Foundations of Mechanical Accuracy by Wayne R. Moore

[dgfoster]

I had a little time in the airport yesterday so I pulled out the computer to pass the time.

I did make the initial setup in my shop as schematically drawn in the first two images. The base is a granite surface plate that is flat (not really important for this use) and stable---very important. On that are 2-4-6 blocks that I ground a few years back to be very flat, parallel, and square though the main important feature for this use is flatness---no rocking whatsoever. Then the three pivot points are as shown. And on top of all that a raw, thermally- stress-relieved 36" Featehrweight Camelback that I milled at each end and centrally where the indicator will make contact. I made a small "puck" 1/2 in diameter and centrally relieved to sit on top of the round probe tip so that there was better/more reliable bearing of the tip on the underside of the SE.

I have not really had time to make certain the pivot points are going to be rock solid. If they are not, I have other plans for support. But, my goal is to have an extremely solid setup that is not subject to disturbance from minor vibration (likeing walking on the concrete slab or traffic passing by on the road 75 feet away.)

THe checker probe will be placed on a fairly robust base I made a couple years to hold a tenths-indicating electronic Federal indicator. Happily its probe barrel and the barrel of the Mitutoyo are both nominally .375 OD. It is convenient to adjust the probe stick-out in this base and the base itself is heavy and very flat so it does not rock. It will slide under the SE in a way similar to it as shown in the photo below.

The base of the support top side.

Underside centrally relieved.

A couple years ago I posted a video of this indicator/base setup being used to look at the flatness of one of my 18" prisms.

My plan is to get everything setup as shown but with a 5 gallon bucket resting in a saddle on the midportion of the SE. Then measure accurately the position of the SE undersurface and confirm that the mount is stable. Then fill the bucket with 25 or 30 pounds of water and measure deflection which I expect to be on the order of .0001" Then, I'll leave the weight in place for varying periods of time. At the end of a given trial period I will release the water via a stop on an attached plastic tube. I figure using water to load and unload is very unlikely to inadvertently side load the setup and introduce error. I will be looking to see if there is any persistent deflection ("set") after loading. Based on Rich's description (which was not really specific as to the number of days elapsed) in about post 30 above, a period of a few days should be enough to make a set occur if one to occur at all. I'll try to keep the setup going for as long as practical to make observations.

Using the same setup, I can see what distortion of the SE occurs with bare-hand contact with the bow of the SE so that a quantitative assessment of just how much disturbance occurs related to local warming of the bow. Will this be worse of better than that of the angle plate shown in another thread. Just after setting up, I did a couple of very quick trials of hand contact and was surprised to see that it look like deflection of the SE base was slower than that of the angle plate. We'll see if that holds.

Finally, I'll also be able to see how much deflection occurs from simple loading. I have measured that previously with equipment ten times less sensitive than the checker Carbidebob gave me.

Temperature variation of the garage will be a potential source of error, of course. But, if you notice we have basically 3 seel columns that are parallel reaching up to the SE sole---the two block/pivots located at the ends of the SE and the probe support. So, slow temp variations will be more or less equally affecting those members. Happily, this is a time in the PNW where day to night temperature variation is relatively mild. I may well set up a standard near the SE that will be of solid steel construction and to which measurements of the plate-to-sole distance can be compared to validate stability of the measuring device.

I am away from the shop for a week. So, I probably will have little to add to this thread for a week. THe exception being if I find relevant information in ASU's Polytechnic library as they hold a copy of ASM Handbook Volume 1 which delves into cast iron. I recently picked up a couple other volumes dealing with machining and casting. Though the information is not totally current, the basics of machining, casting metal, steel, and cast iron have not really changed fundamentally. So, I am learning the books are a concentrated trove of interesting facts.

I do hope people with constructive criticism will not hesitate to chime in. I'd like to hear opinions on how the setup could be modified to do a better job.

Denis

 

[John Garner]

Let's go back to an earlier topic in this discussion, the "It can't bend" phenomenon. My observations over 40+ years working with spaceflight hardware have convinced me that this phenomenon manifests itself primarily in one of two seemingly contradictory ways, the first involving big beefy steel tooling, and the second with composite structures built on that tooling.

