[Bartos]

Hi,
I've started designing this machine around april and it developed from a small (400x400mm working area) fun project to a full scale 1750x1000 milling machine.
When I started thinking about Mill and designing it, I barely knew what is CNC and used very simple methods for my first design. However due to my innate urge for cost/benefit optimization and thx to many threads on couple of cnc forums I came of with current design.

Now, I'm ready to get it built, but I still have doubts about some details and I'd like to ask you to help me with them.

Design spec

Price ~3100$ +/- unknowns
Steel Construcion:
all made of 100x100x5mm steel tubes (4" x 4" 1/5")
Dimensions 2000x1460x400 mm
Working Area 1750x1000x250mm

Weight: Bed 250 kg; Gantry 70 kg; Gantry with all the equipment on it 150 kg

Motor Drivers: 4x Toshiba TB6600 drivers
Stepper Motors: 4x (2 for X axis) Sanyou Denki Sanmotion 103H7823-1730 2Nm 4A 2,4mH
X,Y,(Z?)
1,5 MOD R&P
25 Linear Supported Rails

2,2 kW chinese Spindle + VFD Hitachi X200/NE-S1/ TOshiba VFnC3

Bed design...
according to Inventor it flexes 100 um when z axis is all the way down and there is 1130 N force applied to spindle collet and 20 Nm torque.
What are the forces to be expected when cutting steel?

Attachment 215618
Attachment 215620
Attachment 215622
Attachment 215624
Attachment 215626
Attachment 215616
Attachment 215628

At first I wanted to have it all welded, stress relieved, milled and ground at one place but it turned out that the closest plant that could do it is 100 km away and it would cost 3,000$ - way to much for my budget. So I decided to get it welded by a friend of mine and stress relieved in nearby city and level the surface for the rails using epoxy resin - total cost 800$ - much better...

For this width, its rather impossible to drive it from only one side and driving it from underneath the table would require additional 30 kg of weight, to have the ball-nut stiffly coupled with gantry.
Using one stepper motor running two ballscrews is not practical as equivalent mass of whole system would be 255 kg to handle by only one stepper motor - acceleration would suffer greatly and the price would be quite high. So I decided on R&P due to the lenght, width of X axis, as well as cost of double ballscrews for X aixs.

Couple questions that I have

1. Mounting ball-screws is quite straight forward but I have some doubts about R&P mounting. Does the angel between line of mounting bolt-pinion meshing with rack and rack axis ('a' on the pic) has to be as close to 0* as possible? I haven't seen any discussion about this and it seems to me that the bigger the angel the less stiffness I get and less efficiency.

Attachment 215630

2. What are the benefits and drawbacks of making a reduction using pinions vs timing belts and if timing belts than which one should I use?
What is the width of the timing belt per N of force on it for a given stretch? I'd like to buy the proper one, not just the big one.
Anyone knows informative article on timing belts?

3. For the Y axis - I wounder whether I should cram a rack between Y rails or place it on top of the gantry. It seams to me that best situation is when force is applied at the height of a lower rail - what do you think ?

4. What backlash do you have on your machines with R&P

5. How does Z axis work when driven from rack and pinion?

6. Does it make sens to buy vector controlled VFD for 350$ instead of u/f controlled for 250$ (I'll use Chinese 2.2 kW spindle)? I'm most interested in low RPM region 1000-3000 RPM. Maybe just take regular VFD and hook up an encoder to the spindle.

7. This stepper motors cost 80$ and can get 1,5 Nm at 1000rpm and around 1 Nm at 2000 rpm. Do you know anything similar or better around that price?
Any practical considerations and loose thoughts from those that have some experience in heavier machines would be greatly appreciated.

As to leveling and assembling machine.
I plan to level the machine by
1. making square, leveled, co-planar epoxy pads on the floor
2. placing bed support on this pads and leveling its tops with epoxy and than upside down and do the same
3. placing bed on the support and leveling rail mounting surface with epoxy then upside down and do the same.

Gantry is a little bit trickier as perfect 90* angel is needed between bottom (bearing mounting surface) and front (rails mounting surface) of the gantry. I was thinking to level front surface of the gantry with epoxty, than attach accelerometers to this surface and make sure that they have only vertical reading (meaning they are properly attached). Flip the gantry on the top surface and adjust it until accelerometers have zero reading and level the upper surface.
How did you do it? Do you have a simpler solution?

Best Regards
Bart

[nlancaster]

That will be a very strong machine. And it looks like you have some good plans about making it come out square. Have you read about Madvacs machine build. He has some great ideas for building steel weldment structure routers.

[Bartos]

Thx, I'm trying my best.

No, and unfortunately the site is down (MadVac CNC - home made 4'x8' cnc precision gantry router ?). Does it open at your computer?

[nlancaster]

Opens fine for me, from the link you provided. Maybe it was only down temporarily.

[Bartos]

Its just my university network that is blocking too much stuff. I now opened it on 3G network and will read it.

[dmalicky]

Nice work and great start, especially that you have the gantry as a big tube and the Z car uses a channel-section integrated with the router mounting plates. Those are the best ways to make those parts stiff.

The built-up gantry tube looks pretty time-consuming to build -- have you considered simply using an A500 steel tube? A 6x6 or 8x8 would be very stiff over 3' (even with bulkheads only at the ends, or a diagonal sheet inside). A500 would also be much straighter than a built-up tube. In general, the higher the ratio of (wall thickness)/(tube height or width), and the shorter the gantry, the less important the interior bulkheads are. I'd guess a 6x6x3/16 with end bulkheads might be enough for 3'.

The stiffness numbers you give (100 um @ 1130 N) are excellent--that's 65k lb/in, which is in the steel-cutting range. Is that with the tool all the way down and a diagonal XY load? I'm guessing that 65k doesn't include deflection of the linear bearings? Those could easily cut the stiffness in half. They are harder to account for, but some approximations can be made, especially if your Inventor FEA can do contact conditions. In compression, SBR blocks are probably about 400,000 lb/in (start 40 seconds in):
Round Rail vs. Square Rail Technology - YouTube
Side loading is probably somewhat less. But their flexy direction is tension/pull-off. SBR only has 1 bearing track near the opening--and only on one side:
http://image.dhgate.com/albu_3585948...pport-unit.jpg
And, the open block opens up under tension (TBR is better).

