[joeavaerage]

Hi,
my model was just to show how sensitive the axis bed is to the depth of the gutter, it is not intended to be used in your model...that is for you to do.

Does this sims gives some real indications at all?

If, like me, you are unused to modelling and FEA it is very possible to end up with results that are a far cry from reality. peteeng is very experienced at FEA and still there are times
when he questions whether a particular analysis is realistic.

I find rather than relying on FEA to give me an accurate and realistic impression of a real world structure, I use the analysis to compare one structure with another. The two studies I posted are
a simple but good example, they are identical EXCEPT for the gutter. The results may or may not be hugely accurate of the real world, but as a comparison they are very instructive.
If you like, I use FEA to inform my thinking, but am careful about ascribing the model to be a perfect reflection of reality.

Craig

[peteeng]

Hi Fatam - Yes FEA can be very accurate if set up correctly. In your case its a comparison tool. Like Craig - how sensitive are certain features or global models to loads. 1000N nominal is a good design load at the spindle nose. Then divide that by the nose deflection in um to get a normalised N/um stiffness. This will allow you to compare it to other machines or your other designs... Peter

[joeavaerage]

Hi,
one thing that tripped me up for a long period of time when modelling a complete machine is the constraints.

You might think it logical to restrain the base by 'locking' it in place relative to the rest of the universe. But that is a mistake because it incorrectly make the base much stiffer than it is in reality.
It is by no means obvious to me, or at least was back then, that this was a mistake. Like any other software tool there is a learning curve. The old saying "BS in means BS out" applies
to FEA.

I have attached a rough FE sketch that I posted in another thread. Note how the top of the 'table' (shrunk down to a square) is locked (universally) and a load is applied to the square representing
the headstock. That means that the table on which the machine sits has no part in the load path...and that is as it should be.

Craig

[joeavaerage]

Hi,
another idea I use extensively is to reduce the model to one part.....simplify it until you are looking at the deflection of just that one part. I find that it makes the modelling simpler but
more importantly is less likely to lead to a mistake that will give unrealistic results, ie useless.

With respect to the model I not long ago posted. The actual part is the L shaped frame alone. Not that you can see it, but it is hollow. The 'L' shaped sides are 32mm thick, the top and bottom
are 10mm thick and there is a half dozen internal separators/webs inside out of 20mm. All or these parts are drawn separately but in the model they are all one part, almost like they
had been cast in one piece. That means that I do not have to model a welded part or a bolted part, which for me represents a great opportunity to make a mistake. It does mean however
I have to consider this model ideal and that a real world construction may be somewhat less than this model indicates.

You might ask then what are the square upstand on the base and what is the square section protruding from the column?. They are solid sections meant to represent the table, or at least the small
section of the table that is bearing load, and the Z axis, saddle and headstock respectively. Ideally I would have modelled them as ideal perfectly stiff components, so the results would be from the flexure of the frame
alone....but I'm not sure how to model that. So I made them out of solid steel. Sure they will flex a little, but it is a fair approximation to 'infinitely stiff'.

My intent here is to model the frame and analyse its performance. I can then change say the thickness of the top and bottom plates and try again. It may have a big influence, but then again it may not.
This is a convenient way for me to find out. What I don't want is all the flexure of the rails/cars/ballcrews/axes etc coming in and clouding or obscuring what I'm trying to study and learn.

This sort of granular approach reflects my grave inexperience in modelling physical components in FEA. I would rather have a simplified model that tells me what I need to know with confidence.
That my simplified model might be somewhat better (or worse) than the real world structure is less important to me.

Craig

[joeavaerage]

Hi,
as an example of the sort of thing that can trip you up: the second model you posted showing a moment along the Y axis.
I see you have placed constraint along each bottom edge of the base, this is a mistake. That means the bottom edge (of each side) is locked in space, ie it cannot deflect,
but that is not realistic at all. I will presumably sit on top of your table, but your table is NOT infinitely stiff and therefore neither is the bottom edge of your axis bed
restrained to zero deflection. You have 'bonded' the axis bed to an infinitely stiff mass or body or universe or whatever you want to call it, and so your structure
has effectively an infinitely stiff base.....but that's not real.

Craig

[joeavaerage]

Hi,
another idea that you need to think about is that you cannot restrain and edge, its not realistic.

Lets imagine some part, approximately rectangular, and you restrain or lock one edge in the 'universe'. Then you apply a load to the part......all that force must be balanced by
a restoring force operating along that edge. But that edge is logically, at the microscopic level, infinitely thin and therefor any force applied on an infintesimal body means infinite pressure.
You can now imagine that the results you'd get look really weird.

The correct way is the restrain a part over an area, so that the pressure is not infinite.