In the first case, I'll cite an antenna assembly tool based on a commercial machine-base table. The machine base was roughly 2 foot x 4 foot, with a 1 inch ground steel top welded to a symmetrical support frame made of 1/4 inch thick 4 x 4 inch square steel tubing. The support frame had 4 steel legs of the same size tubing, and each leg was tapped for an1-8 UNC leveling screw.

At one end of the table, a squat-box bracket was pinned and screwed to the table top; at the other end of the table was a slender 5 foot tall tower bracket was pinned and screwed to the table top. Each bracket had a welded plate that were machined in-situ (on a good Cincinnati-Gilbert HBM) to locate flight-structure interface plates to "machine tool precision".

My employer's Quality-Assurance folks witnesses the on-machine inspection of the finished assembly fixture, and then witnessed the setup of that fixture in our employer's facility. The next inspection was of the first Antenna Structure built on the assembly fixture.

Disaster. QA witnessed the re-fitting of the Antenna Structure onto the assembly fixture, and it fit.

The assembly fixture and full list of Actual and Nominal locations of the first-article Antenna Structure went back to the machine shop for correction, which was witnessed by both hands-on QA and their managers.

Having thus corrected the problems, the assembly fixture was returned to the plant and set up again.

The first Antenna Structure built on the reworked assembly fixture went off for first-article inspection. And again failed miserably.

Rinse and repeat a couple more times.

Finally, with both my employer's and the machine-shop folks exasperated, someone had an idea: Let's have the Spacecraft folks take a look at the problem. My team walked the half a block to the Structure Assembly facility, for a briefing on the problem.

Our very first question was about the fixture leveling process.

We learned that the written directions called for 1) raising one jack screw fully off the floor, and lowering the other 3 jackscrews approximately 1 inch from their fully-retracted positions. Then, using a Master Precision Level (such as the Starrett 199Z), position the level midway between and in-line with two of the lowered jackscrews, adjust one of those jackscrews to center the bubble. The the level was to be moved to be in-line between the third partially-lowered jackscrew and whichever jackscrew of the first pair allowed the level to be perpendicular to its initial position. Once that was done, the raised jackscrew was to be lowered to contact the floor.

At the encouragement of the QA folks, my team verified that the table surface between the three "adjustment" jackscrews was indeed level. As soon as we had confirmed that the written instructions had been followed, QA management stated that the Assembly Fixture had been properly set up, and my group needed to explain why the several First Articles had such large errors.

I pointed out that the written process 1) did not specify which jackscrews were to be used for setup, and 2) that the levels of the screw-to-screw lines on the table surface had not been verified.

There was a chorus of "But it's a steel weldment, it CAN'T bend!" So I picked up the level, placed it between the leveling screws in the short direction, showed everyone the bubble position, and then moved the level to the long direction line and showed the crowd the bubble.

Several of the looky-loos protested "It can't bend like that!"

Ultimately, the process was revised to require making all four jackscrew-to-jackscrew lines level, at which point the whole shebang from machine shop to first-article inspection went as intended.

At the other end of the spectrum, we had a roughly 3 foot x 8 foot carbon-epoxy composite flat-panel structure that would not properly mate to the spacecraft structure. The interface design featured 6 inch-or-so square mounting pads that were supposed to be coplanar, on both to-be-mated structures.

On discovering the problem, the powers-that-be directed the smaller of the two structures be taken to a large surface plate and checked for planarity. That structure sat flat on the surface plate, so the p-t-b concluded that the larger of the two structures must not be coplanar.

But careful measurements by my group showed that the coplanarity of the larger-structure interface pads were within measurement uncertainty of our theodolite-based coordinate measuring system. I reported those results, which were challenged because the smaller structure lay flat on the surface table.

Eventually, I was able to convince t-p-b to raise the smaller structure off the surface table on 1 inch x 3 inch diameter parallel spacers. When the structure lay happily on the parallels, the looky-loos basically said "See, we told you so!"

