peteeng

I am glad to know that I am not alone in my doubts and questions. I will establish a set of tests and will test all the designs using them. In the case of dynamic stiffness, my knowledge is none, so I will have to familiarize myself with this topic.

Regarding FEA - I am using FreeCAD FEM. I have the Fusion 360 free hobbyist version, which doesn't have the FEA module. I attached the picture showing what options I have. The results I can display after FEA are displacement magnitude, Tresca stress, strain xx, xy, ... components. I always felt that there should be something that tells me where in the model there is the biggest bending - I expect it would require quite uniform meshing though.

I will use finer mesh in the future.

I have also one more question about bolted design. I know that the thicker the better, but there is the cost factor. Bigger beams with thinner walls are more rigid than smaller beams with thicker walls. At least as long as the walls don't buckle. I wonder how thin I can go. I started from 20 mm (.78") sheets bolted through the face of one of them to the edge of the second one with M10 bolts. But I can make the structure more rigid for the same mass using 15 mm sheets and M8 bolts. But now I am thinking about doing something else - making even thinner walls, let's say 10 mm, and bolting using M10 bolts through the faces of both sheets into solid steel bars. It is hard to explain, so I attached pictures. Picture shows beam sections with 20mm i 10mm walls. Using the second option, there is the additional advantage of easy adjustment of geometry to being square. What do you think?

[peteeng]

Hi PIO- I'll have a read of freecad fem and see where its up to. Can it do preloaded bolts?
1) re mesh in the settings image you are using second order elements. If your solve time is long change this to first order elements and it will be considerably faster. Since we are only interested in deflections second order is a bit over the top (it will give a better stress result but deflection should be the same). But run a model 1st order and 2nd order to check deflection is the same (first order elements if enough of them will give same answer as 2nd order). 1st order elements will give about the same result as a 2nd order and solve really fast. 2nd order are useful for big elements on curvy surfaces as the mid node can provide curvature to the element. But machine parts are generally flat
2) Does freecad do linear buckling? If so run that and it will tell you the probable buckling modes and load ratios
3) Thin structure design vs thick structure design is a bit of a learning curve. Machines probably live in thick structure land. This is to account for shear transfer within the structure and to remove vibration modes. Thin structures vibrate, this is one of the issues using construction extrusions they have lots of thin free edges and they can vibrate alot...
4) I think by the time you make your elements stiff enough they will be thick and won't buckle
5) Dynamic stiffness involves vibration and the ability to damp it. If freecad does modal analysis you can run that and find out what vibration modes are likely to occur and where and the freq. In other systems if you know the damping factors the FE can calculate the damping time and effects. But this falls into transient dynamic solvers and complexity
6) Using brackets is valid. But to correctly model the bolted connection the solver needs to be able to preload the bolt and account for friction in the connection. If you model it as a "snug" bolt it will not be as stiff as a friction connection...
7) Change growth rate to 0.1 will help
8) Z axis height? depends on what you want to do. Have a look at commercial mills and see what they do. My std Z is 350mm
9) anything else ask

Peter

edit - I looked thru the freecad Fe features. Currently it does not support bolted connections. The mesh has to be contiguous thru multibodies. It can have friction surfaces but how can it create the friction zone around a bolt? it can't so in this case bolts would be part of the contiguous mesh. So leave bolts out, model as faying solids and assume that the connection is 90% efficient say... Once you get down the path a bit I could run a model for you... Since strains within machines are very small we can assume that the friction in the connection if designed correctly will not slip. This is what happens at rails and motor flanges etc. So its safe to assume the connection is "bonded" initially... Have fun Peter

Edge bolting is good if the plate is thick. But making edge bolting is fiddly unless youyr CNC has an aggreagate fitting or is 5 axes. I think brackets are the way to go... unless you are happy match drilling and hand tapping all those bolts...

[joeavaerage]

Hi,

I have started to question whether I need this much in the Z axis. And maybe it is a good idea to swap travel distances between Y and Z so it is more flat than vertical.

There we disagree, I am firmly in the camp of the more Z axis travel the better. In addition to the vice, workholding and the tool and tool holder, I always intended at some stage to have a fourth and fifth axis.
They eat into your Z axis travel bigtime.

Several months ago I largely completed my trunnion fourth axis and rotating platter fifth axis. The height of the platter is 240mm above the bed. My Z axis has 350mm travel, so with the fourth/fifth axis in place my
remaining travel is only 110mm. Fortunately my Z axis is movable. That is to say the Z axis bed is cast iron, 700mm long and 235mm wide. It contains the servo, ballscrew and saddle. It attaches to the vertical column
of my 'frame' by four large bolts. Thus if I need more Z axis travel I can unbolt the Z axis bed and shift it upwards.

My design was from the very outset, about four years ago now, going to be modular. That is the X,Y,Z axes were all going to be identical and bolted together. The bed themselves were going to be cast iron, stiff and well damped.
I was going to spare no expense to make the axis beds as good as I possibly could. They would then be bolted to a frame. As it turns out the three axis beds, including the casting, precision machining, ballscrews, rails/cars and servos
cost me $16,000NZD, while the steel frame (laser cut fabricated steel plate) cost only $2000NZD. Thus if I ever decided I need a more rigid frame I could do so and just transfer the axis beds to it.

With the forgoing you can see that a column design offered the greatest flexibility (not meaning compliance here!) in executing my design. Whereas a gantry design is very much less flexible in regard to altering the
height of the Z axis in particular.

I think peteeng has already come to the conclusion that his five axis design 'Epoch' has to be five axis from the outset, whereas his three axis 'Millli' is three axis and will perforce remain so. A column design might allow you
to 'mode' your machine as required.

As peteeng has also pointed out that column designs tend to limit Y axis travel to avoid excessive moment on the column. The travel of my Y axis is 350mm, the same as the X and Z travels. Additionally an important design
parameter is a column design is the space between the Y axis and the column. The larger the space the bigger the parts you can handle, very desirable.....BUT the larger the space the greater the flexure of the column.....so
there is a balance to be struck. Given that my Y axis already has generous travel AND I allowed for a large (relative) space between the column and Y axis I have a machine that has very good capability to accept large
and/or awkwardly shaped parts, but have really 'put the acid on the column design' with regards rigidity.

Were I to redesign my machine I would be a bit more 'stingy' about the Y-to-Column distance to gain a little extra rigidity. Every time you design and build something you learn new things, or perhaps appreciate better old
things....

Craig

[peteeng]

Hi All - Since its currently a design exercise I'd go max expected Z and once a more details are in place cut it back if needed. I don't think the gantry config is any less difficult to change the Z in future then a column design. My Epoch project has gone full circle and now I can see it easily being 3,4 or 5 axis within the same basic structure. Simplification can only happen once the major objectives have been worked through. The front end of this sort of project is fuzzy and it takes time for a newbie to defuzz the project. PIO go for a neat 400x400x400....Peter