Don't worry I have made this exact mistake and is how I learned about it. I found it particularly when I was analysing a new cast iron headstock for my machine. I was restraining on several
edges and getting results that did not look kosher. peteeng tipped me off as to why that is the case. The saddle is 275mm x 275mm square and flat. It is unrealistic to believe that I could bond to
that entire surface, when I was bolting with a string of bolts around the edge of my part to the saddle.

I used peteeng's suggestion that I include a small 'land', say 0.1mm high, and about 4 bolt diameters in diameter at each bolt hole. Now the part was sitting on these raised portions and each land
was locked to the underlying saddle. This is much more realistic, and much better and believable results followed. peteeng uses an even rather more realistic bolt model than that. You can imagine that
my approximation was to effect perfectly and intimately bond each land to the underlying 'universe' whereas the bolt really applies pressure of the land to the underlying material and friction is what
holds them together. As it turns out if you model a small land as I have done verses a rather more complex bolt model often results in the bolt model being somewhere between 85% to 95%
of the idealised model that I use. I have to keep in mind that my model employs an approximation that somewhat overestimates the bonding between one part and another.

Once again I favour simplicity of modelling and the elimination of major modelling errors than exact modelling that may include logical errors that make the results meaningless.

Craig

[Fatamorgana]

Hello, and thanks for all the tips regarding FEA. I must admit, the two last days has gone away to extensive FEA learning, been all over youtube trying to source learning material, combined with tips from you I have made som new simulations. I have also gone away from bonding surfaces where I actuall will be having bolts to attach the pieces together. The one thing im not totally sure of is when I make a contact point for to surfaces that will restrain with bolt, I have ended up with "rough" contact, where after my understanding ""separate" would give more meaning, but after the different simulations I have found the rough contact to give must meaning. I also restrained the feets under y-axis bed as locked points. I also did it with the frame, but for simpleness I will just keep it like that as of now.

I must admit it seems like an awesome tool, the last thing I needed now at this point was a reason to use more time in fusion, but this is much the reason for doing this, the steep angle of learning.

In these examples I have used a force of 5000n on the different axis on spindle. Please give me a point in the right direction if im totally of track here (other than overexaggerate use of force)

[Fatamorgana]

Also, about to ordrer some linear guides/cars. Got a great offer from our favourite shop in Korea. What you think of this Brand?



Sent from my iPhone using Tapatalk

[peteeng]

Thompson is the inventor of linear guides - I expect they will be very good. Peter

[peteeng]

Hi Fatam - In Fusion constraints there are bolt connections. You can place a bolt in hole with or without a nut. Then you specify the bolt size and length and preload. Plus the contact friction co-efficient say for steel 0.35 which you use in the separating contacts of the connected surfaces. Then the connection is correctly connected with bolts. Make sure the faying surfaces are not bonded together and are converted to separating contacts then you can run the model.

Fusion Help | Bolt Connector | Autodesk

If you use rough contacts the connection it can separate but not slide. Sliding is more likely if not enough preload is there, if the connection is in high shear or has high bending loads which induce shear. Or if the bolts are "snug bolts" or there are dowels in the connection then sliding is not possible. If the connection slips then you need more preload or more bolts. But keep playing and you'll sort these things out. Peter

In regard to the design load it does not matter. I usually use 1000N. What does matter is normalising the deflection to the load so you know its compliance/stiffness. Say take your Z load 5000N/441um = 11.33N/um which is quite stiff but not stiff like commercial VMC at 100-150N/um plus...

Now another thing about gutters. Traditionally with machine movement involving ways and gibs we had to have the drive forces being symmetric and at the friction centre otherwise the table would jam. So we had to have a gutter. This is the same as using an F clamp. One side cocks and the assembly jams. This is a stick-slip issue. Using roller cars and linear bearings has changed that. As there is no friction (nearly none and certainly no difference in dynamic or static friction) there is no stick slip. Plus as the cars resist moments the table is insensitive to where you drive it if the bearings are well separated. This means you could drive the table on the base side vs central and not use a gutter (therefore much stiffer). There are commercial machines that do this. Being at the side you have the opportunity to adjust the height of screws and bits to fit the parts. Worth looking at....

In regard to FEA modelling you also need to consider the full structural loop. If you have a large spacing between the spindle nose and the table for instance and you constrain the table then there is a secondary moment created by that distance (sometimes called a "crank"). This moment creates more loads in the machine than in reality. The tool and the cut face are always the same distance apart (zero) if that makes sense. So you need to build a dummy part on the table that comes up to the "tool" so this closes the loop. Then the action and reaction are co-linear and no secondary moments or forces are created... Peter

[Fatamorgana]

Hey Pete

Thanks for clearing out my understandings of contacts.