When we measured, one at a time, the sag of the structure at the mounting points after sliding the parallels out from between the surface table and mounting pad, eyes were opened. A couple of the mounting points sagged only trivially, but others sagged as much as 1/8 inch.

Interestingly, some of the looky-loos walked away insisting that something was wrong with the sag measurements because carbon-epoxy can't possibly bend like that under 1-G.

 

[lucky7]

John, did you work at JPL? If yes, did you work with R Chandramouli?

 

[John Garner]

Lucky7 --

No, I spent 1975 to '79 in the Air Force aligning launch vehicle guidance systems. After that, I came as close to home (San Francisco) as I could afford (San Jose), commuting up The Peninsula to do spacecraft alignment.

My employer does far more commercial work than government work, but is very shy about having its name mentioned online. Even though I've been retired for a few years now, I still respect their sensitivity.

 

[Richard King]

Once the readers learn about 3 points they will see it at cafes and bars. Many tables and bar stools are3 points. Look under granite surface plates, at machines like Heald, Blanchard, Monarch, Hardinge, Cinc. Tool and Cutter, Moore and Pratt and Whitney jig bores, centering machines, camera tri-pods, surveyor tri-pods, etc. The old timers used 3 points a lot. I bet if we asked the machinist who ran surface grinders on the Abrasive forum, many would tell you how they ground parts sitting on 3 points and blocked around them.

 

[dgfoster]

John, given your deep background in the problems of alignment and precise measurement. I take even more to heart your suggestions to for use of glue to attach the pucks to SE. I infer from your recommendation that you would not worry about the stability of the glue over time. I also thought about wicking a drop of thin cyano glue between both the SE pucks and the granite pucks thinking that might solidify their contact. Do you have any concern/experience regarding it growing or shrinking over time and possibly introducing error?

Denis

 

[John Garner]

Denis --

I don't have any significant experience with "soak" durations longer than a few days, but we found that the short-term stability of test setups is significantly better when "gravity (or magnetic) attachments" are backed up with mechanical or adhesive clamping.

In most cases, adhesive clamping has been far easier to implement than mechanical, and a few dabs of hot-melt adhesive at the edges -- NOT between -- the test artifact and its support have usually been 1) quick and easy to apply, 2) effective, and 3) easy to remove when the testing is done.

Incidentally, a cordless hot glue gun is especially easy to use. I don't know if rechargeable models are being sold today, but the old-fashion models that heat a high-thermal-mass glue gun when the glue gun is in its stand have been very, very satisfactory.

We have, from time to time, used a fast-setting epoxy for temporary clamping, between layers of "prescription tape" that's been stuck onto both the test artifact and its support. Usually we scuffed the to-epoxy surfaces of the tape with a non-woven mesh abrasive and ethanol.

That said, I would not use layers of tape and glue for your testing. Small dabs of adhesive at the edges of metal-to-metal interfaces seems much more suitable.

John

 

[CarbideBob]

Iron and aging.
"green' iron machines differently than aged iron from the same source and foundry. Way different and I do not want to see that stuff only 48-96 hours old.
Things seem to settle down after about 2-3 weeks. Better at months and longer.
I do not care who says what or any theories. There is a difference about how it acks and size control at cut and after.
The just in time auto world made this a problem for tooling guys.
One can change cutting tool grades to try to help but it still is
Brake rotor, cast iron block and crank guys know this pain. Customer not happy.

 

[J_R_Thiele]

Denis I know that it is tempting to use a straight-edge as you have them- and know they have been stress relieved- but they are not a good choice when used as illustrated. The arch shape is used as it does not deflect as much as other shapes. That is good when in use as a straightedge- but not for this test- where you WANT deflection. Turn the straightedge on its SIDE with one of your points at the end and the other two about 1/3 of the way to the other end. To do this add a generous weight over the single end point. This leaves 2/3 of the straightedge to sag and twist. If you have the room and another stable straightedge have the sideways deflecting straightedge extend over the edge of the surface plate and use an "upright" straightedge to hold your measuring probe. To load the deflecting straightedge hang a bucket from the end. After getting baseline measurements for a week (?) add dry sand to the bucket. No evaporation and weight change with the sand. To remove the sand use a wet/dry vac with a HEPA filter. You can do this without touching the bucket.