In your current Z car and gantry car designs, with a tool load in the long-axis direction, either the top or bottom pair of bearings will *both* see tension, making that a weak mode. I'd suggest orienting the rails/bearings in opposing directions, either facing towards or away from each other (e.g., like Jerry Burke did on his Big Bamboo machine). Oriented like that:
- a long axis cutting force creates side loads in the bearings.
- a cross axis cutting force creates compression in 2 bearings, diagonal from each other. The other 2 are in tension (an advantage of profile rail--they don't care about T vs C), but at least there are 2 strategically in compression to resist the torque.
That works for the Z axis, but rails on top and bottom of the gantry tube will torque the tube walls and probably deflect them. So, also consider profile rail, at least for the 3' axis (but, they need a very flat surface to not lock up).

SBR/TBR is a good choice for the long axis, as long as the blocks are spaced apart enough. In the side view, I'd suggest getting the tool in between the blocks -- then the gantry won't tip under Z tool forces.

For milling steel, cutting forces can range from 50 to 1000 lbs (200 to 4000 N). Of course it depends on RPM, DoC, etc. Here are a few papers that map a range conditions (note their RPMs are much lower than routers usually use, so forces much higher):
http://www.me.mtu.edu/~jwsuther/Publ.../220_PA002.pdf
http://www.iaeng.org/publication/IME...p1751-1756.pdf

1. Yes, ideally that angle is close to 0 deg for stiffness. +/- 10 deg is probably fine.

2. Gear reduction is more complex to design and make, and a little noisy, but much stiffer. I studied belts for a long time before deciding on gears for our 4x8. There is really no way to get a belt to approach the stiffness of gears, but wider, coarser pitch, big pulleys, and more pretension all help some. Coarser pitch hurts precision, so right away we have tradeoffs. GT2 has the best accuracy and repeatability, but not the stiffest cords (except for $$ carbon fiber versions). Some good pdfs:
http://www.maelabs.ucsd.edu/mae156/A...drive/Belt.pdf
http://www.gates.com/facts/documents/Gf000289.pdf
https://www.gates.com/downloads/down...older=brochure
I found that optimizing a belt drive for stiffness requires lots of $ and compromises, and it's still not even in the ballpark of gears. In our gear drive, the most flexible part is the cantilevered pinion that drives the rack (that also took some work to get it stiff enough).

3. Yes exactly, driving the gantry car between the rails and close to the lower rail is the stiffest way. With lateral tool loads, the entire gantry car pivots about the ballnut or pinion. The higher the ballnut/pinion, the more leverage the tool has to rotate the car. It's often tight to package it right next to the lower rail, though. For the 3' axis, I'd use a ballscrew rather than R&P: simpler, more precise, more accurate, longer life. A 20mm lead is plenty fast for a 3' axis, IMO. If using R&P, pointing the teeth down keeps chips from accumulating in them.

4. I've heard 0.001" repeatability is doable with enough care. CNCRP specs their spring loaded system at 0.002". We're still getting the bugs out of our machine (we use a rigid pinion, so rack alignment is critical). If we can't get 0.002", we'll go to spring loading.

5. Ball or ACME screws are normally used for Z, since it doesn't need to be fast, and a rack would fall without power.

6. Vectors drives are much better for low RPM. I don't think a Chinese 2.2 kW spindle will have much torque at 2000 rpm, though.

Other thoughts:
- Lower the long axis rails (If the rails are high, long axis cutting forces cause the gantry to tip).
- Don't weld to the middle of any tube you want to stay straight. Welding is ok at the ends.
- For those points, I'd omit the short riser tubes up to the long rail tubes. The 3' crossmember tubes can be bolted to the underside of the long tubes. The long tubes can be supported with bolted-on diagonals in the lower frame.

[Blaisun]

Good Day, I like your build. you plan is much more detailed than mine was.. One question. is there a reason that your rack has the teeth pointing up vs down? I would think that rack teeth pointing down would be better due to not letting chips and dust accumulate...

[Bartos]

dimalicky
thx for compliments
First,let me give exact dimensions for the table, gantry and working area as I didn't do it in the first post

HxWxL
working area is as in the topic
6' x 3.2' x 1'

and table dimensions are
6.6' x 4.8' x 1.5' (without the gantry)

gantry
11"x5.9"x4,8' (159") thickness of the wall is 1/4"

I started from square tube and I ended up here with max deflection 87um. Plain rectangular tube with the same dimensions gives 147um.
The bulk heads are crucial here - they reduce deflection by a factor of 2 and do not add much weight.
I fiddled a little with the design and it might be better to return to 10" x 9" or 11" x 9" square tube with no flanges. I guess I went to far with complicating things just to gain 2" (or more precisely not to loose them due to distance between end-mill and beginning of the table). I've added flanges as there where no square tube of the required dimensions so I thought that I'll have it welded out of 1/4" steel plates. But they were there mostly to make protection from debris easier. I guess i can use Plexiglas for flanges - much lighter, and transparency gives possibility to check it at anytime.

To be precise with deflections and loads
Load is 800N in X axis and 800N in Y axis applied to the collet. Total of 1132N at 45* angel
Deflection measured at the collet is 59um, at the tip of end-mill 80um.

I'm aware of the Y axis (I didn't think about Z and their simultaneous "opening" with Y though) bearing alignment and their "opening" under tension, although I didn't give it much thoughts. I don't know how much will they "open" and move out of the axis so I can't make any decision.
What I was thinking is
1. Yes they will open and displace. By how much?
2. I could place them on top and bottom of the gantry and they wont open
3. Bearing blocks wouldn't open but whole supported rail is more flexible in lateral direction.
4. I will certainly loos 2" on depth of working area and thus I need to rise it 2" inches loosing a bit of rigidity.
5. No data no decision -> leave it as is.

