[peteeng]

Hi Ard - There is an old engineering saying that covers a lot of situations. Figure out how stiff it has to be, then make it 10x stiffer. So there are two ways to estimate the machine stiffness as I see it:
1) Look at successful machines for the job and make your machine the same stiffness. say 10-200N/um stiffest machine I found was 650N/um
2) Looking at FSwizard pick a 10mm slot mill - cutting force 12kg in aluminium. 14000rpm 650mm/min CL=0.026mm power = 0.87kW torque = 0.59Nm. So to cut a chip 0.026mm wide we need it to deflect <0.0026mm say. So 120N (12kgf) at 3um and 40N at 1um which probably would work. Do same in mild steel

rpm 2640 T=0.59Nm feed 60mm/min CL=0.0114mm so need ~0.0014mm to cut. tool load is still 12kgf so 120N to move 1um which is about what a new VMC is say around 150N/um... so have a play with that and see if it makes sense. Peter

[ardenum2]

120N to move 1um which is about what a new VMC is say around 150N/um

mind blown... N/um... I took it at face value as a stiffness figure without actually understanding what it meant. This is big, last time this happened, I was watching Dan Gelbart's Building Prototypes series where he explained I can use simple functions to determine straightness/parallelism between two surface measurements hence align rails. They teach you maths but don't really teach you how and where to use it in the real world.

many thanks, I will try to digest it...

[peteeng]

Hi - Yes Dan is a legend. Peter

[pippin88]

Bunnings got back to me. $508 AUD including tax for 5 bags of SikaGrout 5530.
I put in an order.

Maybe your local hardware is nicer to you Peter.

Unfortunately we don't really have local hardware stores anymore in Newcastle. Just Bunnings or one or two other stores (that seem to be keen to mainly deal with business / trades)

[peteeng]

Thks Pippin - Yippee - Your on your way! Don't break a bag on the way home Peter

[pippin88]

Another mob finally got back to me - say might be limited availability depending on current jobs (may be made to order? for wind turbines).
They quoted $70/bag (I guess without tax)
Don't think I can be bothered cancelling the Bunnings order

[ardenum2]

mind blown... N/um... I took it at face value as a stiffness figure without actually understanding what it meant. This is big, last time this happened, I was watching Dan Gelbart's Building Prototypes series where he explained I can use simple functions to determine straightness/parallelism between two surface measurements hence align rails. They teach you maths but don't really teach you how and where to use it in the real world.

many thanks, I will try to digest it...

doesn't all this mean that instead of the usual 1000N I just apply a deflection of 1um and I immediately get how stiff it is(N/um)?

[peteeng]

Hi Ard - correct - applying a deflection vs a force is a simple way to establish the machine stiffness in FE work. But check that the answer is the same both ways when you start doing this just in case your solver handles deflections vs force different. Peter

[ardenum2]

can I setup simsolid to display how much newtons needed to for a displacement of 1um? I went over the result options but the only ones that seem to matter are displacement magnitude and stress. It's weird because I'm getting 90-100N/um(applying 100N and getting between 0.0009-0.0012 displacement) which appears to be very high for such a small size. I am merging all solids into one and applying al 6061 t6 though. I do this mainly to rapidly compare between frames and not really get actual values but it would be possible with E80 uhpc.

do you merge objects into one or take your time and properly assign all connections and materials?

[peteeng]

Hi Ard - In SS (and other FE systems) you can apply a force or a displacement to a feature as an input. It can't be swapped at the output stage. Just set the unit in the force/displacement table. By merging objects are you doing this in CAD (a boolean addition?) or in SS? Its very easy to apply materials in SS to parts. SS can also automatically connects all parts together when imported. Then you can work throough the connections and set them as bonded, friction, sliding etc.

