[peteeng]

Hi Sus - Read the paper. They state that fabricating machines has "failed" a couple of times. Commercial builders have been successfully fabricating machines forever. Casting is a faster production route if you want to make 100s of machines that's why its picked over welding. Bit of a lightweight paper for a conference. Peter

[peteeng]

Hi Sus et al - Looking back through your thread you have always chosen the C frame as the design with an XY table. The C frame design historically is used to allow easy access to the table and to minimise build parts. The structure is usually massive. If you have used a light C clamp you will quickly realise they twist and spread when you load them up. The C shape is a poor structural choice if alternatives are around. This is called an open structure. Closed structures are much more rigid. Like a gantry its loads stay within the load loop. A C frame is a folded cantilever and if you look at the cantilever deflection eqn you can see that the L^3 part is dominant. If you close the loop it becomes 8x stiffer with the same section sizes.

To close the loop on a C frame means bringing an arch over the top of the spindle down to the front of the machine. This is doable but then loading maybe an issue. So maybe turn the machine on its side. Plus the parts would become long to go over the spindle so maybe will not be ideal (lengthening L^3). The other way is to make an triangular prism config so the spindle can poke out the top. This fits better with the machine footprint and if made wide would fit inside the enclosure with no issues really. These sort of exoskeleton machines have been built but they are more complex. Now we have good CAD these things are easier to make. A cubic exo structure is viable as well but triangles are stiffer & shorter.

Chase the Grail - Peter

The machine design notes has some internal conflicts I feel. They push for the damped concrete but in the theory it states stiffer is better. So if you are creating a damped system you may as well benefit from a higher stiffness material (steel) and damp that vs using concrete. Using laminated metal is effectively creating a shear damper system. Airbus uses laminated aluminium (look up GLARE) for its airframes and one characteristic they are happy with is low vibration and acoustic noise. Its lower cost then carbon fibre and has excellent fatigue performance. But for machines fatigue does not happen.

[joeavaerage]

Hi,
my machine is a column design, and the rigidity of the machine is largely tied to the rigidity of the column, Peter is perfectly correct.

Provided you are willing and able to build the column of sufficient dimension and material; it is as stiff as any other design.

Column machines are usually made in four main pieces, the column, and that bolts to the base, the headstock that extends the Z axis saddle to the spindle, and the saddle/X axis/Table that sits upon the base.
Clearly to have a rigid machine the right angle between joint between the base and column must be perfectly rigidly bolted. The column itself is subject to bending when subjected to forces in the Z axis direction,
and the Y axis direction. Under X axis loads the column twists. For this reason the column must be uber stiff against bending and especially torsion. The headstock is subject to bending under Y and Z axis
loads and torsion under x axis loads. It too needs to be uber stiff, but being that much shorter is relatively easy to achieve.

The advantage of a column design is the flexibility with respect to the arrangement of parts and very VERY good part access, the downside is that the components need to be stiff.

I myself chose the column design because it gave me the flexibility to design to get the most from my axis beds. The column and base are one piece, ie no botted angle joint.
They are two 'L' shaped pieces of plasma cut 32mm medium tensile steel. Easy and relatively cheap to fabricate. To that frame I bolt the Z and Y axis beds and the X axis/table sits atop
the Y axis.

For all the supposed advantages that Peter quotes with close frame type verses open type frames you might imagine that all machines would be built that way........but consider that there must be some fairly cogent reasons
why 50% or more of vertical mills on the planet are column types or variants of it.

Column designs are perfectly feasible if you are prepared to design and build the required components, the column being the most critical, correctly. The advantage comes when you
have to get them machined, you can disassemble into large but mainly rectangular chunks, a near perfect fit for the majority of large mills out there that you might select to do the job.

Craig

[suspension]

Thanks both for explanations. As per the structure, I have settled with C frame. Many of the simulations that I did at the start did not give me a huge stiffness benefit for some reason. My reasoning was that such benefits with closed structures is significant for larger machines - but I see that the 8 fold gain in stiffness is independent of L. In any case, I have to stick to C frame now.

Hi Craig,

Did you do any simulations to understand the stiffness of your machine? In my case, I did many simulations with steel plate based designs. I designed the column and base as large as possible for a given total steel mass. The results were disappointing. I figured that the cost of steel, welding, cutting, stress relieving will easily go beyond the cost of say a frame that use a grout (Even expensive ones like Sika 3350) for a given stiffness.

