584,812 active members*
5,324 visitors online*
Register for free
Login
IndustryArena Forum > MetalWorking Machines > CNC "do-it-yourself" > Composite Steel Gantry Mill - Seeking Feedback
Page 1 of 3 123
Results 1 to 20 of 55
  1. #1

    Composite Steel Gantry Mill - Seeking Feedback

    Hello!

    I would love to get feedback on this design that I'm working on. I'd like to order parts as soon as possible, but figured it would be a good idea to get community feedback first in order to avoid any unforeseen problems.

    General specs

    Footprint: 4’ x 4’
    Y Travel: ~30"
    X Travel: ~32"
    Z clearance: 6” (Vacuum chuck/spoiler board will take ~2" of that)
    Transmission: 1610 ball screws
    Guides: 25mm linear rail
    Motors: Nema 23 w/ DM556 drivers (already have on hand, plan to upgrade later)
    Electronics: Arduino Uno w/ GRBL (already have on hand)
    Spindle: 2.2 kw generic Chinese spindle (already have on hand, would like to get a low speed spindle later)


    Design principles
    - Identify and strengthen the weakest link (don’t over-invest in areas that are not the biggest problem).
    - Where possible, achieve rigidity by widening cross sections rather than thickening skins (doubling skin thickness x2’s rigidity, but doubling cross section x4’s rigidity).
    - When making design choices, consider cost, tools at my disposal, and ability to assemble accurately.


    Materials of construction

    Composite skins: steel
    Joining plates: aluminum
    Table core: extruded polystyrene foam
    Core for Y-beams and X-beam: polyurethane foam

    I plan to use epoxy to level the table, X-beam face, and Y-beam sides. I haven’t put much thought into what to do about the meniscus yet. I’ll either use a regular high-viscosity laminating resin from US Composites or a specialized leveling epoxy from Precision Epoxies. I’ve put off researching this since I have been focusing more on other parts of the design.


    Table

    The table is a 4’x4’x2’ polystyrene foam core with 1/8” steel on stop and bottom, and the lightest gauge steel I can find for the sides. I can’t make the table thicker than 2 ft or else I won't be able to get it out of my basement. I would like to make a vacuum chuck for this, and I’m not sure if I will have pipes come up through the table or if the vacuum ducts will come in from the side.

    Reading this manual by Hexcell (https://www.hexcel.com/user_area/con...Technology.pdf) made a few things clear: high performance cores are only necessary in thin panels where the core will be subjected to higher shear stress or if the core will be subjected to point loading. Affordable polystyrene should work fine as a core for the table, since the shear stress will be spread out over a large volume. Local compressive failure won’t be a problem because the workpiece, spoiler board, vacuum table, and steel skin will all help spread out the load. The steel on the sides will be thin and only serve to protect the foam from impact damage. There will also be edge and corner protectors.


    Y-Beams

    My goal with the Y beams is to cut down on the length of the X-beam arms in order to reduce the moving mass and increase the rigidity of the machine. I went with a triangular shape because I think its main weakness will be bending in the X direction. The triangular tube will be made out of 1/4” plate and will be filled with polyurethane foam.


    X-Beam

    This is my thinking: The semi-circular X-beam handles torque better than a square tube, and the flat front face allows me to mount the Z-actuator assembly. One great thing about this gantry is that instead of resonating at one low frequency, like a cylinder or square tube, it will be broken up into several higher pitched nodes. Like the Y-beams, it will be made from 1/4” steel plate and filled with polyurethane, which will not only help with damping, but also significantly raise the natural frequency (not so much by adding mass, but by raising the rigidity of the flat faces).


    Ball Screws

    I haven't completely worked out the mounting yet, so please help me out. If you have better ideas or suggestions for the mounting plates, I would love to hear them. Note: I have to use two screws on the X-axis, otherwise the Z-assembly will be susceptible to torque.

  2. #2

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Here are some preliminary simulation results. I ran these simulations before starting this thread, and I've made some small adjustments since then. I haven't had a chance to run simulations since making the updates since each one takes my computer 2 hours to run . IMPORTANT: these simulations do NOT account for slop in the rail carriages, and assumes the ball screws are perfectly rigid, but I think they are still impressive nonetheless.


    X DEFLECTION
    200N into the end of the spindle produces 3.268 microns of displacement.
    (61N/um)


    Y DEFLECTION
    200N into the end of the spindle produces 3.515 microns of displacement.
    (57N/um)


    Z DEFLECTION
    200N into the end of the spindle produces 4.34 microns of displacement.
    (46N/um)
    You can see that the strain is mostly in one spot on the 0.5" aluminum plate. Instead of making the plate thicker I simply extended the Z-rails to see if that will help, but I haven't run a simulation since then to see the effects.

  3. #3

    Re: Composite Steel Gantry Mill - Seeking Feedback

    The steel skin and foam of the X-beam will weigh about 200 lb. With all the plates, motors, spindle, etc.. it could weigh 350 lb. So do I need bigger Y-rails? And I'm guessing I will need bigger screws, at least for the Y axis.

    Can't wait to get some feedback... Thanks!

