[digits]

Hi guys - I have been wondering how us DIY'ers might go about building an accurate machine with reasonably long travels (say 2000mm/6 feet) using off the shelf linear rails etc.

It would seem that one of the biggest problems is making the rail mounting surfaces accurate along such a length without access to an even bigger machine to simply grind or mill the surfaces flat. I was wondering if anyone had sucessfully tried turning the problem on the head, and used the linear rails as the accurate surface to which to align a machine base.

Feel free to shoot me down, but, the rails are straight during manufacture, and so presumably straight when unstressed and resting on a flat surface like a mill's table or a large suface plate. If you could machine some accurately flat short cross pieces, couldn't you then mount them to the back of a pair of rails, and create a rail + sleeper affair, with the rails parallel and at the same height.

If the 'sleepers' had horizontal slots in them, you could then through bolt them to the verticalish side of some angle iron which you'd milled vertical slots into. The horizontal side of the angle iron would have slots or holes to allow it to be bolted down to some solid chunks of imprecisie iron that would run along the length of the rails. If you then adjusted and tightened all the bolts to get the imprecise bits as solid and square as possible, wouldn't you still have a set of precisely aligned rails, now bolted onto a pretty imprecisie but solid base? If you then added a dozen or so adjustable feet to the I-beams, wouldn't you end up with a large floor-standing machine base with precisely aligned linear rails?

Here's a quick CAD mock-up of what I'm on about:

[Drassk]

The reason linear rails have all those mounting holes is that they don't hold themselves straight, you have to mount them so that they're straight. In other words, they can follow a precision surface but they aren't one themselves.

Check this document out for a description of what I'm talking about, look at pages 12-17:
http://www.automation4less.com/pdfs/hiwinaglg.pdf

[digits]

Hmm, I'm not sure I read that datasheet the same way - you are not trying to distort the rail to straightness with the mounting bolts, but restrain it. The rails are precision ground flat and straight in the factory - surely they don't curl up like bananas as soon as they pop off the grinder...

Looking at page 7 of this Nook pdf http://www.nookindustries.com/pdf/NookProfileRail.pdf it seems to say that the rails are straight to within 170um across the length of a 2500mm rail on the lowest precision (C7) grade. C5 is to within 60um.

And on page 8 they have a paragraph all about rail straightness - basically saying that the rails have to be straight to allow you to mount them to a straight surface.

So, assuming that they actually are straight, I guess you could carefully align a couple of rails face-down on a suface table, and then epoxy a paving slab to the rear sides to keep their relative positions constant...

Or have I just had too much coffee and not enough sleep?

[DIYaholic]

digits,
I would think your idea is possible.....
But is it practicle?
First, how accurate are the 2 "I-beams"? Then, how precisely can you align them? Same goes for the cross ties. How accurate are they? You would probably have to precisely set/shim/align each one seperately. The next step would be to precisely set/shim/align each linear rail. Seems like a lot of work with way too much room for error.
I think the real question is: How flexible are the linear rails and can you use shims at attachment points to maintain their straightness?
Just a non-mechanical engineer's thoughts. Good luck with your design.

Randy,

[digits]

The point of my hair brained scheme is to fix the rails in place at the places they want to be in when they are unloaded. The I-beams and most of the rest of the 'base' are assumed to be warped in all sorts of nasty ways. Therefore the bolting points for the rails are fully adjustable, and can be tightened up to fix the rails in position without pushing or pulling the rails out of their natural, more accurate shape.

When you finally flip the thing over, the base will look like it's been run over by an elephant, but the rails should still be parallel to eachother. I'm guessing that some adjustable feet could then be used to securely support the warped I-beam structure on the floor without altering its shape.

Like I say, it's all a bit hair-brained, but it might just work...

[Jason3]

Hey Digits,

Funny Randy should mention railways - check out the THK JR type rails - they are intended for use as a structural member and will absorb some mounting error. I've used them and found them quite nice to work with.

Best regards,

Jason

[Drassk]

Jason's mentioned a decent solution, but the product is quite different from normal linear rails.

The problem isn't how accurate they are, it's how stiff they are. In the end, a linear rail is still a steel bar and will sag and bend just like one unless it's mounted to a reference surface. The accuracy to which they're machined just means that they can maintain the accuracy of the surface they're mounted to rather than degrading it.

[digits]

Jason, those rails do look interesting!