Once I decide for Y it won't be easy to change it latter but i can definitely change the Z cart at any time.
It is' also possible that in future (if this machine starts making money) that I'll change supported rails for HIWIN linear rails and that will solve all the problems

X axis
I cant really place the tool between blocks. Distance between front face of the gantry and tool in long axis is 7 inches. If I put the block that far ahead, distance between blocks would be 14" - that's quite a lot of unused travel.
Fortunately, center of gravity is almost perfectly between the blocks. If I find it to be a problem I can always change mounting of bearing blocks

I was looking for a paper like http://www.me.mtu.edu/~jwsuther/Publ.../220_PA002.pdf for a long time - thanks a lot.

1. So I thought right - I'll make some models of it soon and update the pictures.

2. How about backlash? I thought that if rack+pinion has a backlash problem then gear+gear would have more problem.
Thanks again for belts related PDFs - I don't have enough time now but I'll surely go trough them in near future.

3. I'll recalculate prices again and see how much of a difference in final price it will make.

6. I think that (please correct me if I'm wrong) U/F controlled VFD will behave in the following way.
For low given RPM, let's say 1500 rev/min it will rotate at almost 1500 rev/min with high current but because there needs to be a proper slip to achieve high torque it will have to go down about 300 rev/min under load to achieve get this slip and achieve proper torque.
This would than mean that closing a feedback loop, with encoder on the spindle, would allow VFD to set frequency high enough to maintain RPM under the load. Is that correct?

Lower the long axis rails... - I don't get it
Don't weld ... even if i get it stress relieved? I need to get those bulkheads inside, somehow.
for the last point I don't get the last sentence. Also bolting might be more troublesome and expensive for me then welding. Friend of mine, will weld it all together. However I'll make suggested change for crossbeams and analyze it with FEA.

Thx a lot for all the comments are PDFs

Blaisun
Thx, I had almost 8 months to get my plan to a decent shape
You're absolutely right, I'll change together with X axis stepper motor mounting plate and post some new pictures soon.

Best Regards
Bart

[dmalicky]

Great to hear it's helpful. Thanks for clarifying dimensions--I obviously missed those earlier. Now I'm getting why you need such a large gantry tube: 12" is a lot of Z travel/clearance/lever arm, and that combined with the high X rails require a longer gantry tube than is typical (and as you know bending deflection goes by L^3).

I certainly agree bulkheads help stiffness a lot. In wood, they're easy. In steel, I look for the easiest and least $ way to get stiffness. One option is to upsize the A500 tube (8x8, 10x10, 12x12) using only end-bulkheads; if that were stiff and light enough, great. If not, alternatives to welding include riveting/bolting in a diagonal sheet, or bolting-in some half-bulkheads that only contact the front and bottom faces. A ~1/8" diagonal sheet stabilizes the cross-section extremely well, and there are ways to install it for a tight fit. What A500 can you get locally? Or Metals Depot and Discount Steel can ship big A500 for much less than the $800.

Those are excellent stiffness numbers for the structure, esp with a 12" Z. I'd suggest including realistic SBR/TBR bearing compliance, as deflection will go way up in reality, undercutting all that work to make a stiff structure. I've done this by modeling the bearing blocks with a custom tuned E (or modifiying the geometry) so they flex the right amount under load. E.g., using the 400k lb/in stiffness from youtube, a 400 lb load should deflect each block 0.001". So I mod the E or geom until that's true.

3. Yes, definitely important to look at. I'd suggest some mini-models to compare 1) a lateral load on a block/rail vs 2) an experiment or estimate of the pull-off stiffness--see below.
4. This will probably be small compared to bearing losses.
5. Yes, often there isn't enough data for a decision; in these cases we can make a reasonable engineering judgement. Clearly, the 1 ball-track off to the side will not be very stiff for pull-off. And we do have some data. First, the 400k lb/in compression stiffness. For pull-off stiffness, we could guestimate (1/3 of the 400k?), or infer it using the load capacity charts (a measure of ball loading and thus correlated to deflection):
http://www.designworldonline.com/upl...oad-rating.jpg
At 90 deg, we see 1.0 capacity which corresponds to 400k lb/in. At 270 deg (pull-off), the capacity is around 0.4. So we could estimate pull-off stiffness as 400*0.4 = 160k lb/in. Now, those numbers are for high quality Thomson bearings with symmetric ball tracks near the opening. And that estimate is just for the bearing shell; the SBR block will open up, reducing stiffness further (thus, TBR). So I'm guessing an SBR shell+block could be around 100k lb/in.
If your Inventor can do contact, you can model the two stiffnesses independently. If not, I might assume ~150k lb/in for both directions (pull-off on the lower rail dominates tool deflection).

Earlier this week I made an xls to calc tool stiffness based on bearing stiffness; see the end of this post: http://www.cnczone.com/forums/diy_cn...ml#post1404620
Plugging in AB=10", BC=12", and k=300k (2 bearings/rail), the math shows Kc=48k lb/in. That's just for the Y bearings; Kc for Z would be similar. Kc for X will be much better since it's not cantilevered; I'll ignore here. Your structure's Kc is 65k lb/in, so inverse addition of two 48k springs and one 65k spring yields an aggregate stiffness of 18k lb/in. Yikes. So this is why I'd suggest either:
A. Do an FEA to find the lateral stiffness of a supported rail; if better than pull-off, orient them opposite each other. And use TBR rather than SBR.
B. Cut costs elsewhere to get Hiwin on Y and Z from the start. An LG20C Z1 is 1800k lb/in in all directions! 2 of them are 3600k, for a Kc of 570k lb/in. On our 4x8, LG20C on the Y and Z axes were only $560 total.

For the X axis feet, 14" sounds about right. Our 4x8 uses 16" feet on skate bearings. Check out some ~heavy duty routers -- their foot length is similar. And they have profile rail for X, so an SBR foot should be even longer.
http://www.technocnc.com/CNC-Routers...outer-side.jpg
http://www.cncroutercentral.com/CNC-...c-Router-a.jpg
And the more Z clearance, the longer the feet: http://www.multicam.com/images/router/8000-r.jpg
One general mechanical principle here is if you want to control something (like a tool hanging down 12"), you need a similar or greater lever arm to control it well. The specific principal is to keep the foot bearing blocks always in compression. Tool loads (especially Z) could cause tension if there is overhang. SBR/TBR is a very reasonable cost save for the X (vs profile rail), if kept in compression.