  4. #4
    Join Date
    Jan 2018
    Posts
    9

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Do you have a good target for stiffness? The Bamberg thesis is the only source I can find that lists stiffness values as targets for performance classes (10-25 N/um for VMC, 50 N/um for a precision grinder)

    I have been struggling with the fact that stiffness sums in the inverse, so the lowest stiffness has an outsized magnitude effect on the overall stiffness of the system. The entry level 25mm rails are ~600 N/um radial stiffness, but my frame is 500 lb before I get to 200 N/um. My 200 N/um target is so that the system stiffness ends up being 50N/um before considering the spindle. I don't know if that is a good target for my goal of a bench-top mill that thinks it is a Kern.

    You will get a much faster run time if you "schematic" your rails. Have a configuration in your model that fills in all of your mounting holes and whatnot that don't have a large impact on the result.

    Do you have a good grasp on the concept of cross-sectional moment of inertia? My reading comprehension has been terrible lately. The "rule of thumb" is to have as much material as possible as far away as is possible from the neutral axis, assuming that the strain is effectively transferred. You seem to have the first part on point, a high modulus (steel) skin with a large perimeter section. I'm a bit concerned with the second point. The skin might be much more effective if you added ribs to prevent buckling.

    Piotr Fox Wysocki recently posted the models for his build that may be useful for you to peruse. My simulation showed that his gantry was a huge weak point of the design. 20 N/um if I remember correctly.

    What kind of accuracy are you hoping for? I don't know much about the epoxy leveling. My uneducated impression is that it is used to make "Class C" surface plates, like you would use to measure the suspension in motorsports.

    Can you explain how you applied fixtures in your simulation? it looks like the whole bottom surface of the rail mounts are fixed. This would give you an absurdly high result for stiffness. A more correct way would be to serarate each component between one point of interest to the next (eg, your Y-Carriages to your Z-carriages) and treat one side as fixed and the other as an applied load. The applied load would be the equivalent force as would be expected from your cutting force. For example, my applied load was a pair of forces of 2N that twists the gantry/bridge to represent cutting in -Y.

    Also, go with the MEGA2560. The few extra I/O are worth it.

  5. #5
    Join Date
    Jul 2018
    Posts
    6248

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Hi Paul - You have to establish a target static stiffness to get the benefit of FE design. Also you need to place a dummy vice or typical fixture on your bed and restrain that while putting a load on a dummy tool. Make the tool infinite stiffness so its not deflecting. You will hit trouble with thin skins. If you look up the Milli thread you will see the development of composite/steel / aluminium and concrete designs over the last few months. You say its a MIll and Blockerra mentions Kern. If you are aiming at a modern typical VMC stiffness then you will have to achieve 150N/um static (and Kern are above that) by observation and 100's of hrs on FE you aren't anywhere near that yet. Epoxy levelling is dubious IMO. Precision and others specify a minimum thickness of 6mm. Epoxy has a stiffness of 3GPa so its really not much better then a rubber sheet to mount parts on. If you go thinner then 6mm surface tension, exothermic expansion and various flow effects take over and it won't self level.

    Mills are thick to remove local deflection issues and vibration. Thin skinned sandwich cores are fine for global stiffness but in a mill the size you have you run the problem of getting loads to transfer to the global structure effectively, so local deflections dominate and are large. If you have a non linear solver on your system (or a large deflection solver) its worthwhile running that to pick this sort of thing up. Linear Static solvers are very mesh sensitive to thin/thick structures. You need a mesh with at least 3 elements across the thinnest part. Good Luck keep at it you have a while to go... Peter

    The Bamberg thesis is really good but its 20 years old and modern machines are much much stiffer then quoted in his work... The stiffest machine I have found is 650N/um measured... typically a bonded model would be 50% efficient compared to reality. So if you want to get to 50N/um you need a model at a 100... and as the total stiffness is the sum of the inverse stiffness "structural loops: of the system, as mentioned the least stiff "loop" can be the dominant compliance....

    eg if you have retrained the bottom of the triangular rails then its over restrained and will be considerably stiffer then when you place it on a machine base...Keep at it Peter

    You have shown the deflection with the Z at its lowest position, you need to do it at its highest position (least stiff position)

  6. #6
    Join Date
    Jan 2018
    Posts
    9

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Quote Originally Posted by peteeng View Post
    Blockerra mentions Kern.
    My machine is designed to think its a Kern in the same way that a security guard thinks they're a cop.

    Quote Originally Posted by peteeng View Post
    If you are aiming at a modern typical VMC stiffness then you will have to achieve 150N/um static (and Kern are above that) by observation and 100's of hrs on FE you aren't anywhere near that yet.

    The Bamberg thesis is really good but its 20 years old and modern machines are much much stiffer then quoted in his work... The stiffest machine I have found is 650N/um measured... typically a bonded model would be 50% efficient compared to reality. So if you want to get to 50N/um you need a model at a 100... and as the total stiffness is the sum of the inverse stiffness "structural loops: of the system, as mentioned the least stiff "loop" can be the dominant compliance....
    Do you have a suggestion of a direction to read into for a stiffness value as it relates to performance? I took a class in "research methods for technical projects" but the class was counter-productively scheduled beside a year 1 nursing class.