Hi Pete,

My main concern now with grout is only shrinkage. I will see if Sika 214LK will give me better results. When I prepared samples I used for EM and shrinkage testing with previous grout, I mixed it with aggregates of size up to 15mm although according to TDS it should be 10mm max. Also, I may have put a little bit of more water than what was instructed (Say 10% more). Could these lead to reduction of EM from 47 to 30? And also to give a shrinkage of around 0.5um per day although according to TFS it is non shrink?


Thanks
Sus

[jackjr-123]

Did you do any simulations to understand the stiffness of your machine? In my case, I did many simulations with steel plate based designs. I designed the column and base as large as possible for a given total steel mass. The results were disappointing. I figured that the cost of steel, welding, cutting, stress relieving will easily go beyond the cost of say a frame that use a grout (Even expensive ones like Sika 3350) for a given stiffness.

According to my sims a steel column will be about 50% stiffer than a UHPC (E45) column of the same volume and mass. But you're right that the steel one will be more expensive and requires more work.

[peteeng]

Hi Sus - "non shrink" is an industry term not a physical absolute. Some grouts shrink overall and some grow a little. Growing grouts are used so they fill a hole and lock in. The less water used the less is the drying shrinkage. Water addition shouldn't affect EM but will affect shrinkage. I'm not a concrete chemist. I have found local Sika technical people good at answering various questions. In regard to aggregate size this is a function for concrete pumping I expect. The actual size does not really matter. The concrete guy I spoke to said that the bigger the aggregate the better the EM would be... But then you have to be careful at edges and surfaces that it fills in correctly. Usually there is a combo of large and small agg. Say 15mm and 3mm... But I feel the EM under compression test is biased high especially with big agg. I like the idea of a single size agg that is relative to the smallest gap you need to fill so it has lots of agg in that gap. Say 5mm pea agg then segregation does not occur. Peter

[joeavaerage]

Hi Sus,

Did you do any simulations to understand the stiffness of your machine?

Yes I did. I considered the column/base as one piece. It is constructed of two pieces of plasma cut 32mm steel, separated by bolted 20mm rectangular sections.
The overall width is 240mm and the column or upright part is 1400mm high and the base is 1100mm long. My original intention was to plate the both top and bottom
with welded 10mm plate. This would have in effect created a rectangular tube. This is what I modelled and simulated. Considering just this structure I ended up with a stiffness
of 125N/um in the Z and Y directions and 90N/um in the X direction.

I was trying to achieve an overall machine stiffness of 50N/um, and therefore I thought the results I obtained were credible and good enough.

When I actually came time to construct the machine I ran out of time (business pressures), and towards the end money also. I did not weld the 10mm plates in, I assembled the machine without
them, always meaning to come back and re-do that. The problem was not really the cost of the plates or hiring a welder to weld them in but rather the cost of sending a 250kg steel construction
1500km to be stress relieved.

The simulation results I obtained were with the plating, and are obviously much better than what I've actually got. Subsequent measurement of overall machine stiffness have worked out between 20N/um
and 31N/um depending on the direction of applied force. It is less than I hoped for, but then I've been using this machine daily for the last three years in this condition.....so its usable as is.
If I ever get the time I will strip it down and weld in the plates. I have recently discovered that there is a heat treating oven of sufficient size in Christchurch, so that might save some considerable expense.

Ultimately while you will design and simulate for a given stiffness, you just have to bite the bullet and construct what you think is the best possible compromise and live with the results. With unlimited time and/or
budget you can design and build anything....but that is unrealistic. There are many projects around the world, and even quite a few threads on this forum that never get beyond the design stage.
All that effort in drawing parts, running simulations, testing materials, experimenting with different structures....and that is as far as they have got after several years. I did some design and simulation, drawing
heavily on what I had learnt from my first mini-mill, but then I actually built it. So in the time that others are discussing this and that with all different material and construction ideas my machine is operating
daily and I use it extensively to make a living. Is it perfect?; not at all, is it the best possible combination or parts, materials and techniques?; not at all, does it work?; yes and very nicely thankyou.

Craig

[suspension]

Hi Sus,



Yes I did. I considered the column/base as one piece. It is constructed of two pieces of plasma cut 32mm steel, separated by bolted 20mm rectangular sections.
The overall width is 240mm and the column or upright part is 1400mm high and the base is 1100mm long. My original intention was to plate the both top and bottom
with welded 10mm plate. This would have in effect created a rectangular tube. This is what I modelled and simulated. Considering just this structure I ended up with a stiffness
of 125N/um in the Z and Y directions and 90N/um in the X direction.

I was trying to achieve an overall machine stiffness of 50N/um, and therefore I thought the results I obtained were credible and good enough.

When I actually came time to construct the machine I ran out of time (business pressures), and towards the end money also. I did not weld the 10mm plates in, I assembled the machine without
them, always meaning to come back and re-do that. The problem was not really the cost of the plates or hiring a welder to weld them in but rather the cost of sending a 250kg steel construction
1500km to be stress relieved.

The simulation results I obtained were with the plating, and are obviously much better than what I've actually got. Subsequent measurement of overall machine stiffness have worked out between 20N/um
and 31N/um depending on the direction of applied force. It is less than I hoped for, but then I've been using this machine daily for the last three years in this condition.....so its usable as is.
If I ever get the time I will strip it down and weld in the plates. I have recently discovered that there is a heat treating oven of sufficient size in Christchurch, so that might save some considerable expense.

Ultimately while you will design and simulate for a given stiffness, you just have to bite the bullet and construct what you think is the best possible compromise and live with the results. With unlimited time and/or
budget you can design and build anything....but that is unrealistic. There are many projects around the world, and even quite a few threads on this forum that never get beyond the design stage.
All that effort in drawing parts, running simulations, testing materials, experimenting with different structures....and that is as far as they have got after several years. I did some design and simulation, drawing
heavily on what I had learnt from my first mini-mill, but then I actually built it. So in the time that others are discussing this and that with all different material and construction ideas my machine is operating
daily and I use it extensively to make a living. Is it perfect?; not at all, is it the best possible combination or parts, materials and techniques?; not at all, does it work?; yes and very nicely thankyou.

Craig

Hi Craig

The final stiffness you measured includes everything that matters I guess - such as rails, rail blocks, etc. So overall 20-31 stiffness you get looks to be good.
I am trying to visualize the frame construction based on what you just described and some of the past pictures you posted, but it is not clear to me. Any possibility for you to post a rough sketch of it?

Thanks
Sus

[joeavaerage]

Hi,

The final stiffness you measured includes everything that matters I guess - such as rails, rail blocks, etc. So overall 20-31 stiffness you get looks to be good.

Yes, it includes all the components. It is rigid enough to do aluminum quite nicely and will do steel, but carefully. I had hoped for better, and one day when I get around to plating the top and bottom
sides of the frame it will match my intended design stiffness of 50N/um. It is useful as it is now, and I cant really afford to take it out of service....so when I get around to plating it is still to be answered.

Craig

[suspension]

Hi,



Yes, it includes all the components. It is rigid enough to do aluminum quite nicely and will do steel, but carefully. I had hoped for better, and one day when I get around to plating the top and bottom
sides of the frame it will match my intended design stiffness of 50N/um. It is useful as it is now, and I cant really afford to take it out of service....so when I get around to plating it is still to be answered.

Craig

Hi Craig

This clearly shows how it is built! I honestly admire the simplicity of your designs! I will also try a similar approach for my machine and do a FEA and see.

Did you rely on the plasma cut when getting column portion right angled to the base portion?

Why can't you bolt the L plates to 20mm plates? This way you can avoid stress reliving.

Thanks
Sus

[peteeng]

Hi Craig & Sus - The "open" beam depends on shear stiffness and the "L' maximises the corner stiffness in shear. If you bolt faces to the L (plating it) then the section becomes closed and much much stiffer in bending and torsion. I did some numbers on a 300x300mm grout column with E45. The 300x300mm steel version would only need to be 10mm thick to be the same stiffness. The grout version would weigh 216kg/m3 and the steel would only be 91kg/m. I still see steel as the candidate. Peter

[joeavaerage]

Hi,

I honestly admire the simplicity of your designs!

Thankyou. Simplicity is a very large motivator for why I did things this way. Simplicity has meant a smaller design effort but more in the actual building of it. I date the build from the day
I got the ballscrews from Korea. The ballscrews determined the size of the axis bed castings. From that date it took about one year, from getting the pattern made, having the molds made,
poured and fettled. Then the axis beds were machined and vibratory stress relieved. Took about a month (I took time off work) to construct thereafter. That is to say getting the frame components
plasma cut, making the ballnut mounts, support bearing mounts, saddles etc. and assembling the whole thing.

The essential point is I spent a modest amount of time designing and planning but more time building....and that's why I have a machine to show for it.

Why can't you bolt the L plates to 20mm plates? This way you can avoid stress reliving.

The separators are bolted through the sides, 12mm high tensile bolts torqued until my eyes bulged! This structure is not stress relieved. My intention was to weld plates top and bottom such
to make in effect a rectangular tube. Thereafter it would be stress relieved. Truth is though that this structure is not machined at all, so would stress relieving be strictly necessary??
The Z and Y axis beds (cast iron) bolt to the frame and a shimmed at each corner which holds them orthogonally, so the frame does not require machining at all.

Peter is 100% correct the effective stiffness of the structure depends on the shear stiffness of the right angle. If the structure were plated top and bottom it would become much stiffer again, and that
was my plan....until time and money pressures diverted me.

I have, since I made this, found that there is a heat treating oven of good size still operating in Christchurch and I may be able to get it stress relieved without a trip to Auckland and back, 3000km round
trip. I have also wondered whether I need bother about stress reliving at all, its not like this structure is machined at all, as I noted above.

Second idea is just to make a new frame. The plasma cut steel for this structure cost me $2000NZD (including tax) and weighs 206kg (total). I could just get it re-made, or even re-design and make anew,
maybe with 50mm sides for instance. Only when it is complete that I would strip my machine down and transfer the axis beds to the new frame. In the overall scheme of things $2000NZD is not a huge
amount. The axis beds (all three) including the casting, machining, ballscrews, linear rails/cars and servos cost $16000NZD....so the frame is a small to modest fraction of the investment.

Third idea is to cast the frame in two pieces, a column and a base. Only when that was complete that I would transfer the axis beds to it.

As you can see having a frame as a distinct structure to which you bolt modular axis beds gives you some flexibility with respect to that frame.
The problem is not so much the frame, I can re-make it OR stiffen it by welding plates top and bottom with or without stress relief, Or cast a two piece frame....but its time.
I cannot afford really to take my machine out of service for a month or so. Even if I could take it out of service I don't really have the time, say a couple of hundred hours, to spare.....
my business would probably collapse if I took that sort of time out. Consequently my machine is being used as is, perhaps less stiff than it might be, but still useful as is.

Craig

[joeavaerage]

Hi,

Did you rely on the plasma cut when getting column portion right angled to the base portion?

Yes I did. The right angle is only as good as the machine that did the plasma cutting, but that machine must be pretty damned good, because when I got the 'L' shapes I found the
right angle to be very good by measuring between the tips I found less than 1mm error from square. I was pleasantly surprised!

I have tried to sneak in there and have a closer look at their machines. They have oxy-acet for 50mm and bigger, plasma up to about 40mm, and I'm not sure about the laser, I think its about 20mm
in steel, but they get really crabby about anyone entering their workshop.

You may have seen the very top of the frame poking out above the Z axis bed, note that you are seeing the plasma cut edge, its not machined, nor did I even bother grinding it, painting or anything
else. I did what was required to build/assemble it and start using it.

Craig

[peteeng]

Hi Sus et al - I've just run some generative stuff on a column. I started with a 300x300x1000mm high column. The top, bottom & front I preserved. I told the system to maximise stiffness and to target 100kg. It takes around 22 iterations to get there but it hits 98kg each time in steel. I also shelled it to 20mm thick and it came up with the same geometry (a thick webbed I beam) so its smarter then I thought. I then gave it a torque and it came up with a slanted web at 98kg. Since a machine has torque in two directions this shape isn't good. But if you reflect the shape its a heavy square tube. I shall add rails and cars to the next one. The solid steel column is 750kg....Next run I'll ask it to target 50kg...Peter

from theory and practice a structure in torsion will have no holes in it as the shear stiffness required is the same along its length, as the torque is the same along its length. A cantilever requires a stiffer base then along its span as the bending moment decreases along the span. Ideally if you minimise torsion stress its a circular shape with constant stress along it. In terms of a square or rectangular shape the corners will be the most stressed area and need to be thick to stop the corners from lozenging.

column 1 - generated from solid
column - 2 generated from 20mm thick shell
twist 1 - generated from shell LHS view
twist 2 - same model but other side

So I think the two plate "spine" with faces is a good start. The rails can be mounted along the thick web to transfer shear loads properly some internal webs to aid assembly then plate the outside (bolts or welds) prefer bolts.