Drassk - I do get what you're saying - the rails are just steel bars at the end of the day. I guess what I'm wondering is whether you can just support them sleeper style at regular intervals along their length, rather than having to have a continuous, accurate surface below them. They will obviously sag/deform between the mounting points but round rails that are only supported at their ends are accurate enough for some applications - I would guess this would be a lot better.

If I can get the construction of the machine structure down to just drilling, aligning and bolting up, I should be able to build some pretty monsterous machines. Drills have the advantage that you can take them to the workpiece. Are there any magic machines that you can just clamp to a wonky I-beam and have it milled/ground flat along its entire length - if so where can I get one?

[mactec54]

I bet the THK JR rail cost is high

digits
You can not use the normal rails or even expect them to work as a structural member, for mounting the rails, You have to think of them as a bearing , for a low grade rail almost anything goes But they still need to be supported, If you want to mount precsion rails you have to mount them as per the manufacture spec in both directions, The straightness tolerance as well as the flat surface tolerance that is needed for them, & yes some are like a bananas when you get them, they are not straight or flat

[lgalla]

If you pickup a long rail it will wobble like a wet spaghetti noodle.It cannot support its own weight.Just laying it down will not be enough to ensure accuracy.Rails must be mounted to precision structural surfaces.The mounting bolts are slightly smaller than the holes in the rail allowing tweaking into perfect alignment.Sorry I do not know how to explain it any better.
Larry

[digits]

OK, so I need a large precision surface to mount my rails on - how do I measure the flatness of this arbitrary length surface to within 25um/m (~0.001"/40")?

And if you were trying to build the world's largest machine, how would you machine this mating surface precisely flat without access to a machine large enough to do so?

[lgalla]

Here is a big one,240"long.
http://www.cnczone.com/forums/showthread.php?t=19418

[digits]

Here is a big one,240"long.
http://www.cnczone.com/forums/showthread.php?t=19418

Thanks - there's lots and lots of useful and interesting stuff in there! :cheers:

[RICHARD ZASTROW]

digits, what if you assembled your adjustable frame with the rails facing down on Larry's famous flat surface plates? The rails would be as flat as the surface plate and the structural beams adjusted to them. If all is properly adjusted and secured in place, you could then turn the assembly over with the rails facing up.

Properly secured and leveled, you would have met your object, no?

Or am I day dreaming (hallucinating again)? I've been prone to do that since they doped me up in the hospital last December. LOL

Dick Z

[digits]

digits, what if you assembled your adjustable frame with the rails facing down on Larry's famous flat surface plates? The rails would be as flat as the surface plate and the structural beams adjusted to them. If all is properly adjusted and secured in place, you could then turn the assembly over with the rails facing up.

Properly secured and leveled, you would have met your object, no?

Or am I day dreaming (hallucinating again)? I've been prone to do that since they doped me up in the hospital last December. LOL

Dick Z

Well, that is pretty much what I described in my 1st post and had everyone disagree with!

Having read a lot of posts about aligning rails, it would seem that really long rails really are rather floppy and therefore not naturally straight.

The only rails I have are some 35mm Hiwins. My longest is only 900mm, and it appears to be both straight and strong enough to support its own weight. When placed on the flattest thing I have - 450mm of aluminium tooling plate I can't see any light between the plate and the rail.

Ofcourse there's looking flat/straight and measuring it. I have the rail up on 2 1-2-3 blocks about 400mm apart which are standing on the tool plate. I have set up a dial-indicator which claims to read 2um/0.0001" on a carriage placed on the rail. The tip of the indicator is about 150mm away from the carriage, so I think I should be measuring the vertical sag/bow in the rail across a 150mm distance. I zeroed the indicator when it was above one of the two supporting 1-2-3 blocks, and then took a measurement at halfway between the blocks. I read 32um of deflection. If I add 10kg of iron weights right next to the measuring point I can increase the deflection to 34um - this seems pretty good to me.

What I really need to do is to get the rail and blocks on to a large surface plate, and then use a height gauge to measure the sag directly. My gut feeling though is that this rail is too stiff to be aligned on a surface plate by its weight alone.

The apparent stiffness does however lead me to believe that if I were to support the rail only at its mounting holes which are 80mm apart, there would not be much deflection between them. I would imagine that if I had fully adjustable 'sleepers' at the bolt hole positions, I could level a few of them against a precision reference - e.g. a granite straight-edge 500+mm long, bolt them down tight, and then slide my reference along about half its length. I reckon this should give me a locally flat surface, but with an accumulated error along the length of the rail, but I suppose I could use the laser/tight wire technique to try and avoid this.