  7. #7
    Join Date
    Jul 2018
    Posts
    6248

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Hi Block and others - You have to make the stiffest machine possible within your resources. Hobby machines have been measured at 1N/um up to 5N/um. They cut steel just, but poorly. A modern VMC will be at least 150-200N/um. One extreme machine I found quoted 650N/um for cutting titanium in aerospace stuff. I can't find where I snipped the VMC stiffness from but it was about 20 years old data. Bamburg covers most things technically correctly although I think his conclusion has some bias when you read his work maybe he had changed direction a little at the time.

    There are so many people now with access to FE that its actually learning about FE not about machine stiffness. For rigidity just pick a number and shot at it. But depending on how you restrain the model, how you interpret bonded connections and bearing compliance and bolt compliance/friction all needs to be factored in unless you model these things correctly in detail. For instance a linear static solver assumes that the global shape of the structure is the same before load is applied and once the load is applied ir load & deflection history is ignored. In this case with 6mm steel facings on styrofoam "oil canning" and membrane effects are neglected. A 6mm steel sheet like this can be pushed in with your fingers, Pauls rails are sitting on "air" and this initial membrane deflection is ignored by the static solver. If you built this and statically tested it you maybe disappointed FE vs reality. If you use too coarse a mesh in a model that has thin and thick bits and you relied on the automesher to mesh the structure you will also be disappointed again. So you have to read up on how the mesher works and what a good mesh should look like. Automeshers are remarkably good these days (I come from a time where I built meshes manually took days) but garbage in garbage out is still applicable. I use very high end FE and I still have to run mesh checks and sanity checks before I believe some outputs. Then there is real or dynamic stiffness vs static stiffness and that is a different matter. Just make it as stiff as you can, a machine can never be too stiff. Make simple models and get a good understanding of what they are doing, if your capable do manual calc checks. Then build up in complexity. I see quite a few complex models in here and weird data... I use a design load of 1000N for rigidity checking so the maths is simple., I also use a forced deflection of 0.001mm and the FE figures out what load is needed to move it 1um. Saves you doing the math. Then depending on your motion system forces pick a high number to check things don't break. If its stiff enough the stresses are very low. Check my threads for Brevis, Milli and Maximus these cover most areas of machine design. Good luck and keep at it.. Peter

  8. #8
    Join Date
    Jan 2021
    Posts
    8

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Axis travels:

    X: 4 to 50 m; Y: 3.5 to 10 m; Z: 1 to 3 m

    Milling head performances: (Electro spindle / Gear driven spindle)

    Power: up to 150 kW / 45kW
    Torque: up to 300 Nm / 1200 Nm
    Speed:

    Up to 30,000 rpm / 6,000 rpm

  9. #9
    Join Date
    Mar 2021
    Posts
    41

    Re: Composite Steel Gantry Mill - Seeking Feedback

    For the controller I love grbl and have tried many variants. You don't want the old UNO, google grblHal. If you want spindle control, 6 axis, Bluetooth, RS485 and look at the ESP32. Or my favorite w 5 axis, spindle control and Ethernet look at the teensy. The form factor of the teensy by Phil Barret is much better. My router is the teensy (green) and my gasket machine is the esp32 (black) as seen in the pics.

  10. #10

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Thanks everyone for the responses.

    Quote Originally Posted by Blockerra View Post
    Do you have a good target for stiffness?
    First of all, thanks for the response.

    I am less concerned with chasing a specific N/um figure than I am with making a balanced machine that performs well in relation to its cost.

    I plan to use the machine for the following:
    - grills for speaker cabinets (mild steel)
    - aluminum corner protectors from sheet aluminum
    - aluminum tooling for low volume RTV plastic part production (speaker cabinet handles)

    In the future I may want to:
    - mill aluminum faceplates for amplifier chassis's
    - use the machine for milling aluminum/steel for unexpected hobby purposes

    I hope that gives you an idea of what I want the machine to be able to do. As you can see, it's not really aerospace level stuff.

    Quote Originally Posted by Blockerra View Post
    The skin might be much more effective if you added ribs to prevent buckling.
    The foam effectively transfers the load between skins. If lightweight core materials were not capable of transferring the load between skins then composite sandwich panels would not be a viable geometry. Check out the sample problems in this Hexcell PDF: https://www.hexcel.com/user_area/con...Technology.pdf. It is written about honeycomb composites but it applies just as well to non-honeycomb composites. According to those formulas, my structures are sound.

    The table is a textbook sandwich panel in which the skins are under tension and compression, and the core is under shear. I considered adding metal supports in the foam, but the formulas and sample problems in the Hexcell PDF showed me that that would actually be a waste of time and money. The modulus and strength of the core hardly factors into the formulas for deflection at all.

    In the case of the Y-beams and X-beam, the foam is only adding local rigidity to the faces of those beams. Doubling the cross section of a beam x4's it's rigidity. As you increase the cross section the beam gets more and more rigid, but the unsupported length of the faces of that beam grows and becomes exponentially less stiff. So the overall stiffness is growing, but the rigidity in the center of the skins of those faces gets worse and worse. This is where the foam comes in, as it increases rigidity in those spots in the beam that would otherwise be flabby and resonant.