>2. Backlash:
Yes, gear+gear will add some, but it's minimal for 2 reasons:
- Those shaft positions can be accurately controlled, much better than the pinion-rack.
- The angular backlash at the output shaft is small, since it's connected to a big gear (~4:1 ratio). I forgot what it calc'd to on ours; pretty sure it was trivial.

>6. VFD:
Peak torque is around 90% slip: http://lhp.co.in/images/c_motor_drv_11.jpg
But notice the current is massive there and even more at higher slip. The motor will overheat after a few seconds.
I'd probably get coated carbide bits and run them faster. E.g. this post and above (impressive as this machine is only ~3k lb/in): http://www.cnczone.com/forums/cnc_wo...ml#post1237624

>Lower the long axis rails... - I don't get it
The main challenge with the 'feet' bearings is to prevent tip from two forces (side view FBD):
1) Acceleration loads acting at the CG of the gantry assembly -- so, about the middle of the tube.
2) Longitudinal cutting forces acting at the tip of the cutter -- so, as low as table height.
In the side view, each of those loads has a lever arm that tries to tip the gantry; the bigger the lever, the higher the foot bearing loads, and the more tip. The ideal X rail height is in-between the gantry CG and the table. The exact height depends on which factor is more important to you. Since this is a steel cutting machine and speed is less important, I'd favor the X rails being much closer to 2) than 1).


> Don't weld ... even if i get it stress relieved?
> for the last point I don't get the last sentence.

I like to design so a stress-relief isn't needed: it adds outsourcing cost that could be better spent on profile bearings, and it's difficult to transport large parts. For example:
For the gantry: A500 plus a bolted/riveted diagonal sheet, or bolted-in half-bulkeads.
For the X rail tubes: Lower the X rails so they are near the table. Then those tubes deflect minimally, so bolting is fine. With the X rails lowered, only 1 longitudinal tube is needed per side. Welding is great for any triangulated structure below, and is no problem at the ends of the X rail tubes.

Yes, hard to explain; here's a near-final iteration of our 4x8 machine as an example, designed for max stiffness at least cost.
Attachment 216472
The gantry is alum 8x8x1/4 with a bolted and riveted diagonal sheet. The frame tubes are steel: X-rail tubes are 2x4x3/16 and all the lower tubes are dainty 2x2 and 1x3 16 gauge. Pretty light, but triangulation is *extremely* stiff and so also insensitive to welding strain-relief over time. Sand is a cheaper way to add mass if it needs it. Two of the 1x3s (#1 and #8) are welded to the 2x4s; the others are bolted. The diagonals supporting the middle of the 2x4s are welded to each other, but bolted to the 2x4 via a plate. We did get a high spot of about 0.015" there; a hand grinder took care of it. Otherwise, the 2x4s stayed flat within about 0.010" and we shimmed the rails to within 0.004". IIRC, FEA showed K at the tool was between 50k and 80k lb/in, depending on contact conditions (sliding bearings), constraints, and load direction; haven't tested it yet for real but am guessing it'll be around 40k.

[Bartos]

David,
thank you for your reply. I read it all, now I need some time to analyze it all and to put it all in the software. Unfortunately I'll be quite busy for a week so I'll post a reply in a week or two.

Best regards
Bart

[dmalicky]

An update, here are some stiffness plots for SBR20:
http://www.cnczone.com/forums/linear...ml#post1412778

Note, in my prior posts in this thread, I see I mixed up the compression stiffness of two blocks (400k) with one block (200k). The experiments show a single SBR20 block is also about 200k lb/in in compression. In tension, a wide range but it averages about 60k.

[Bartos]

Hi again,
thx for your advice - now it can be assumed that result are quite precise.

I did couple things to gain stiffness, and here is a new drawing.


First of all I changed all rails to HIWIN (thx to you so I will skip all the points related supported rails.
Picture is quite self explanatory but I'll write briefly what I did.
I switched the table upside down so that sides do not contribute to later deflection.
Thx to that I could reduce table width (while maintaining working space) by 10" and thus reduce block spacing (10") and gantry width(4").
"gantry with legs" type of design allows for wider support on the sides and much lower deflections. Even though it seams much bulkier weight of the gantry is almost the same - 162 lbs; I've reduced wall thickness from 6/24" to 5/24"

By changing design I could reduce deflection on the collet from 68 um to 40 um (without block stifness). After adding block stiffness into design (800k lbs/inch as on youtube video) deflection is now 86 um / 0.0034" (when force is acting from the back of the gantry) and 100 um / 0.0039" (when force is acting from the from of the gantry). If I'm not mistaken its around 60k lbs/inch when a spindle is all the way down. Around twice that much when the spindle is 2" from the top. That I guess gives possibility to do some precise machining on small objects mounted in a vice.

Deflections are on the pic
Attachment 218336

I've looked up couple stores around me and I've finally found one that offers HIWIN blocks and rails reasonable prices so I decided to increase the budget and use HIWIN HGH20CAZAH blcks and HGR20RxxxxH rail on whole machine. Partially because I took a look at price difference between supported rails and linear rails in a perspective of a cost of whole machine and not rails themselves. Also I've noticed at the stores that good open blocks cost the same as HIWIN blocks.

I certainly agree bulkheads help stiffness a lot. In wood, they're easy. In steel, I look for the easiest and least $ way to get stiffness. One option is to upsize the A500 tube (8x8, 10x10, 12x12) using only end-bulkheads; if that were stiff and light enough, great. If not, alternatives to welding include riveting/bolting in a diagonal sheet, or bolting-in some half-bulkheads that only contact the front and bottom faces. A ~1/8" diagonal sheet stabilizes the cross-section extremely well, and there are ways to install it for a tight fit. What A500 can you get locally? Or Metals Depot and Discount Steel can ship big A500 for much less than the $800.

I think we talk about quite different numbers when it comes to costs, so let me clarify.
Steel 0.45 $/lbs
Stress relieving 0.15 $/lbs
Welding 0.33 $/inch
I think that this prices look quite different in USA.

I can get 8x8 easily and 10x10 is something that I'd have to look up in couple places. I think that, given complexity of gantry design (bulk heads), it might be easier to have it welded from steel plates. My friend can cut the steel to proper dimensions so cost won't increase to much.

If your Inventor can do contact, you can model the two stiffnesses independently. If not, I might assume ~150k lb/in for both directions (pull-off on the lower rail dominates tool deflection).

Yes, I've found Contact option that allow to add stiffness between objects and that's how I got new numbers.

B. Cut costs elsewhere to get Hiwin on Y and Z from the start. An LG20C Z1 is 1800k lb/in in all directions! 2 of them are 3600k, for a Kc of 570k lb/in. On our 4x8, LG20C on the Y and Z axes were only $560 total.

What are the LG rails ?. I can't find them in HIWIN catalog. DO you have any numbers regarding HGH20 blocks?

For the X axis feet, 14" sounds about right. Our 4x8 uses 16" feet on skate bearings. Check out some ~heavy duty routers -- their foot length is

Right now there is 10" from front face of fron block to rear face of rear block. I'll play with the design and see if I can get better stiffness.

As for rest of the post I'll get back to it when I have more energy. I was working in Inventor for 10 hours today and couple more yesterday, I've got no more power.

Best regards and thx again
Bart

[dmalicky]

Nice updates -- it looks stout and those are great stiffness numbers. Those V-shaped gantry legs help lateral stiffness a lot, and to 'shorten' the tube. All of which you know. That's terrific you can model with contact. Does it also let all the linear bearings slide on their rails? I like that to show any stiffness losses from 'racking', and also the best ballscrew placements.

For the FEA model, I like to include a tool of realistic length, but large diameter and very high modulus. That way, the displacement due to cosine effects of an slightly angled spindle are accounted for, but we don't also have excess deflections from a flexy tool.

I was thinking you're in the US which is why I mentioned Hiwin (here, automationoverstock.com has old stock Hiwin for good prices). But Hiwin is one of the lower priced options for profile rail elsewhere, too. Those are very attractive cost rates -- where are you located? I'm in southern California, and here I doubt one could stress relieve a spoon for $24 (0.15 * 162 lb)!

The LG blocks are old stock and no longer made. I see Hiwin's current catalogs don't have the stiffness charts like the old ones did. You could try asking Hiwin, or estimate the stiffness by comparing an old LG block at the same preload class and load rating. Section 2-1-7 of the old catalog
http://www.automation-overstock.com/...arguideway.pdf
has the stiffness numbers, and the load ratings start 8 pages later. HGH is said to have improved stiffness, so the LG stiffnesses should be conservative. My off-the-cuff guess is that an HGH20 Z1 would probably be 40 kg/micron or more. So for 2 bearings, that's probably a stiffness of ~4 million lbs/in in the xls -- fantastic!

If welding the gantry, keep in mind it will distort quite a bit, so include some thick 'pads' for the rail mounts -- those will be machined off to get level surfaces. Distortion can be minimized by extensive and rigid fixturing, tight joint fit-up, lots of tack welds, and checking as you go and trying to correct distortion by a strategic weld order. Also, most welds don't need to be full length; stitches are fine for many, and the less heat, the less it will distort. Welding does have the advantage that you can tailor the wall thickness and tube sizes to be ideal for your design. E.g., it's often helpful to decrease the wall thickness and increase the tube dimensions (especially X, since the Z is limited by realistic rail placement) for a better stiffness/weight ratio. Lower weight is helpful so the motor and drive costs don't get too excessive, and/or so the motors/drives can accelerate the gantry at a fast rate.

If using an A500 tube, we're discussing ways to install a diagonal or simple bolt-in bulkheads currently in another thread, starting around here: http://www.cnczone.com/forums/diy_cn...ml#post1413940

On leveling with epoxy, I've learned recently this may not be a good idea for a really stiff machine. We did it on ours, but I also found the epoxy seemed to be a soft foundation for the rails. Later in that same thread, Wizard says that "The unfortunate reality is that epoxy stiff enough to support the rails is usually to stiff to flow." So for a metal-cutting machine, it appears that epoxy probably isn't a good solution for rail mounts. That leaves machining them (the prefered solution, because then a ledge can be made for rail alignment) or building something that is level by design/process (I'm not sure that would be precise enough for a machine like yours).

For analyzing foot length, I like these tests:
- Model an accel/decel load at the CG of the gantry--that will check tipping tendencies. The 'g' level is harder to estimate, but I think 0.5g is a pretty aggressive number. Many machines run 0.1g or less, because the gantry mass is large or the motors are small.
- Give the tool a pure Z load, like for drilling. That can be a worst case for tipping if the tool overhangs the foot bearings.

Yes, CAD and FEA are great time sinks!

[dmalicky]

Another thought on welding the lower frame: there is quite a bit of welding to the inside and underside of the long X tubes (under the rails), and no welding on the opposite sides (top and outside). That will likely cause the tube to bow up and out--as those welds cool, they contract the bottom and inside of the tube, forcing those bows. If the tube has enough of its own stiffness compared to the effect of the welds, the effect is smaller, but it's still a risk. Stress relieving should correct a good amount of the distortion, but not all of it. A common solution is to also weld long flat bar stock to the top of the tube, which would be be machined away when making level rail mounts. The welding to the top of the tube will also help correct the upward bow, but could also overcorrect it, so hard to predict where it will 'land'. (Those risks and extra work are why I only weld to the ends of that tube, using bolts in the middle, but many others weld and then level with either machining or epoxy.)

[Bartos]

Nice updates -- it looks stout and those are great stiffness numbers. Those V-shaped gantry legs help lateral stiffness a lot, and to 'shorten' the tube. All of which you know. That's terrific you can model with contact. Does it also let all the linear bearings slide on their rails? I like that to show any stiffness losses from 'racking', and also the best ballscrew placements.

Yes, it does look better and I like it much better this way .

here is one more drawing with dimensions

Attachment 218536

Inventor has following contacts:
Bonded
Separation
Sliding / No Separation
Separation / No Sliding
Shrink Fit / Sliding
Shrink Fit / No Sliding
Spring
but I don't know if it's possible to combine Sliding / No Separation with Spring.

I was thinking you're in the US which is why I mentioned Hiwin (here, automationoverstock.com has old stock Hiwin for good prices). But Hiwin is one of the lower priced options for profile rail elsewhere, too. Those are very attractive cost rates -- where are you located? I'm in southern California, and here I doubt one could stress relieve a spoon for $24 (0.15 * 162 lb)!

I've found HIWIN HGH20CAZAH blocks for 45$ with tax so it's quite cheap compering to USA and even to automationoverstock.com.
I've seen prices for BOSCH rexroth and NSK and this is completely different price level and probably precision level. I think that HIWIN is more than enough for me at this time.
I'm in Poland and when I saw that you're form USA, I figured out that you think about completely different rates for welding etc.

The LG blocks are old stock and no longer made. I see Hiwin's current catalogs don't have the stiffness charts like the old ones did. You could try asking Hiwin, or estimate the stiffness by comparing an old LG block at the same preload class and load rating. Section 2-1-7 of the old catalog
http://www.automation-overstock.com/...arguideway.pdf
has the stiffness numbers, and the load ratings start 8 pages later. HGH is said to have improved stiffness, so the LG stiffnesses should be conservative. My off-the-cuff guess is that an HGH20 Z1 would probably be 40 kg/micron or more. So for 2 bearings, that's probably a stiffness of ~4 million lbs/in in the xls -- fantastic!

I had done as you said, I compared preload and took 40 kg /um as a good approximation and as you say - results are fantastic and 4 mln lbs/in is quite a lot.
With new stiffness I got the following results (force 800 N in X + 800 N in Y, applied to collet and measured at collet)
65um / 0,0026" when applied from front
55 um / 0,0022" when applied from the back
33 um / 0,0013" when applied from the bottom - answer to last part of your post

For the FEA model, I like to include a tool of realistic length, but large diameter and very high modulus. That way, the displacement due to cosine effects of an slightly angled spindle are accounted for, but we don't also have excess deflections from a flexy tool.

So I've made the tool 2" long and 4/5 wide and applied a force (same as always - 800 N in Y and 800 N in X) to bottom 1/2" of the tool and deflection changed quite a lot.
Now I have ( collet/tip of tool )
force acting from the front 90 um / 202 um - 0,00354" / 0,0080"
force acting from the back 80 um / 190 um

If welding the gantry, keep in mind it will distort quite a bit, so include some thick 'pads' for the rail mounts -- those will be machined off to get level surfaces. Distortion can be minimized by extensive and rigid fixturing, tight joint fit-up, lots of tack welds, and checking as you go and trying to correct distortion by a strategic weld order. Also, most welds don't need to be full length; stitches are fine for many, and the less heat, the less it will distort. Welding does have the advantage that you can tailor the wall thickness and tube sizes to be ideal for your design. E.g., it's often helpful to decrease the wall thickness and increase the tube dimensions (especially X, since the Z is limited by realistic rail placement) for a better stiffness/weight ratio. Lower weight is helpful so the motor and drive costs don't get too excessive, and/or so the motors/drives can accelerate the gantry at a fast rate.

It's very unlikely that I would machine rail mounting surfaces, for two reasons.
1. I've found only 1 company in 60 miles radius from my place that could do it. They charge 3000 USD for material, welding, stress relieving and machining.
I can get it welded, stress relived and epoxy leveled for 800$. 2200$ is more than half of the price of whole project.

Most of the companies don't have machines that could do it and those that have, say that this design is too small for them, their working field is 10x10 yards or so and gantry can lift up to 200k lbs and they say that they would have to charge me so much that they won't even offer it.

If using an A500 tube, we're discussing ways to install a diagonal or simple bolt-in bulkheads currently in another thread, starting around here: http://www.cnczone.com/forums/diy_cn...ml#post1413940

Thx, I will read it trough.

On leveling with epoxy, I've learned recently this may not be a good idea for a really stiff machine. We did it on ours, but I also found the epoxy seemed to be a soft foundation for the rails. Later in that same thread, Wizard says that "The unfortunate reality is that epoxy stiff enough to support the rails is usually to stiff to flow." So for a metal-cutting machine, it appears that epoxy probably isn't a good solution for rail mounts. That leaves machining them (the prefered solution, because then a ledge can be made for rail alignment) or building something that is level by design/process (I'm not sure that would be precise enough for a machine like yours).

I thought about this and I've figured out a cheap solution to this problem if it occurs - and as you say it probably will.

First - to maximize stiffness, I will "level" mounting surface by placing steel balls of different radii on the surface and then pour epoxy on the balls to cover them with around 0,2" of resin.

If that's not enough I will have a 0,5" or 1" thick steel plates precisely ground (to +/- 0,0008") and I'll place them over the epoxy resin. I hope that it'll work and it will be definitely much cheaper than machining whole construction. I guess I can have them ground for under 100$ + material (30$).

For analyzing foot length, I like these tests:
- Model an accel/decel load at the CG of the gantry--that will check tipping tendencies. The 'g' level is harder to estimate, but I think 0.5g is a pretty aggressive number. Many machines run 0.1g or less, because the gantry mass is large or the motors are small.

I have calculated max acceleration with a load safety factor for a stepper motor of 1/2. I took under consideration ballscrew inertia, motor inertia, and bearing drug. Max acc for X Y Z is 0,30g 0,44g and 0,8g respectively. This numbers should probably be divided by 3 to allow for acceleration during cuttings.

Deflection under 800 N acting on a CG in X direction was only 15 um / 0,0005".

Yes, CAD and FEA are great time sinks!

Yes and I'm thinking to leave the design as is and start making it instead of perfecting it. It looks quite good now. I'll only play with bottom of it and see how much weight I can reduce before it starts flexing under the gantry.

Thank you
Bart

[Bartos]

Small update...

I changed a design of the bed, a little bit. I removed lower rails since they did not contribute to stiffness and reduced wall thickness from 5 to 4 mm (form ~1/5" to ~1/6").
Now everything is much lighter and as stiff as it was. I reduced bed weight from 220 kg ~550 lbs to 120 kg ~250 lbs.
Cross-members deflect in Z axis around 19 um / ~0,001" under a 70 kg / ~150 lbs steel block and this deflection doesn't change much under horizontal force (1120 N ). I guess that's a good result.
Attachment 218676



Now, since the bed is so simple - I wonder whether I should have it welded (weld it myself using stick welder?) or screw it together. If welded - do I need to stress relieve such a simple construction or just let it lay outside for 2 weeks or so?

Anyone has some thoughts?

Best regards
Bart

[dmalicky]

We see some very nice metalworking products from Poland here -- Bison chucks, and we used to be able to get master precision levels from there, but I've not seen them in a few years. I'm guessing machine tools are a focus of your country?

Those stiffnesses at the collet look great.
- For X and Y at the tool, I'm guessing your tool was bending a lot, despite the 20mm size. You could try a higher modulus to made it rigid.
- Good to see the Z stiffness is high -- HGH20 on the X axis helped that a lot.
- Another way to check stiffness is to apply equal and opposite loads to the tool and workpiece, since that's what happens during actual cutting. Then monitor net displacement between the two. Sometimes that shows a weak mode in the table (but probably not with your rails at table height).
- Also, what constraints are you using for the FEA? You probably know the problems of over or artificial constraints.

For Inventor's contact options, if it can't do Sliding + No Separation with Spring (which is a common type of limitation for FEA packages), there is a fairly easy workaround (superposition) since this is a reasonably linear problem:
- Run a standard case with no contacts and find the *standard deflection*
- Run the Spring case and calc the *extra deflection due to springs*
- Run the Sliding case and calc the *extra deflection due to sliding*
Then the total real deflection will be = *standard deflection* + *extra deflection due to springs* + *extra deflection due to sliding*
In my models, sliding has added (very roughly) 20% to the total. A low Y ballscrew helps minimize that.

Yes, I hear you on the machining costs.
- I've not seen the ball-bearing idea before. It's intriguing, although the process sounds more difficult to me (measuring each surface height, getting the right ball for the right place, not mixing them up, placing enough of them to support the rail...), but you may have something in mind.
- I like the ground steel bar idea; I think 8mm would be thick enough, and thicker is probably not be better since lateral loads on the HGH cause a torque on the epoxy.
- There is also the traditional leveling method--spotting and scraping. Weld on sacrificial hot-rolled steel bars, roughly 30mmx15mm. Use a precision straight edge and/or surface plate coated with ink to mark the high spots. Scrape, file, grind, or sand. Repeat. That method is good for as much flatness accuracy as you have time (but it will take a lot). Rail mounts are easier than ways since they don't need to be very smooth, just flat, and small gouges can be ok. There are some good videos on youtube, and posts...
Want to get into Hand scraping - Page 2
http://www.cnczone.com/forums/genera...education.html
Some use a hand grinder to speed up the process, at least for roughing--not as precise but they make it work. gingery_machines @Wiki - scraping

Your acceleration analysis and results look good.

Are the holes in the gantry tube for access to the rail bolts? If not, I'd check whether the weight savings would be more effectively achieved with thinner walls than holes.

Most of us use a taller lower frame so the table is at a more ergonomic height. But with the table off the floor, accel/decel loads can cause a lot of shaking. Diagonal braces fix that and they can be very thin wall.

If the bed is welded, stress relieving is the only way to be sure it will be stable. Room temp creep relief is really slow--years and decades. As you know, I bolt the mid-tube joints to avoid all that.

I'm impressed with your work--it's rare I see an undergraduate working at this level. On desiging vs building, I almost always find it's better to spend a lot of time on the design. CAD/FEA are slow, but far faster and less expensive than redoing a build that isn't right. I've learned to revisit designs over a period of many months--simplify, improve performance, try a new approach, reduce cost, incorporate something new I learned. But during that period, it's also good to build 'critical function prototypes' -- if there's some question that can't be answered by analysis, or if I just need to get a 'gut feel' for something, or test whether an idea will work, then I build a mini/partial prototype (maybe in wood) as an experiment to answer those questions. E.g., maybe weld up some scrap steel like a mini gantry, and measure distortion. Or practice leveling techniques. Then that info can get reincorporated into the design up-front, instead of having to apply band-aids later.

[Bartos]

We see some very nice metalworking products from Poland here -- Bison chucks, and we used to be able to get master precision levels from there, but I've not seen them in a few years. I'm guessing machine tools are a focus of your country?

Thank you, that's nice to hear that.
I don't think that tools are a focus but heavy industry or steel related industry might have been until 1989'.

They are still on the market BISON-BIAL S.A. and they should be available in US.

Those stiffnesses at the collet look great.
- For X and Y at the tool, I'm guessing your tool was bending a lot, despite the 20mm size. You could try a higher modulus to made it rigid.
- Good to see the Z stiffness is high -- HGH20 on the X axis helped that a lot.
- Another way to check stiffness is to apply equal and opposite loads to the tool and workpiece, since that's what happens during actual cutting. Then monitor net displacement between the two. Sometimes that shows a weak mode in the table (but probably not with your rails at table height).
- Also, what constraints are you using for the FEA? You probably know the problems of over or artificial constraints.

For now I will not do any simulations as I need to focus on some projects at university for which a deadline is for next week. But when I finish them I'll get back to this project.
But I guess (from previous simulations) that increasing Y rails spacing by increasing height of gantry front face I'll gain more stiffness then by anything else.
I'm not sure what you mean by artificial constraints as I've never taken any classes in FEA etc. but I guess, simply speaking, it means not making it as it is in reality.

For Inventor's contact options, if it can't do Sliding + No Separation with Spring (which is a common type of limitation for FEA packages), there is a fairly easy workaround (superposition) since this is a reasonably linear problem:
...

Yes that's indeed very simple way around. I will check this when i get back to this project.

Yes, I hear you on the machining costs.
- I've not seen the ball-bearing idea before. It's intriguing, although the process sounds more difficult to me (measuring each surface height, getting the right ball for the right place, not mixing them up, placing enough of them to support the rail...), but you may have something in mind.
- I like the ground steel bar idea; I think 8mm would be thick enough, and thicker is probably not be better since lateral loads on the HGH cause a torque on the epoxy.

As to bearing balls I think it should be quite simple to do.
Let's assume that ends of my frame bend upwards after welding. So now looking from the side, we have a banana shape \_____/ like that.

1. Put a straight edge on top of that and find point of biggest depth - let's mark this point as point A and depth as D1 this will also be diameter D1 of our biggest balls.
2. Now drop couple "D1" balls and they will all roll down to the lowest point - lest call it B. Now rock/adjust the bed until point B is in point A.
3. Drop more balls of D1 radius until they start to stick out above the straight edge by the difference in diameter between D1 and next smaller balls "D2".
4. Drop D2 balls on the left and right side of D1 balls and repeat process until the smallest balls

For one gantry side it should be absolutely trivial, for 2 sides we will not have the same banana shapes. Points A on both sides will be shifted and thus we will have to adjust level of the bed so that total distance between points A and points B is at minimum or area between level and steel frame is at minimum thus minimizing resin volume and maximizing stiffness.
For both shapes being convex it should be even easier.

- There is also the traditional leveling method--spotting and scraping. Weld on sacrificial hot-rolled steel bars, roughly 30mmx15mm. Use a precision straight edge and/or surface plate coated with ink to mark the high spots. Scrape, file, grind, or sand. Repeat. That method is good for as much flatness .....

I've briefly read about this some time ago but I'm not sure if I can get an equipment for this here and if I want to do this.
But I might weld on some steel plates on the bed just in case of epoxy leveling failing to work and screwing steel plates on epoxy fails to work too.

Your acceleration analysis and results look good.

Thx, I played a bit with numbers, two days ago, and I added some more cells to my excel calculator. Among others "max cutting force @ full acceleration" and "max cutting force @ reduced acceleration". Now for stepper motor that I'v selected (Sanyo Denki 103H7823-1730) I can get following forces during acceleration at speed up to 2 inches per sec - utilizing only 60% of motors torque.

available force @ acc

X 1050 N @ 0,2 g
Y 993 N @ 0,13 g
Z 993 N @ 0,2 g

By the way, do you know this store PLC Center MRO Inventory and Industrial Repair Service - PLCCenter.com, are they trustworthy? It's the only store where I can find this motors and they seem to be the best motors price-wise that I've looked at. And I consider only those that have full spec and torque curve. I'll be in Chicago soon and I was thinking to buy the motors from them in advance and bring them with me

Are the holes in the gantry tube for access to the rail bolts? If not, I'd check whether the weight savings would be more effectively achieved with thinner walls than holes.

I played a lot with FEA with all of the futures of the design (wall thickness, holes diameter etc) and for each I've calculated um/kg gain and adjusted the features until I got the best ratios. But this was before I changed the design and gantry. The numbers might have changed now and I should probably readjust them.
Holes on the top are because they did not reduce stiffness and on the back to help mounting rails; to give a welder access to the inside ribs and of course to reduce weight.

Most of us use a taller lower frame so the table is at a more ergonomic height. But with the table off the floor, accel/decel loads can cause a lot of shaking. Diagonal braces fix that and they can be very thin wall.

I have something in mind. I will post it as soon as I draw it.

If the bed is welded, stress relieving is the only way to be sure it will be stable. Room temp creep relief is really slow--years and decades. As you know, I bolt the mid-tube joints to avoid all that.

I guess that I will screw it together using big screws. Stress relieving would probably be very impractical due to the size of lower frame.

I'm impressed with your work--it's rare I see an undergraduate working at this level. On desiging vs building, I almost always find it's better to spend a lot of time on the design. CAD/FEA are slow, but far faster and less expensive than redoing a build that isn't right. I've learned to revisit designs over a period of many months--simplify, improve performance, try a new approach, reduce cost, incorporate something new I learned. But during that period, it's also good to build 'critical function prototypes' -- if there's some question that can't be answered by analysis, or if I just need to get a 'gut feel' for something, or test whether an idea will work, then I build a mini/partial prototype (maybe in wood) as an experiment to answer those questions. E.g., maybe weld up some scrap steel like a mini gantry, and measure distortion. Or practice leveling techniques. Then that info can get reincorporated into the design up-front, instead of having to apply band-aids later.

Thank you, I guess that I should have selected studies related to engineering or machine building
That's a very good approach, I couldn't agree more .


Best regards
Bart

[dmalicky]

Ah, yes the Bison chucks are available here. I just learned the level company (VIS) went out of business... Precision Levels

Right, for FEA constraints, the goal is to simulate reality. Unfortunately nothing in reality is rigid, while FEA constraints usually are. It's also easy to apply excessive constraints, like constraining a surface to be perfectly rigid. As a rule for FEA, stress and deformation results "near" the constraints are usually inaccurate. But in a gantry CNC, we usually don't care about the lower frame, as it doesn't move much and has low stress. Ideally, we would put a contact condition (with friction) on each foot, but that may be beyond Inventor's capability, and would be slow to solve even if it could. But we don't need to work that hard for this problem. Lets say you have 4 feet, numbered and arranged like this (view from above):
3 4
1 2
Where the 1-2 direction is the X axis, and 1-3 the Y axis. Then I'd suggest these constraints, as a reasonable set that will solve quickly:
- Foot #1: Constrain a single point on the foot in X, Y, and Z -- this could be considered the pivot point for the entire model.
- Foot #2: Constrain a single point in Y and Z. This will prevent the model from rotating about the Z and Y axes (respectively), and give expected support from the floor
- Feet #3 and #4: Constrain a single point in Z only. These will prevent rotation about the X axis, and give floor support.
Inventor may give a warning that you have point constraints; that's ok, it just means stresses there will be innacurate. And those constraints put all the X & Y cutter force into feet 1 and 2. That may not matter if the frame is stiff enough. Or, just apply an equal and opposite force (~co-linear with the cutter force) to a block workpiece on the table. Then the feet take nothing except the weight. (To check the effect of gantry accel, apply a m*g force to the approx CG of the gantry... only Foot #1 will react, but if it alone can do it with minimal deflection, 4 real feet will do it better.)

I hadn't thought of that ball bearing plan--thanks for the explanation. I think I'd prefer a stiff plane to mount to, rather than points of stiffness. But your plan is clever.

I've never ordered from PLC Center, but I believe it's a normal business. They have been around for a while.

That's interesting that the holes didn't hurt stiffness--I will try that on my models, thanks. That's great you did a parametric study!

Yes, you would be a good mechanical engineer!