The above will allow for more sag to occur- but also twist. With one sensor the deflection can be measured. If you had two sensors you could measure twist also. You could detect the twist with a spirit level, but it would be hard to actually measure it without disturbing the setup. I know there are electronic systems that could measure the tilt- but that only works if you have one.

 

[johansen]

The method i was going to use was to cut a 1/2" square bar 20 inches long (more like 3/8 after clean up) from a cast iron table i have. Cut the ends parallel, bolt one end to the table and stretch it. Clamps would be used to hold it rather than drilling a hole

You could then experiment with various kinds of stress relief, such as making a torsion pendulum from it. As long as you dont touch the ends, its relatively simple to measure its length to a good accuracy.

 

[IceCzar]

Iron and aging.
"green' iron machines differently than aged iron from the same source and foundry. Way different and I do not want to see that stuff only 48-96 hours old.
Things seem to settle down after about 2-3 weeks. Better at months and longer.
I do not care who says what or any theories. There is a difference about how it acks and size control at cut and after.
The just in time auto world made this a problem for tooling guys.
One can change cutting tool grades to try to help but it still is
Brake rotor, cast iron block and crank guys know this pain. Customer not happy.

I've worked with a lot of materials with files. The feel of a file on a material tells a lot. I can definitively explain the case hardening of a limestone. But I've recently noted the same "feel" when working iron ( which I've been doing most weekends Plus)

https://www.practicalmachinist.com/forum/attachments/compress_20230128_163913_3047-jpg.385462/

 

[dgfoster]

Once the readers learn about 3 points they will see it at cafes and bars. Many tables and bar stools are3 points. Look under granite surface plates, at machines like Heald, Blanchard, Monarch, Hardinge, Cinc. Tool and Cutter, Moore and Pratt and Whitney jig bores, centering machines, camera tri-pods, surveyor tri-pods, etc. The old timers used 3 points a lot. I bet if we asked the machinist who ran surface grinders on the Abrasive forum, many would tell you how they ground parts sitting on 3 points and blocked around them.

Seems likely.

Denis

 

[Richard King]

The thread started -

Foundations of Mechanical Accuracy by Wayne R. Moore​

and now your talking about heat treating, cast iron straight-edges .

Why don't you start new posts so more people can learn. I bet many people see the title and skip it. My 2 cents worth

 

[dgfoster]

Denis --

I don't have any significant experience with "soak" durations longer than a few days, but we found that the short-term stability of test setups is significantly better when "gravity (or magnetic) attachments" are backed up with mechanical or adhesive clamping.

In most cases, adhesive clamping has been far easier to implement than mechanical, and a few dabs of hot-melt adhesive at the edges -- NOT between -- the test artifact and its support have usually been 1) quick and easy to apply, 2) effective, and 3) easy to remove when the testing is done.

Incidentally, a cordless hot glue gun is especially easy to use. I don't know if rechargeable models are being sold today, but the old-fashion models that heat a high-thermal-mass glue gun when the glue gun is in its stand have been very, very satisfactory.

We have, from time to time, used a fast-setting epoxy for temporary clamping, between layers of "prescription tape" that's been stuck onto both the test artifact and its support. Usually we scuffed the to-epoxy surfaces of the tape with a non-woven mesh abrasive and ethanol.

That said, I would not use layers of tape and glue for your testing. Small dabs of adhesive at the edges of metal-to-metal interfaces seems much more suitable.

John

Thanks for the additional information. The idea of not getting any glue/tape between the part and its supports seems logical.

I just bought a simple 60 watt corded glue gun as I use hot glue all the time shipping castings. The higher wattage compared to my previous one is a plus. A cordless one sounds handy.

Denis

 

[MCritchley]

Iron and aging.
"green' iron machines differently than aged iron from the same source and foundry. Way different and I do not want to see that stuff only 48-96 hours old.
Things seem to settle down after about 2-3 weeks. Better at months and longer.
I do not care who says what or any theories. There is a difference about how it acks and size control at cut and after.
The just in time auto world made this a problem for tooling guys.
One can change cutting tool grades to try to help but it still is
Brake rotor, cast iron block and crank guys know this pain. Customer not happy.