    Quote Originally Posted by Blockerra View Post
    Can you explain how you applied fixtures in your simulation? it looks like the whole bottom surface of the rail mounts are fixed. This would give you an absurdly high result for stiffness. A more correct way would be to serarate each component between one point of interest to the next (eg, your Y-Carriages to your Z-carriages) and treat one side as fixed and the other as an applied load. The applied load would be the equivalent force as would be expected from your cutting force. For example, my applied load was a pair of forces of 2N that twists the gantry/bridge to represent cutting in -Y.
    I want to make sure I understand you correctly. In Fusion 360 there are "constraints" (used to account for objects that are not directly represented in the model but that factor in somehow structurally) and "contacts" (used to describe the relationship between different bodies that are directly represented in the model). I used 4 constraints: one on the long edge of the bottom of each Y-beams. I used those constraints because I did not run the table in the simulation for simplicity's sake, although I plan to. There are no constraints directly on the rails if that is what you mean. All of the contacts between bodies are defined as rigid. I thought there would be less error if I modeled as much of the machine at one time as I could, although I left out the table because 1) I was just doing preliminary simulations, 2) it would have taken my computer much longer, and 3) the table is very stiff.

    Blockerra, thanks for the resources and input!



    Hi, Peter, and thanks for the response.

    Quote Originally Posted by peteeng View Post
    Hi Paul - You have to establish a target static stiffness to get the benefit of FE design.
    I don't agree with this. FEA has already helped me identify a couple weaknesses I would have not seen until after building it. Setting a specific N/um figure would be fairly arbitrary, since there seems to be no agreement on how to equate that N/um figure into real life machine capabilities. If 10-25 N/um was good enough for a professional mill 20 years ago, then it should suffice for my purposes today, even if it won't be the most cutting-edge aerospace mill ever made.

    Quote Originally Posted by peteeng View Post
    Also you need to place a dummy vice or typical fixture on your bed and restrain that while putting a load on a dummy tool. Make the tool infinite stiffness so its not deflecting.
    Thanks, that is a great idea!

    Quote Originally Posted by peteeng View Post
    You say its a MIll and Blockerra mentions Kern. If you are aiming at a modern typical VMC stiffness then you will have to achieve 150N/um static (and Kern are above that) by observation and 100's of hrs on FE you aren't anywhere near that yet.
    Unfortunately, I'm not doing contract work for NASA yet. My goal with this design is to achieve a very high level of performance in relation to cost and weight.

    Quote Originally Posted by peteeng View Post
    Epoxy levelling is dubious IMO. Precision and others specify a minimum thickness of 6mm. Epoxy has a stiffness of 3GPa so its really not much better then a rubber sheet to mount parts on. If you go thinner then 6mm surface tension, exothermic expansion and various flow effects take over and it won't self level.
    I suspect that this will not be a problem because of how the structure will be loaded. There is a large surface area between the Y-beams and the table, and the forces are well distributed. Even though the epoxy itself isn't very stiff, it will still increase the vertical bending stiffness of the Y-beams by increasing the distance between the bottom of the Y-beam and the table. The bottom of the Y-beam, epoxy, and table now present sandwich structure of itself, albeit ones with asymmetrically stiff skins.

    Quote Originally Posted by peteeng View Post
    Mills are thick to remove local deflection issues and vibration. Thin skinned sandwich cores are fine for global stiffness but in a mill the size you have you run the problem of getting loads to transfer to the global structure effectively, so local deflections dominate and are large. If you have a non linear solver on your system (or a large deflection solver) its worthwhile running that to pick this sort of thing up. Linear Static solvers are very mesh sensitive to thin/thick structures. You need a mesh with at least 3 elements across the thinnest part. Good Luck keep at it you have a while to go... Peter
    Where exactly would you expect to see local deflections in my design? I just want to make sure that I understand you correctly: Are you saying there will essentially be some "slop" in the faces of the skins that has to be stretched out before it starts to be loaded? If so, I believe this would only be a potential problem on the front face of the gantry, which is 2 feet tall and made of 1/4" plate. All the other faces are relatively narrow compared to the thickness of the plate. Thanks for the input, how do I go about getting a non-linear solver? The only simulations I've run are static tests on Fusion 360.

    Quote Originally Posted by peteeng View Post
    The Bamberg thesis is really good but its 20 years old and modern machines are much much stiffer then quoted in his work... The stiffest machine I have found is 650N/um measured... typically a bonded model would be 50% efficient compared to reality. So if you want to get to 50N/um you need a model at a 100... and as the total stiffness is the sum of the inverse stiffness "structural loops: of the system, as mentioned the least stiff "loop" can be the dominant compliance....
    I'm definitely not going for 650 N/um here. But it is good to know that I may be able to expect 50% of simulated results. It will be really interested to build this thing and see how it compares to simulations. I will be very happy if it is 50% of simulated results!

    Quote Originally Posted by peteeng View Post
    eg if you have retrained the bottom of the triangular rails then its over restrained and will be considerably stiffer then when you place it on a machine base...Keep at it Peter

    You have shown the deflection with the Z at its lowest position, you need to do it at its highest position (least stiff position)
    Yes I restrained the bottom of the triangular rails for simplicity's sake since I was just working on optimizing the top portion of the machine.

    Quote Originally Posted by peteeng View Post
    A 6mm steel sheet like this can be pushed in with your fingers, Pauls rails are sitting on "air" and this initial membrane deflection is ignored by the static solver.
    Okay, I see what you were saying earlier. To clarify, the gantry face is actually not sitting on air, it is attached to foam. That may not seem like a big deal, but it is. Foam may be a weak material, but it doesn't need to be strong for the job it is doing.