Once I had my planar sleeper surface, I could then bolt the rails to it, using the bolts to constrain (straighten) the rail.

I guess the design of the sleeper would depend on the inaccuracy of the main frame - ideally they'd have enough adjustability to be bolted on to a raw surface of a structural element, and I would probably avoid any tapped holes, to reduce the required accuracy of the drilling.

[Jason3]

Hey Digits,

I think the difference with what Richard proposed is that you wouldn't be relying on the rail being straight to establish an accurately planar mounting array.

You shouldn't need a surface plate to see if the rails is pretty close to straight - can you set the dial indicator up on the tooling plate near the centre of the rail with it still on the 123 blocks each end so the plunger is on the top surface, take a reading then flip the rail over and see if the measurement is the same? This should give you some idea how straight the rail is. You could do the same with the rail on it's side too.

If you support the rail every 80 mm in the case of a 35 mm rail, it's not going to bend between the supports unless you approach the rated load, partly because the bearing itself is not a point load - most 35 mm linear bearings have at least 80 mm or so of balls on the rail so it will tend to span between the bolts in the rail. If it is paired up with another bearing bolted to the moving axis then that will also help hold it straight.

Cheers,

Jason

[digits]

Ah, OK, sorry Richard - re-reading your post, I guess you're suggesting I use Larry's surface plates to pre-straighten the rails, then I bolt the frame on to the back of the rails without altering the alignment, and then remove the surface plates? So the surface plates become a temporary straightening jig, I guess?

Flipping the rails is a good idea Jason - I guess sag will cancel out, but an actual bend will measure as a double error.

The problem I have is getting a stable enough setup to allow me to flip the bar and trust the repeatability. My tool plate is non-magnetic so the indicator will move. I guess I could try this on the mill's table though - the rails only supported on 2 points, so the flatness of the table doesn't matter.

[klaabu]

For starters I would like to agree with everyone who stated that long rails will bend and sag like a rubber band under it's own weight as I found out the hard way. To make matters worse my 35 mm rails (both 3048 mm long) are mounted on the sides of the base table on very imperfect 6" x 3" x 1/4" rectangular tubing.

Here's how I did it:

1. Made base plates out of 2-1/2" x 1/2" hot rolled steel. Drilled and machine tapped all 40 M8 holes every 80mm on each plate. Drilled all minor holes (40 per plate) for 5/16-18 for mounting. I actually used the same drill bit for both as M8 and 5/16-18 are very similar in size.

2. Devised a gizmo on the picture for tensioning music wire. The other side is a mirror of this one and has just a little hole drilled instead of a guitar tuner.



3. Clamped base plate to the table frame tubing while aligning the top side of the plate to the tensioned music wire.

4. Drilled 5/16-18 minor holes through the frame tubing using holes in base plates as guide with magnetic drill press.

5. Removed the plate from frame tubing. Tapped holes in the table frame using 5/16-18 gun tap, enlarged mounting holes in base plate to 3/8" to accommodate for 5/16" socket head cap screw. Countersunk the top row as the screws there will actually be hidden under the rail.

6. Put it back on the table this time with screws and shoved shims behind base plate until .001" feeler gauge did not fit between wire and the plate without disturbing wire first. Thickest shim used was .03", thinnest .002".

7. Now it was time to align the top of the base plate to the wire.



8. Put on the master rail with the reference side up and repeated alignment steps for side and top using music wire set up.



For the reference, shimming and straightening base plate took longer than half a day but because of that shimming and aligning the master rail took less than an hour.

9. Repeated steps 3-5 for base plate on the opposite side of the table.

10. Built a contraption to probe base plate for supplemental rail to ensure parallelism with master rail in 2 directions.

[digits]

Wow - thanks for that! A picture really does paint 1000 words!

That is a very impressive looking machine you're building there! So hot rolled steel base-plates milled or ground flat in any way, or are they just raw? And did you have a machine big enough to drill them all the way along, or did you drill it in sections?

Cheers.

[klaabu]

Thank you digits.

Base plates are raw. I have no means to machine 10ft of surface without re-clamping it 6-7 times. That's what I had to do for drilling as far as I remember and each time I'd re-zero x-axis to the last hole.
Precision rails on plates with mill scale on didn't turn out to be much of a problem though - there were very few places where I had to prop it up with .002" and .003" shim stock pieces.