Those castings are probably knocked out of the sand within minutes of solidification! Most proper foundries will let castings cool overnight and then blast and green for a bit. Greening is a well known phenomenon, I’ve read a few papers long ago, I believe the casting is 99.999 settled after a week or two.
I feel that greening and age related stress relief are two different things.

I would think that a difficult to machine and inconsistent batch of brake rotors would be easier to machine if they were thermally stressed first.

 

[dgfoster]

Denis I know that it is tempting to use a straight-edge as you have them- and know they have been stress relieved- but they are not a good choice when used as illustrated. The arch shape is used as it does not deflect as much as other shapes. That is good when in use as a straightedge- but not for this test- where you WANT deflection. Turn the straightedge on its SIDE with one of your points at the end and the other two about 1/3 of the way to the other end. To do this add a generous weight over the single end point. This leaves 2/3 of the straightedge to sag and twist. If you have the room and another stable straightedge have the sideways deflecting straightedge extend over the edge of the surface plate and use an "upright" straightedge to hold your measuring probe. To load the deflecting straightedge hang a bucket from the end. After getting baseline measurements for a week (?) add dry sand to the bucket. No evaporation and weight change with the sand. To remove the sand use a wet/dry vac with a HEPA filter. You can do this without touching the bucket.

The above will allow for more sag to occur- but also twist. With one sensor the deflection can be measured. If you had two sensors you could measure twist also. You could detect the twist with a spirit level, but it would be hard to actually measure it without disturbing the setup. I know there are electronic systems that could measure the tilt- but that only works if you have one.

Your points are well-taken. On your suggestion, I may indeed use a different casting or, more likely, part of a scrap casting (they happen!) as a test piece. For instance I could saw off the bow and uprights from a 36, thermally stress relieve it, and then, after milling a surface, clamp it down to two supporting 2-4-6 blocks cantilevering about half or 2/3rds the slab.

Adding in twist considerations might be interesting. But I think I want to keep the experiment as simple as possible and try to induce "creep."

Based on some time I spent last night at the ASU Polytech Library in Mesa, I think "creep" is the correct term for "taking a set." Using that terminology I skimmed a couple chapters of ASM Handbooks Vols. 1 and 4. According to what I read there, creep should not occur in grey cast iron unless subjected to prolonged relatively high stress at significantly elevated temperatures---they looked at 700F. So, we shall see what actual real-world experimentation shows.

BTW, another important bit I learned from the librarian was that all of the ASM Handbook Volumes are available FREE online if you search "ASM Handbook Volume 1 (or whatever one you are interested in) Free PDF" The printed set costs around 1K per volume new though some can be found used for 50 to 100 dollars and the entire set is 10K. So nice to be able to pick through this rather authoritative resource from home. The print set occupies about 36 inches or more of shelf space and probably weighs 150 pounds. I am finding tons of relevant information concerning grey iron and heat treating. (I am thinking, at some point, of starting a thread here on "nuggets" I and I hope others will mine from those Handbooks. There are a numberof topics previously discussed here in other threads that are addressed) But it was also fun to look through volumes devoted to machining, grinding, welding etc. Some of the information is a bit dated. But the important fundamentals have not really changed much in the last decade or two.

Denis

 

[dgfoster]

Iron and aging.
"green' iron machines differently than aged iron from the same source and foundry. Way different and I do not want to see that stuff only 48-96 hours old.
Things seem to settle down after about 2-3 weeks. Better at months and longer.
I do not care who says what or any theories. There is a difference about how it acks and size control at cut and after.
The just in time auto world made this a problem for tooling guys.
One can change cutting tool grades to try to help but it still is
Brake rotor, cast iron block and crank guys know this pain. Customer not happy.

Those castings are probably knocked out of the sand within minutes of solidification! Most proper foundries will let castings cool overnight and then blast and green for a bit. Greening is a well known phenomenon, I’ve read a few papers long ago, I believe the casting is 99.999 settled after a week or two.
I feel that greening and age related stress relief are two different things.

I would think that a difficult to machine and inconsistent batch of brake rotors would be easier to machine if they were thermally stressed first.

MCritchley,

Your thoughts are exactly in line with ASM's comments on the same. See Vol4. Interestingly they addressed brake rotors as an example. They also talk a lot about how shakeout procedures have a major influence on residual stress and point out that the great majority of grey iron castings will have minimal retained stress if allowed to cool in the mold sand prior to shakeout. That comes as no surprise as the 1948 US Navy article cited here numerous time in the past made the same points.

They also flatly state that "Stress Relief. The relief of residual stress is accomplished by heating the iron to a temperature at which the stress is relieved by rapid creep. Of course, complex shapes must then be cooled uniformly so that stress is not reintroduced. Vibration has been promoted as a method for providing stress relief to iron castings. This procedure has not been demonstrated to be successful in a valid test." Pg 1453 vol.4

Denis

 

[dgfoster]

I've worked with a lot of materials with files. The feel of a file on a material tells a lot. I can definitively explain the case hardening of a limestone. But I've recently noted the same "feel" when working iron ( which I've been doing most weekends Plus)

https://www.practicalmachinist.com/forum/attachments/compress_20230128_163913_3047-jpg.385462/

It sounds like you may be finding the "skin" on the cast cast parts you are working on. Most likely those large casting (like 99% of grey iron castings) were not stress relieved---no need to. But, the great majority of retained stress in such parts will be in a very thin skin which may not file smoothly and evenly like sweet soft castings that have either been stress relieved or machined. (Shot blasting, commonly done as a cleanup procedure on raw castings to make them look nice, induces surface stresses as I think most are aware.)

Denis

 

[eKretz]

Ohhh, I like this thread. Good stuff. A couple points: someone mentioned that ground surfaces test slightly harder than milled surfaces. This one is pretty obvious I think... Hardness tests are generally done with penetration testers. Surface finish can actually make a difference in such a test. If pressing something into a perfectly flat surface vs a "spiky" rough surface, which do you think will penetrate further from the extreme top of the surface, easier? Also, proper grinding methodology needs to be followed. If you're burning the surface, you're obviously going to alter the hardness. Sharp, open wheel...

Lessee, what else. Iron moving and relieving stresses, stored outside, yada yada. Way back when, this was probably true. These days, I think not so much with proper thermal stress relief.

3-point support - VERY good. Not always possible, especially in large machining setups when rigidity is needed. I used to have a method for this. I would initially set up on 3 points, and let the area where I was going to put the last parallel sag under its own weight. Stick an indicator on the part, down to the table. Start jacking up the sagged side with the Enerpac, while monitoring the parallel on the other corner. As soon as that opposite corner parallel loosened up, I'd read from the dial indicator how much the "sag" corner had lifted. Split the difference by adding half that amount under the sag side with shim. On very big stuff, even on three points stuff will sag at the unsupported corners. Very difficult to accurately check with gravity acting on everything...

Very large fabrications or castings with smaller cross sections are extremely difficult to set completely stress-free. They're just too flexible under their own weight. A benefit to this is that usually it doesn't matter, since they'll conform to whatever you set them on or bolt them to.

And yes, whoever said that: EVERYTHING moves. I have proved this on more than one occasion to people who were similarly deluded about the stiffness of steel or iron. On large machines, the jacking screws are absolutely there for a reason.

 

[Richard King]

Those castings are probably knocked out of the sand within minutes of solidification! Most proper foundries will let castings cool overnight and then blast and green for a bit. Greening is a well known phenomenon, I’ve read a few papers long ago, I believe the casting is 99.999 settled after a week or two.
I feel that greening and age related stress relief are two different things.

I would think that a difficult to machine and inconsistent batch of brake rotors would be easier to machine if they were thermally stressed first.

A student brought a competitors angle straight edge to a class and the edges of the angle were hard as glass. We had to use a Dumore grinder to grind them low so we could scape it. I figured it was pulled out of sand to soon or when they stress relieved it they didn't cook it long enough, got it to hot or again pulled it out to fast.