    Peter, thanks for your time in responding and for all the attachments that you generously provided. I can't wait to check them out.

  11. #11

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Waynekofuco, thanks for joining the thread.

    I know about grblHAL. I've heard that it is not very user friendly yet since it is new. Is that correct? I only need 3 axis, but I know grblHAL has some other features that I might benefit from in the future. I'm not especially savvy with electronics and software, I will have to use something fairly user friendly. Since I have the Uno/GRBL already and don't need a lot of advanced features, I would need a compelling enough reason to switch.

  12. #12

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Quote Originally Posted by peteeng View Post
    A 6mm steel sheet like this can be pushed in with your fingers, Pauls rails are sitting on "air" and this initial membrane deflection is ignored by the static solver.
    It seems like you are combining two separate potential problems:

    1) Slack introduced during fabrication.
    I don't see this being a problem.

    2) Rigidity of the face plate in its center.
    One of us is misunderstanding how the foam works in this structure.

    The foam provides continuous support to the face plate and transfers shear between the different skins. When you bend something, the outside layers gets put under tension and compression, and the core transfers shear between them. If the steel face is being pressed on while "resting on air", like in your example, a large portion of the inside of the plate is being put under shear, which is inefficient. The benefit of filling the beam with foam is that now when you press on the front face plate, the face is only being put under tension, the foam is seeing the sheer load, and the backside of the beam is being put under compression. This is much more efficient, and in the example you gave, would bring the whole beam into the equation faster.

  13. #13
    Join Date
    Mar 2020
    Posts
    179

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Thoughts for the OP-

    Foam core: what about thermal expansion? Polystyrene has 7x the CTE of steel (https://www.engineeringtoolbox.com/l...ents-d_95.html). Combining the two sounds like an absolute nightmare to me trying to get stable, precision surfaces. Probably just won't happen.

    I also wouldn't dismiss what Pete is saying about local vs global stiffness. The products that do use foam as core in panels structually, i.e. boats, etc, use the foam to create panels with high-modulus materials on BOTH sides of a relatively thin foam core, so that the high-modulus skins are doing most of the work. And, these structures as Pete noted may have good global stiffness due to the shape, but generally poor local stiffness. On lightweight boat hulls "oilcanning" is definitely a thing. By doing it your way, with a very thick core, you're relying on the foam a whole lot more, such that a vertical load on the middle of the table is essentially only resisted by the compressive modulus of the foam.

    Epoxy leveling: many people talk about this. I looked for reports of actual successes with the technique (in context of CNC linear rail mounting) and couldn't find any. Or at least not enough to want to commit to it. It's possible, but it seems like most people that were successful ending up doing a lot of manual scraping to get them actually flat. I decided not to do it, personally. Also I would definitely use a very low viscosity epoxy, the laminating grades are pretty thick and unlikely to self-level to a machine-precision quality

    My advice? If you want accuracy, use a granite surface plate for the table. I went to all sorts of trouble and expense to fabricate, heat treat, and machine a welded steel structure, which is certainly about max rigidity given the design and is the shape I wanted, but a granite surface plate would be a lot easier with guaranteed flatness. Of course if you really need the 30x32" work area that's going to be a very large piece of granite. I'd try to make it work with a 24x36 surface plate. Only have to buy that then drill some holes, instead of all the fabrication fuss.

    Also, how are you fabricating the gantry? Are the steel panels edge-welded together? If so, that's gonna create a lot of distortion. If not, the material isn't working as a single piece. You could bend tabs and bond with bolts along the edges, I guess.

  14. #14
    Join Date
    Jul 2018
    Posts
    6248

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Hi Paul - You are describing what's called "sandwich action" for this to work the entire element needs to be placed in bending. The sandwich is the same as an I beam. The web transfers shear load (analogous to the foam as you explained) But this is a global effect not a local effect. What mechanical properties are you using for the foam? For a local load the foam is not stiff enough to transfer direct loads. If there is a "hard point" ie a place where load is being introduced or being reacted with, typically a stiff web is placed there to deal with the local conditions. Even in structural I beam steel construction you will see shear webs at connections to keep the flanges stable.

    The sandwich principle in global bending transfers longitudinal shear from one skin to the other. Transverse shear is a different matter. Again if you push on styrofoam with your finger it will crush. The steel skin smears the load but local deflections still occur and its deflections that machine design is all about. In your design the rails maybe close enough to the corners so this deflection does not show up much. Run the same model without the foam and see what happens you may find the foam is not actually doing much...

    If you publish your foam properties I shall model something to demonstrate... Peter

    I appreciate you don't want to do aerospace contracts but you do want to Mill aluminium, if you look at commercial machines of the same size that you plan, you will see they are tonnes of cast iron. They would not do this if they did not have to. Many people think that they can just go slow with a less stiff machine. This is not the case. A Mill to cut aluminium efficiently has to be very stiff.

  15. #15
    Join Date
    Jul 2018
    Posts
    6248

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Hi Paul - To explain longitudinal and transverse shear.
    1) If you get a soft cover book and hold it up so its a beam and get someone to push down in the middle you will notice as it bends the pages slide over each other. This is the shear that is being resisted when acting as a beam. This is what the I beam web and foam core does this is a global effect. 2) If you hold up a loaf of sliced bread as a beam and push in the middle the slice you push on tries to slide down the loaf. This is transverse shear and can be local or global. If the foundation of the skin is not stiff as they generally are then the skin deflects and compresses the core see image. 6mm steel is a stiff skin and may not do this, modelling can determine this... but linear solvers may not pick up small transverse deflections due to membrane action... you need to do a test model to figure that out with your FE system.

