Moving Gantry column bearing spacing - open discussion

[--JawZ--]

No I have tested the IGUS system and it is not suitable for CNC machine use, even at Hobby level, it would be a very bad choice if you where to build a new machine, IGUS can be used for other sliding systems that don't need any kind of accuracy

The design you have with the round rails would not work with supported round rails as drawn the Bearings would have to be open, round rail bearings also have large clearance between the rail and bearing some have small adjustments that can be made but are ok for hobby level machines, if you want any kind of accuracy then you would use pre-loaded profile linear rails

I was talking more in general terms of cnc machines, a metal or even a wood router/mill cnc machine should not use IGUS rails, but a low weight 3D printer or a laser, could and can use the IGUS rails with good results.
I have however, not been able to do any long term tests of this kind of system yet. As I've recently just started testing the IGUS rails out on a custom designed 3d printer made up out of alu, carbon fiber and both IGUS rails and roller wheels in v-slotted extrusions. The design also welcomes the HWIN rails, so if my testing shows that I can't use the IGUS system, I will just move to the HIWIN system.

The design was just a mockup of a placement idea I had, and was constructed while I had some time of at work. No real design thought beyond the general placement was considered.
And I agree with you, that pre-loaded profile linear rail system is a better option. It was just to add an option of something someone might think of, and having it in here with both pro's and con's. Which would allow them to decide to choose a similar, or different design. If it would fit most of their requirements, not only in the machine, but also in their pocket.

[mactec54]

I was talking more in general terms of cnc machines, a metal or even a wood router/mill cnc machine should not use IGUS rails, but a low weight 3D printer or a laser, could and can use the IGUS rails with good results.
I have however, not been able to do any long term tests of this kind of system yet. As I've recently just started testing the IGUS rails out on a custom designed 3d printer made up out of alu, carbon fiber and both IGUS rails and roller wheels in v-slotted extrusions. The design also welcomes the HWIN rails, so if my testing shows that I can't use the IGUS system, I will just move to the HIWIN system.

The design was just a mockup of a placement idea I had, and was constructed while I had some time of at work. No real design thought beyond the general placement was considered.
And I agree with you, that pre-loaded profile linear rail system is a better option. It was just to add an option of something someone might think of, and having it in here with both pro's and con's. Which would allow them to decide to choose a similar, or different design. If it would fit most of their requirements, not only in the machine, but also in their pocket.

Almost anything will work for a 3D printer you are not doing anything serous with hobby level 3D printers, you sound like you are reinventing the wheel as all these types of guiding systems have been well tried out by most hobby builders, which move on quite quickly once they see the results, you don't have to use Hiwin there are may much cheaper that will do the same job, you have to choose carefully though with the cheaper rail sets, the steel round rails work well for 3D printers and don't have to use supported rails because of the low loads, so you have lots of choices, The IGUS bushing system they use is not reliable and not easy to get adjusted correctly

[peteeng]

Hi Mactec - So the machines built with twin rails and 4 cars each side did they do better then a single rail and 2 cars each side? regards Peter

[peteeng]

Good morning CNCers - So Jawz sent me a step file of his gantry & gave permission for showing the images. I stripped it down to the structural parts. Its a combo of extrusions and cnc plates. I imported it into simsolid and used an auto connect command which bonded all the parts together. I then applied a "push" load to the spindle collet of 1000N while the cars were restrained. This is my standard design load. The collet moved 0.123mm. This is a static stiffness of 1000N/123um or 8N/um. Most if this deflection is in the collet as you can see in the image. So I made the collet "rigid" to see what the machine does vs just the nose of the spindle. It then moved 0.090mm so the static stiffness is 1000/90= 11N/um. As the model is bonded ie all connections are 100% efficient I expect the real figure to be around half the calculated one.

I then deleted one rail to see what the "free" deflection was and its 2.71mm. Then with the rail in its 1.1mm deflection. The columns are two parallel plates and I would suggest that these be addressed as in the exaggerated images you can see these winding. So the gantry is stiff, the columns I feel are the weak link. My other comments are as Jawz has access to commercial CNC to cut his parts why not build a plate gantry vs using extrusions? It will be locally and globally stiffer. The columns have cover plates on them (not shown here) and these could be made to be structural to become better in torsion. Thanks Jawz Peter

[peteeng]

Modelling Weirdness - Hi All - I decided to put in a 150mm spacing to complete the series and then chart it. But the 150mm deflected less then the 200mm so I played around with the model and couldn't find out why. I decided it may have been the complex sliding contact with the rails and cars as they include the grooves and stuff. So I built a simplified rail and car using only 3 contacts and wow, the same result. So I think its something to do with the deflection of the bearing beam. But its connected to the car and the car to the rail which has an infinitely stiff connection to the "floor" this is weird... So the general rule about spacing may not be correct?? More investigation needed. Hate mysteries... but we are talking about deltas in the 0.01mm band...Peter