My toolchain is CAD - alibre or fusion or rhino. make the assembly - export as step or a Parasolid. Mostly Fusion now.
SS - import and have SS automatically connect all parts together
SS - select all parts of one material and assign to that material, select other parts and assign etc until all materials assigned. Then in the assembly menu you can display by material to check you have correct materials
SS - apply forces or displacements.
SS - if needed work through connections at parts and change to friction sliding or whatever
SS - if bolts are in preload the bolts
SS - change the solver settings to high resolution

Hit GO - Peter I think thats it OPPs forgot you have to apply constraints otherwise the part gets booted to Mars

your rar file seems to be an image not a compressed file...

[ardenum2]

yes I merge solids into one in NX and I load .prt file directly to simsolid, it supports native nx files which is very convenient. Merging is an old habit because meshing in NX is a horror, if it doesn't crash it takes hours even for one solid piece.

It's probably because it's .rar not zip. I use 7zip for compression, I think it's been with me for almost 20 years now.

[peteeng]

Hi Ard - Then bring a NX assembly into SS and let it connect it together. The step file came into fusion clean. Do you want it stressed? Cheers Peter

edit- just checked it came in as a single body... send it as an assembly.

[ardenum2]

Hi Ard - Then bring a NX assembly into SS and let it connect it together. The step file came into fusion clean. Do you want it stressed? Cheers Peter

edit- just checked it came in as a single body... send it as an assembly.

thanks, I'll take you up on that offer when my final design is finalized and properly defined in simsolid, only wanted to show you what I run simulation on. These aren't even assemblies but single bodies that are parametric within one file, much easier and faster to find geometric boundaries and limits to movement than working with assemblies.

I find myself suggesting to people to build their stuff on slabs but I haven't actually made any concepts about it. So this is what I'm doing now on a 900x600mm piece. Since I'm comparing between frame designs I even simplify rails and carriages to their basic geometric forms. Won't give me real values but will tell how frames compare to each other.

M1 below, I was sure it would be stiffer than the other one, the main ram is quite buff... thats 0.0017 vs 0.0012. I think I have to include the spindles too

[ardenum2]

2) Looking at FSwizard pick a 10mm slot mill - cutting force 12kg in aluminium. 14000rpm 650mm/min CL=0.026mm power = 0.87kW torque = 0.59Nm. So to cut a chip 0.026mm wide we need it to deflect <0.0026mm

for a 0.026mm CL deflection must be less than 0.0026mm? where does this come from?

[peteeng]

Hi Ard - I stated that you figure out how stiff it has to be, then make it 10x stiffer. If the machine deflects the width of the chipload it can't cut it slides. So if the chip thickness is 0.026mm we want the machine to not deflect more than 0.0026mm so it can cut the intended chip. Stiffer is even better. The wizard says it requires 12kgf to do the cut so that's 120N so at the tool we have 120N/0.0026mm or 3um say. 120/3=40N/um to achieve this cut. From the image attached we can see that this stiffness will work. I have tried to find the "Machine Design" book with no luck. But it was a very old reference, machines have become much stiffer since this was published. But a 20N/um machine would be typical of the time. Peter


https://www.glue-it.com/tools/millin...ine-stiffness/

[ardenum2]

.

[ardenum2]

Hi Ard - I stated that you figure out how stiff it has to be, then make it 10x stiffer.

ah okay

FSwizard is weird you set one variable then another and that other resets or changes the first one which you have to change again, bizzare.

do you have the name of that book? or is the name Machine Design?

[ardenum2]

huh this is interesting, I doubled Z height (spaced the carriages twice as far in Z) but deflection only went down to 0.0019 from 0.0021.

[Buoyen]

Bearing blocks dont have anywhere near the same stiffness as solid steel. So to get decent results on this analysis you need to model the correct stiffness at that point.

[peteeng]

Hi Ard & Bouyen - For a sensitivity analysis the material stiffness does not matter as long as the models are kept "apples for apples". As the design refines you can make the bearings from aluminium. This is close to what bearings are. But real bearings have different stiffness in different directions and you have to make a variable stiffness car model to accurately capture that deflection. Peter

Hi ard - I've been concentrating on a 4x8" router at the moment. Peter