    Typically in sandwich construction (aerospace, marine, industrial) at hard points the core is removed and it goes back to a solid or a web is introduced at that point. If you look at typical cast iron mill construction they always have rails mounted on a web never on a "membrane" ie something that can deflect like a hammock. Foam or air same thing in terms of machine stiffness requirements. The same comment is true when a designer places a rail in the middle of a SHS or RHS side, its on air and can deflect slightly, plus it can vibrate. Thin panels under transverse load deflect easily until membrane tension develops then the global structural stiffness takes over. These are fine points but you seem interested in this area and its in development/CAD time that these things can be eliminated... Peter

  16. #16
    Join Date
    Mar 2021
    Posts
    41

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Quote Originally Posted by catahoula View Post

    Epoxy leveling: many people talk about this. I looked for reports of actual successes with the technique (in context of CNC linear rail mounting) and couldn't find any. Or at least not enough to want to commit to it. It's possible, but it seems like most people that were successful ending up doing a lot of manual scraping to get them actually flat. I decided not to do it, personally. Also I would definitely use a very low viscosity epoxy, the laminating grades are pretty thick and unlikely to self-level to a machine-precision quality
    I could not agree more with this. The problem with low viscosity though is the cure shrinkage and apparent low self surface tension that creates a bowl effect. I attempted to sand the bowl out of my 6x6x24 gantry tube using my 12x18 granite surface plate to my 'leveled' tube, but it was futile. I wound up removing the all the leveling epoxy and sandwiched JB cold weld between the tube and surface plate to fill the majority of the natural concavity of the tube. What this revealed was how unflat/level my lapping hade made the tube. My fix was to use jack screws on mating surfaces and inject JB weld into the gaps. You might be able to see what I mean in the pics.

  17. #17

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Quote Originally Posted by catahoula View Post
    Foam core: what about thermal expansion?
    First off, great name! I love my Catahoula more than I ever thought I could love a dog.

    Thermal expansion is a really good point. I will have to think about that. It is probably possible for me to add chopped fiber to the PU for the sole purpose of balancing the expansion rate to that of steel. I believe CFRP has an approximately neutral thermal expansion rate because the fiber and epoxy balance each other out.

    Quote Originally Posted by catahoula View Post
    I also wouldn't dismiss what Pete is saying about local vs global stiffness. The products that do use foam as core in panels structually, i.e. boats, etc, use the foam to create panels with high-modulus materials on BOTH sides of a relatively thin foam core, so that the high-modulus skins are doing most of the work. And, these structures as Pete noted may have good global stiffness due to the shape, but generally poor local stiffness. On lightweight boat hulls "oilcanning" is definitely a thing.

    Sandwich technology is used in many industries besides boatbuilding, especially aerospace. My structures are covered on all sides by a high modulus skin (steel), although it is asymmetrical in some cases. Sandwich structures CAN have poor local stiffness. That depends on the overall geometry. The lightweight boat hulls you are talking about are deflecting because the core is thin. Deflection is of course dependent on other factors as well: skin modulus, thickness, core thickness and modulus, but the core thickness has an exponential effect whereas the skin thickness and modulus have a linear effect. The core modulus factors in very little. I'm sure boat designers weigh out all the variables and have to balance many factors including cost. Since I have never built a boat, I would assume that if many boats have this problem, then designers don't give this aspect very much priority for whatever reason. My design is different in a vast number of ways than a cheap boat.

    Quote Originally Posted by catahoula View Post
    Epoxy leveling: many people talk about this. I looked for reports of actual successes with the technique (in context of CNC linear rail mounting) and couldn't find any. Or at least not enough to want to commit to it. It's possible, but it seems like most people that were successful ending up doing a lot of manual scraping to get them actually flat. I decided not to do it, personally. Also I would definitely use a very low viscosity epoxy, the laminating grades are pretty thick and unlikely to self-level to a machine-precision quality
    Perhaps I can be a guinea pig then!

    By laminating grade, I mean the kind used for laying carbon fiber or fiberglass, not the kind used to laminate bar countertops. The epoxy used for laminating cloth is as thin as you can get without going to a specialty resin system. I haven't figured out yet if the specialty leveling epoxies are pretty much the same thing as the laminating epoxies but relabeled, or if they are actually superior.

    Quote Originally Posted by catahoula View Post
    By doing it your way, with a very thick core, you're relying on the foam a whole lot more, such that a vertical load on the middle of the table is essentially only resisted by the compressive modulus of the foam.
    That is not how sandwich structures work. Read through the sample problems in the Hexcel PDF on page 14. https://www.hexcel.com/user_area/content_media/raw/Honeycomb_Sandwich_Design_Technology.pdf

    Quote Originally Posted by catahoula View Post
    My advice? If you want accuracy, use a granite surface plate for the table. I went to all sorts of trouble and expense to fabricate, heat treat, and machine a welded steel structure, which is certainly about max rigidity given the design and is the shape I wanted, but a granite surface plate would be a lot easier with guaranteed flatness. Of course if you really need the 30x32" work area that's going to be a very large piece of granite. I'd try to make it work with a 24x36 surface plate. Only have to buy that then drill some holes, instead of all the fabrication fuss.
    A huge granite surface plate would put this machine in an entirely different price bracket. Regular kitchen granite countertops are not flat enough, I already checked. If I could have scaled down the work area for what I need to mill, I would have already.