[peteeng]

Hi All - I thought about the issue and decided to make the gantries perfectly rigid. Everyone wants a rigid gantry and this isolates the result to the bearings and rail. And this gives the logical deflection series. 150mm- 0.009mm 200mm - 0.004mm 300mm - 0.002mm and 400mm - 0.001mm. But these are down in the 0.001mm band. So it seems the gantry is complex and requires individual analysis to get the deflections down. Cheers Peter

[mactec54]

Hi Mactec - So the machines built with twin rails and 4 cars each side did they do better then a single rail and 2 cars each side? regards Peter

They are much better when you have a tall Gantry anything over 250mm to 300mm of Z axis travel and they can be a smaller rail size than if using a single one

[--JawZ--]

Hi all,

After seeing Peters results here, and the insight he has given me. I made a new design of my Gantry.
And thought it would be a good idea to share the results here, now this is not a 1:1 comparison to Peters simulations, but it can give some insight of the difference between previous design vs new design.

Previous Gantry "beam" design content;
1x 8Hx220Wx2040L mm (~8.9kg)
1x 10Hx220Wx2040L mm (~11.7kg)
2x 160Hx80Wx2040L mm Light extrusion (~21kg/piece)
16x 80Hx40Wx60L mm Light extrusions (~0.2kg/piece)

New design;
1x 12HxW220xL2040 mm (~13.7kg)
1x 12HxW198xL2040 mm (~12.6kg)
1x 12HxW178xL2040 mm (~11.5kg)
1x 10Hx200xL2040 mm (~11kg)

This was not only a change in the beam itself, I also extended the cars spacing as this thread is about, from 2/3 of a car spacing, to 2 car spacing.
Also I thought I would check to see the placement of the cars, either on the outside of the gantry, as per Peter's simulation. Compared to my inside placed cars structural strength and stiffness.

So I have included some images, I might have done something wrong in the simulation to generate an accurate representation of how the loads are going to be applied to this gantry design, but it will at least show a difference that can show a good enough estimation between each placement. I believe.
Also to note, this is a bonded sim setup. Setting up a separated bolt connected setup in Fusion 360, is a really time consuming, as far as I know, task compared to the bonded approach.

Front_FreeDeflection_CarsInside Isometric_FreeDeflection_CarsInside

Front_FreeDeflection_CarsOutside Isometric_FreeDeflection_CarsOutside
So for the 1 side stiff and 1 side loose I used a 1000N force load, as Peter did. Then I made the car bolt holes stiff on one side, while applying the load on the other side to see the difference in stiffness the column re-design had, as well as the extended car spacing.
If this could be compared directly to Peters result, it would have shown a significant increase in stiffness, but bear in mind I have made two changes here, so the car placement is most likely not the reason for this big change. And it's not setup as Peters's setup either.
Having the car on the outside, as per Peter's test design, yielded a displacement of 1.646mm at the most.
While the inside car placement, as per my design, yielded a displacement of 1.534mm.

Now does this prove the inside car placement is best? It depends, on how stiff the rest of the columns and gantry beam is. If the columns have weak walls and base plate, the outside car would be prone to more bending along the width of the gantry. Which would cause the pinions, in this case, to be in an angled contact towards the rack/racks. This would lead to pre-mature and uneven wearing of the rack and pinion. As well as the cars themselves, but the biggest wear would be on the rack and pinion system.
But then again, if the the design is that weak, both options would introduce uneven tear to both of the linear motion systems. And without any practical experience or more in depth theoretical knowledge on this matter, I might be wrong with assuming the inside placement is, for me, the better option here?


Front_Stiff_DownForwardLoad_CarsInside Isometric_Stiff_DownForwardLoad_CarsInside

Front_Stiff_DownForwardLoad_CarsOutside Isometric_Stiff_DownForwardLoad_CarsOutside
The second set of images, handles the inside or outside placement of the cars on the columns only, this shows the difference the changed placement of cars had on the gantry. So this is a bit more on topic with this thread.
For this simulation, I applied a 1765N load on top of the gantry beam, in the middle, To simulate the X and Z-axis weight with accessories, with some plus margin. I also applied a remote/"push" load to simulate the spindle at it´s lowest point during operation, this load had the usual 1000N load applied to it.

The difference here is quite obvious, based on how I set this sim up. The inside placement made the gantry stiffer, due to it having less length to bend.
The column base plate, based on my sim setup, showed minimal difference. Which is a bummer, I need to re-think the sim setup to showcase this difference a bit better, but it seems to suggest that the inside car placement is more forgiving on the bending of the column base plate though....? But just a smidge better, in this sim setup.

Peter might have better results to give in this regard, if this is worth checking in to. I think it is, but we'll see.