As I have learned from the simulations I can see that the gutter really weakens my design, so I was really interested in your thoughts about rethinking the design. I quickly edited the design to see if Im on the track of something, I also removed the gutter.




Sent from my iPhone using Tapatalk

[peteeng]

Hi Lars - Having the rails on the column sides means the rail lands have to be machined very accurately as this removes one direction of adjustment. If your confident of machining on spec (especially parallel) like this then that's fine. I meant to leave the rails where they are but move the drive to the side. Have you moved the drive for the Table/Base? Peter

[joeavaerage]

Hi,
having the rails on the side is tricky, I would avoid it.

The simple expedient is to make the gutter as shallow as possible and then increase the thickness of the slab until you get a stiffness that is suitable.
Lets take a ridiculous case, where the gutter is 30mm deep but the slab is 200mm thick.....then the gutter will make very little difference to the overall stiffness of the
axis, but then 200mm is a damned thick, heavy and expensive piece of material. A 100mm thick slab might be a better choice, it will not be as stiff, but neither will it be as badly affected
by a 30mm gutter than say an 80mm slab, with only a modest increase in weight and cost.

This is a very common case when designing....you have to make a compromise between stiffness, size and cost. There is a balance to be struck.

Craig

[peteeng]

This is where I suggested to place the drive. An image maybe explain better - Peter

[joeavaerage]

Hi Lars,
the pics you posted of the linear rails/cars is that they are 30mm rails...correct? If so they a 42mm high.....correct?
The ballscrews you pictured are 'cut flange types'.....correct?

If so they could be as high as 48mm....correct?

In that case the gutter need be 6mm ONLY!!!!

More like you want some clearance and so the height of the ballscrew and ballnut mount might be 55mm say, the gutter still only need be 55-42=13mm deep. In an 80mm slab
of steel a 13mm gutter is hardly going to ruin its stiffness!!

If the gutter were 40mm deep in an 80mmm slab then I'd say yes, you probably need to re-think what you're doing but a 13mm gutter??....., do not introduce a whole bunch of complexity
with respect to construction and multiple setups in your machine just to solve a problem that is just not that important.

Craig

[joeavaerage]

Hi peteeng,

Why???? A small or modest gutter is not going to have an adverse influence on Lars design....your solution adds complexity to solve a non existent problem.

If Lars is being smart he could lay his 80mm slab down on the bed of his mill and do the gutter, the tops (rail mount surface), the rail drilling, the ballscrew mount drilling,
the limit switch/home switch mount drilling....ALL in one setup, without ever having to move the slab until its complete. This is by far and away the most straight forward
and accurate way to do so, and especially for a hobby machinist. Simplicity and ease of construction trump trying to gain a few percent in stiffness at the expense of multiple
setups.

Craig

[peteeng]

Hi Craig - Its a suggestion to remove the gutter. Its not a complication & it can be simpler. If your chasing stiffness then every bit counts. If you go this route you can even raise the middle bit between the rails and increase the base & column stiffness more. I simply put fwd alternatives, there's always more ways to skin the possum.... Peter

[joeavaerage]

Hi peteeng,
lets assume for the moment that Lars uses two 80mm thick slabs for the Y an Z axes. Lets further assume that he has no gutter at all. Lets go further and assume that the right angle
joint can be made ideal, ie perfectly stiff, and so to the headstock and 'crank' as you call it, the upstand representing the X axis.

The ONLY flexible components left are the Y and Z axis beds, and that they are as stiff as an 80mm slab of steel. Were you to analyse this structure you might find
stiffness of N_x, N_y and N_z N/um in each of the load directions.

Now analyse the same structure but with a 13mm gutter. I would expect about a 20% degradation of the stiffness of each axis bed and thus all the stiffness parameters
( N_x, N_y and N_z) would also degrade by 20%. certainly a degradation but hardly earth shattering. Were Lars to chose to spend an extra $250
to get 100mm slabs even with a 13mm gutter the net result would be better 80mm slabs WITHOUT a gutter.

I just got quoted 1050,- dollar for 3 pcs of 300x800x80 and 1250,- for 300x800x100

From the point of view of simplicity an extra 25% of the cost of the slabs is an easy decision to get a better result than 80mm slabs with no gutter.

Craig

[peteeng]

Hi Craig - Thats for Lars to figure out, it's his possum. It's simply a suggestion. Lars will work out what's best for him. Peter

[joeavaerage]

Hi peteeng,
well the question is if you cannot make an axis bed stiff enough out of an 80mm slab of steel with a small gutter then what the hell can you use?. Beryillium is not a practical answer.

Craig