[joeavaerage]

Hi peteeng,
generative design is all very well, but is it suddenly and magically produce a design that is twice as stiff or equivalently half the weight for the same stiffness? Were your original design
a real dog....then yes the generative design would be a major improvement, but in our case the basic elements of the design are already well known structures
for which we have intuitive good sense and experience that they are suitably stiff already and generative design may improve it.....but not markedly.

Take my structure for example, and in its complete form it had 10mm plates top and bottom, we know from long experience that sort of structure will perform well.
Is it optimal....probably not, however if it were optimised would it change radically? I rather suspect not.

I am not averse to the idea of using software aided design to improve a basic design, and if you are talking a production run where only a few kilos saved per item represents tons saved
over the run, well yes that is highly worthwhile. In the current circumstance where Sus is building a one-off machine that will never be replicated (completely at least) this sort of generative design is
less useful.

Using my own structure as an example you can very much see that I have focused on simplicity in design and most certainly simplicity in construction to minimise the cost with good
stiffness. My advice to Sus it start with a simple design using structurally sound ideas that is eminently practical to construct using available materials and techniques and then use FEA to analyse it.
If is sufficiently stiff then box on and build the thing. I have used bolting to construct my basic frame but then was going to 'let in plates' and weld. Welding is such a versatile and ubiquitous process
that to ignore it is to miss out on some simple yet extremely stiff structures....if you have to stress relieve then stress relieve.

It should not take months to come up with a simple and stiff design for a hobby machine. Were it a prototype for a production run, then maybe. In the mean time all this design, re-design, re-think,
re-design is burning calories with no tangible result.

Craig

[peteeng]

Hi Craig - GD gives another perspective to the design, another tool in the toolbox. You may say there are precedents to follow but I have been involved in projects where ideas have been generated via AI that have worked out to be novel, doable and far better then the run of the "mill" thinking takes you. That was involved in competitive sport and the result there was a winning one. In fusion GD you can input a required stiffness and it will work at it till done, saving me hrs of work, maybe days. So even if it gets to the "usual" answer then you know its a reasonable answer... To answer your first question, yes I think it will.... Peter

[joeavaerage]

Hi,
this thread has been running for eleven months and Sus has done the great big circuit through EG, UHPC only to arrive at a fabricated, be it bolted or welded or both, steel column design, ie the most
basic tried and trued solution. Surely its time to settle on something and actually make it.

Craig

[peteeng]

Hi Craig - That's up to Sus - Development projects often don't realise. Its the nature of the beast. Auto companies have spent years and B$$$ of certain projects only to shelve them. Same in electronics and I expect ice cream design as well. I've been involved in product development for over 40 years and none of this surprises me... All of the projects in the forum that don't realise; at least they have been documented and the learnings can be passed on. Makers are driven by various mysterious forces. Otherwise we'd just get a second job and buy a machine... IMHO no fun in that.

Project timelines are personal and if the thread contributes to the CNC community then its worthwhile, irrespective of time. .. The Grail is out there. Peter

My current Grail is for a supplier of small harmonic drives, if anyone has one I'd appreciate a contact...

[suspension]

Hi Craig

I did a quick design and simulation of a similar structure to yours but the stiffness I get in X is very low. (Even with 50mm L shaped plates). For example, in the design attached, X stiffness is 37, Y - 62.5, Z - 500. This is probably why I dumped steel plates at the start. Even this structure will cost more than the grout based design I have. I may be doing something wrong in here though. Hopefully if possible, please take a look and let me know.

Although I have not settled with the frame construction, there is good progress in other fronts . I have sourced almost all the parts, learned quite a lot, etc.

Hi Pete

Thanks for the GD based designs. Unfortunately I do not have that facility in my fusion360. But I can get some good insights from these simulations you have done.

Thanks
Sus

[joeavaerage]

Hi,
remember that the design I analyzed was plated top and bottom, much increasing the column torsional rigidity which in turn increases the rigidity in the X direction
Remember also that I have two very stiff cast iron axis beds bolted to the frame which go a long way to simulate the plating effect, namely increasing the torsional rigidity.
The two side plates are locked so that they cannot shear relative to each other and vastly increases the stiffness.

Craig