    Quote Originally Posted by catahoula View Post
    Also, how are you fabricating the gantry? Are the steel panels edge-welded together? If so, that's gonna create a lot of distortion. If not, the material isn't working as a single piece. You could bend tabs and bond with bolts along the edges, I guess.

    Yep, edge welding. Yes, it will create heat distortion. Can you describe to me why this will be a significant problem? Like Peter mentioned, everything is a compromise.

    Quote Originally Posted by peteeng View Post
    Hi Paul - You are describing what's called "sandwich action" for this to work the entire element needs to be placed in bending. The sandwich is the same as an I beam. The web transfers shear load (analogous to the foam as you explained) But this is a global effect not a local effect. What mechanical properties are you using for the foam? For a local load the foam is not stiff enough to transfer direct loads. If there is a "hard point" ie a place where load is being introduced or being reacted with, typically a stiff web is placed there to deal with the local conditions. Even in structural I beam steel construction you will see shear webs at connections to keep the flanges stable.
    A sandwich structure is LIKE an I-beam, but it differs in that a sandwich structure has continuous support, whereas an I-beam has only local support. You are correct that point loading can cause local compressive failure if the applied load divided by area exceeds the core compressive strength. but I believe that the linear rails and steel skin will more than adequately diffuse the load.

    Quote Originally Posted by peteeng View Post
    The sandwich principle in global bending transfers longitudinal shear from one skin to the other. Transverse shear is a different matter. Again if you push on styrofoam with your finger it will crush. The steel skin smears the load but local deflections still occur and its deflections that machine design is all about. In your design the rails maybe close enough to the corners so this deflection does not show up much. Run the same model without the foam and see what happens you may find the foam is not actually doing much...
    Yes, if you push on bare styrofoam with your fingers you can crush it. On the other hand, a major use for styrofoam is going underneath concrete foundations of buildings. It is all about how well the load is spread out. As I've said, in the Y-beams and X-beam, the foam is not adding to the global rigidity of the beams or adding rigidity near the corners, it is only adding rigidity to the middle of the faces which would otherwise be unsupported.

    Quote Originally Posted by peteeng View Post
    If you publish your foam properties I shall model something to demonstrate... Peter
    That would be great! Unfortunately Fusion 360 has been stuck running a simulation all day. I will let you know what properties were used in the simulation after it unfreezes.

    Quote Originally Posted by peteeng View Post
    I appreciate you don't want to do aerospace contracts but you do want to Mill aluminium, if you look at commercial machines of the same size that you plan, you will see they are tonnes of cast iron. They would not do this if they did not have to. Many people think that they can just go slow with a less stiff machine. This is not the case. A Mill to cut aluminium efficiently has to be very stiff.
    Frankly, this kind of mindset cannot lead to innovation. Standard practices always exist for good reasons, but they are not always perfect. I just want to make sure I understand... are you telling me that it is impossible to make a machine that will adequately serve my purposes without tonnes of cast iron? If so, I happily take on the challenge! The weight of a machine is only a loose indication of its rigidity. It is possible to create a much less or much more rigid machine depending on the geometry you choose.

    Quote Originally Posted by peteeng View Post
    If you get a soft cover book and hold it up so its a beam and get someone to push down in the middle you will notice as it bends the pages slide over each other. This is the shear that is being resisted when acting as a beam. This is what the I beam web and foam core does this is a global effect. 2) If you hold up a loaf of sliced bread as a beam and push in the middle the slice you push on tries to slide down the loaf. This is transverse shear and can be local or global. If the foundation of the skin is not stiff as they generally are then the skin deflects and compresses the core see image. 6mm steel is a stiff skin and may not do this, modelling can determine this... but linear solvers may not pick up small transverse deflections due to membrane action... you need to do a test model to figure that out with your FE system.
    Do you think that the X-beam will actually see a load like you are describing on its face? That is an honest question. Thanks.

    As it stands, none of the structures in this design will see point loading significant enough to fail under local compression.

    Quote Originally Posted by peteeng View Post
    Typically in sandwich construction (aerospace, marine, industrial) at hard points the core is removed and it goes back to a solid or a web is introduced at that point. If you look at typical cast iron mill construction they always have rails mounted on a web never on a "membrane" ie something that can deflect like a hammock. Foam or air same thing in terms of machine stiffness requirements. The same comment is true when a designer places a rail in the middle of a SHS or RHS side, its on air and can deflect slightly, plus it can vibrate. Thin panels under transverse load deflect easily until membrane tension develops then the global structural stiffness takes over. These are fine points but you seem interested in this area and its in development/CAD time that these things can be eliminated... Peter
    Whether or not it is appropriate to add local reinforcements to the core is dependent on the magnitude and area of the applied load. No, foam and air not the same thing; I'm not sure how to respond to that. Yes, it is silly to mount rails in the middle of large RHS, but that is a false analogy. Cast iron is not the only way to make a great machine.