Cheers //
Anders

[peteeng]

Hi Anders - Could you upload the images into the forum via the advanced features? links fail over time and some of the images are smaller then small stamps for some reason. What really matters with the simulations I think is to model push or pull loads at the tool or the collet. Then convert that to a N/um static stiffness then if everyone starts doing that we have a common comparison between designs. Some machine makers do quote their XYZ static stiffness. I sent a note to the Datron technical head and he said he would find out the machine stiffness for me but didn't get back to me. I think the min static stiffness should be in the std specs of a a machine. I think the more fwd thinking builders would benefit from this. They spec pretty much everything else!!

The bearing and ballscrew people spec the N/um stiffness as well. Cheers Peter

In your case the delta LHS/RHS column stiffness is important as you don't want the pinion twisting on the rack. But your gantry assembly is much sturdier then lots of other working designs...

[--JawZ--]

Hi Peter,

I would if I could, the edit post disappears after I have edited the post once. Unsure why that is.
Images small? I did make it so the preview is of a thumbnail size, but the links should bring you to the flickr full images. But if I can edit the post, I will be sure to make the changes suggested.
The push load is at the point of where the collet would be in the design, I did not include it due to the fact it would have taken a lot longer to simulate the various changes. And I was only interested, at this time, in the displacement of the parts that can be viewed in the image.
If the static stiffness is as simple as dividing the applied force with the amount of displacement, I will be sure to add in that bit of info into the post. Once I figure out how to edit the post...

How would the calculation be done, however, when having to loads, each affecting their own axis of moment. Would you combine the loads into one, and then use the formula F/D=N/um. Or is it more to it perhaps? I have never been in a position where I needed to learn these formulas, so I have no experience utilizing them in this manner.

Yes precisely, I just wanted to see how much of an impact this could have on A design, I figured it would be a good bit of info to have here.

[peteeng]

Hi Jawz - You have one hour to edit a post, after that its frozen. So you will have to add the info in another post. Combined axis loads are not quoted. Machines are linear so if you wanted to know the stiffness in a particular direction you could vector the stiffness. I've attached an extract from a thesis on machine stiffness that compares machines. Its old and modern machines are much stiffer now. Also machines are symmetrically stiff so a 1000N force in the +X direction should result in the same deflection as 1000N in the -X dirn. With "soft" machines this may not happen by the way especially in the cross (push pull as I call it) gantry dirn due to the poor torsional stiffness of some designs and the large offset the gantry provides. (technically this is because the load is not applied at the shear centre of the system) Keep making Peter

[shmuu102]

Question about rail orientation . No one seems to be asking about the difference between rails parallel to the plane of movement or perpendicular (most usually 90deg CCW )

What are the pros and cons of each orientation?

would there be any benefit to combining the two: a square rail on top parallel (possible 2 carts) and another (round FSBR to save cost) 90 deg on the side below the ball screw (or other drive mechanism) with just 1 cart?

maybe the round rail cost savings would not be worth giving up a interchangeable parts benefit.

[peteeng]

Hi Shmuu - Good thoughts. 4 cars are actually statically indeterminate ie one car is redundent. But 4 cars could be more stable. The combination of square cars and round cars is actually a good thought. sq cars/rails can resist moments in all directions, This means the load does not get past the first car, So in a group of cars the rear cars could be round to provide stability but not much load capacity as its not needed. In terms of cost if you use square cars you may as well use square everywhere. If round then round everywhere to reduce cost (and performance) Some may argue that why dilute the build quality using mixed bearings? Once you buy the square rails the cars are cheap so use 4 in a group vs 3. If cost is really the driver then use round and get less performance. You also pay for what you get. You can get expensive rounds and cheap squares... also the stack height of the two systems is different so you have costs in machining if you mount sqs and rounds on the same surface.

You don't usually interchange parts, a car may wear out so you flick it and put a new one in. Typically the closest car to the load fails first as its the one doing the most work....so do you pull the machine down regularly and rotate bearings like car tyres? maybe a good idea...Keep thinking.. Peter

[pippin88]

Square rails (most) have equal load capacity in all directions.

Unsupported round rails also do but are floppy floppy floppy

Supported round rails do not have equal load ratings - depends on direction of force. Strongest with force pushing down on the top of the car, weakest pulling up on the car.

So I don't see an advantage of being 90 degrees with square rails.

Round rails - can be good to put 180 degrees to each other.

[shmuu102]

So you see no difference in square rails different orientation? parallel (on top of the side supports) vs square rail 90 deg CCW (on the outside of the side supports)?

It seems like most commercial machines have them on the outside. Is this solely to keep them out of the debris being thrown from the cutting ?

[peteeng]

Hi Shmuu - They are mounted on the side so the machine can be side loaded. It also makes the column easier to mount but mainly its the loading I think... Square rails have the same capacity in their three directions and three moments. As a general statement their orientation does not matter. Merry Christmas I'm full of cake.... Ho Ho Ho Peter