    Quote Originally Posted by Waynekofuco View Post
    I could not agree more with this. The problem with low viscosity though is the cure shrinkage and apparent low self surface tension that creates a bowl effect. I attempted to sand the bowl out of my 6x6x24 gantry tube using my 12x18 granite surface plate to my 'leveled' tube, but it was futile. I wound up removing the all the leveling epoxy and sandwiched JB cold weld between the tube and surface plate to fill the majority of the natural concavity of the tube. What this revealed was how unflat/level my lapping hade made the tube. My fix was to use jack screws on mating surfaces and inject JB weld into the gaps. You might be able to see what I mean in the pics.
    I think you did not plan out the epoxy leveling thoroughly enough. The meniscus has to be accounted for. I haven't gotten to planning it out very much yet, but a few options come to mind:

    1) Temporarily extend the surface of the object being leveled, then break off the extra surface containing the meniscus
    2) Leave the meniscus on the object permanently and plan to work within its perimeter
    3) Choose the correct barrier so that the meniscus is about neutral. This would be addition to option 1 or 2. Maybe Precision Epoxies could guide me on if this is possible.

  18. #18
    Join Date
    Jul 2018
    Posts
    6248

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Hi Paul - You seem to be set on a course so I shall only respond to significant stuff. 1) edge weldiing of a 6mm box such as you have drawn will introduce significant distortion. I have worked as a welder so speak from experience. By significant I mean relative to building a CNC mill. How will you level a surface as you propose, more epoxy? 2) laminating grade epoxies are thixotroped ie they have additives to make them NOT flow, they develop hot spots when curing and crawl and orange peel. Been there done that...Model a 6mm epoxy bed under your rails and see how they go... 3) you mention strength a lot. There is no element in a CNC machine that is under threat of strength failure 4) every part of a machine deflects and every micron adds up to lots of deflection at the tool. Especially parts that are a long way away from the tool due to magnification of the deflection. You have held the base of your columns as infinitely stiff. This over restrains the model. 4) Polystyrene is 80-90% air so in terms of machine design its air...Its great to be innovative but most innovations fail at its first , second and maybe 3rd attempt. This forum is a shark tank... take note of what people say it will save you lots of $$$ and heartache. 5) no, machines do not have to be made from CI but it is a benchmark and machine builders are moving into concrete and mineral casting. I have modelled a lot of sandwich machine bases and they have not worked out in my simulation experience... I would pick concrete before I went down your current path... Peter

  19. #19
    Join Date
    Mar 2021
    Posts
    41

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Quote Originally Posted by PaulAmelang View Post
    I think you did not plan out the epoxy leveling thoroughly enough. The meniscus has to be accounted for. I haven't gotten to planning it out very much yet, but a few options come to mind:

    1) Temporarily extend the surface of the object being leveled, then break off the extra surface containing the meniscus
    2) Leave the meniscus on the object permanently and plan to work within its perimeter
    3) Choose the correct barrier so that the meniscus is about neutral. This would be addition to option 1 or 2. Maybe Precision Epoxies could guide me on if this is possible.
    You may be right, but I use epoxy quite a bit for the product I build. I'll offer the following. Leveling epoxies are designed to have super low surface tension so as to find the lowest point of any given area. This is the meniscus effect. This does not account for shrinkage. Leveling epoxies shrink. You can mitigate the effect with fillers like calcium carbonate, but this also mitigates the leveling effect. Yet another issue is the exothermic properties epoxy directly related to volume. In areas where there are variations of the volume of epoxy will change the cure and leveling effect.

    Ill ad another option to your list that I think would yield a 'leveling' outcome.

    4) Do not attempt to contain the epoxy at all. Protect the sides and allow any excess drip/run off. Mind you this wont be a clean process. It will force self surface adhesion making it somewhat convex that will offset the convexing nature of the cure.

    Obviously the above requires that the surface being leveled is concave and not convex. Oh, and lets hope there is no twist in the steel, convex or not...

  20. #20
    Join Date
    Jul 2018
    Posts
    6248

    Re: Composite Steel Gantry Mill - Seeking Feedback

    Hi Paul - To get back to your original assertion that a semi-circular section is stiffer then a square section. If an RHS is made the same envelope as the semicircle it is 66% stiffer then the semi-circular section in torsion. It is also stiffer in the other directions. The examples are 200mm high and 100mm deep... but actual dimensions do not matter... Peter

Page 1 of 3 123

Similar Threads

  1. Seeking feedback on large bench mill design
    By embraced in forum Benchtop Machines
    Replies: 9
    Last Post: 10-04-2015, 07:48 AM
  2. Replies: 6
    Last Post: 05-23-2015, 02:42 AM
  3. New steel mill/router design. Feedback appreciated.
    By geekness in forum CNC Wood Router Project Log
    Replies: 11
    Last Post: 06-19-2014, 05:24 PM
  4. 2nd Build - Steel Tube, Moving Gantry Style, Mill / Router
    By InventIt in forum Vertical Mill, Lathe Project Log
    Replies: 73
    Last Post: 02-26-2012, 08:15 AM
  5. Composite Gantry
    By Karlf in forum DIY CNC Router Table Machines
    Replies: 10
    Last Post: 12-25-2007, 09:31 AM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •