Standard (Or Generally Agreed) Method For Measuring Run-out 8

[gcubed888]

First post in this forum - starting with a question I hope someone can help me out with; hoping there might be a definitive answer to this one.

The company I work for currently measures run-out (TIR) on 16 cylinder diesel engine crankshafts by using a steel-topped table, two vee-blocks (not solid vee-blocks, but with rollers) and a dial indicator. Now, some have suggested that a level, solid granite table should be used and others have also suggested that solid vee-blocks should be used instead of vee-blocks with (little bit worn) rollers. Furthermore, because of the length/weight of the crankshaft others have suggested that multiple vee-blocks are required to counter possible deflection of the crankshaft under its own weight.

My question is this - is there a standard method (ISO, ASME, BS, etc.) for measuring crankshaft run-out (TIR) using vee-blocks and a dial indicator? Any insight into standard methods and/or other methods and practical considerations?

Any help would be greatly appreciated.

 

[Belanger]

To check runout requires a datum axis, and one issue is that vee-blocks don't always yield a reliable datum axis. As the part rotates in the vee-blocks it could ride up and down slightly, thus throwing off the runout readings. A better type of holding device would be collets or even chucks to get more encircling around the datum feature(s).

John-Paul Belanger
Certified Sr. GD&T Professional
Geometric Learning Systems

http://www.gdtseminars.com

 

[MikeHalloran]

Solid v-blocks, collets and chucks are not good for crank journals, and collets and chucks get expensive in the sizes needed for big Diesels, so roller vee blocks are about the least awful means of support available and practical.

I hope the top of the steel table is at least a couple of inches thick, and reinforced beyond that, i.e., a foot or two deep overall; you need a reference surface that stays pretty flat under the considerable weight of a large crank. I.e., it should look like, or be, part of a large machine tool.

It's probably possible, if a little inconvenient, to rotate the crank within its own block, and remove the bearing shells and caps at the journal you want to indicate. ... but you have to make sure the block is properly supported and leveled on a big rigid table, because the blocks of big engines are flexible, too.

>>> What does the shop manual say? <<<


Mike Halloran
Pembroke Pines, FL, USA

 

[gcubed888]

Thanks for great advice so far, but I'm surprised that there is no official recommended set-up re type (and number) of vee blocks, type of table, how to conduct measurements, etc. Nothing from any official or unofficial body that I can see out there. Our supplier recommends we use 5 vee blocks on a granite table (16 cylinder crankshaft) - is that to combat deflection (sag) in the crankshaft if suspended between two vee-blocks? Nothing makes sense (no reasons why) - all seems rather arbitrary; maybe we should just follow the seemingly arbitrary directive and be done with it - will cost a lot of money though.

 

[pmarc]

I am afraid you will not find any document that could be globally recognized as a standard or a guideline for runout measurements.
As it was already stated solid v-blocks (even placed on perfectly stable granite table) are not the best option for establishing repeatable and reliable datum axis especially when form errors of portions of the shaft assigned as datum features are significant (e.g. three-lobbed). If rollers you want to use instead v-blocks are worn, the measurement uncertainity will not be smaller, that's for sure.

As for usage of additional v-block to limit the deflection of the shaft, keep in mind that whenever possible a method of supporting the part for inspection should reflect how the part is supported in working conditions. Is your crankshaft supported in 5 places in reality? If yes, then your supplier is correct. But if not, you will be checking a geometry in really non-functional condition, which may cause a lot of troubles in a future.

 

[dingy2]

Are we talking about circular runout or total runout? It does make a difference.

V-blocks for checking runout does not really follow the requirement since, usually, the datums are developed from full diameters rather than a V contacting the surface of the datum feature at either 45 or 60 degrees. If one diameter is a bit larger than the other, it would also throw off total runout (not circular though) since the readings would increase as we move along the surface of the feature towards the larger size. In other words, the center of the datum feature would be higher on the larger size datum feature. The use of chucks on each end is much better.

The dial indicator must contact on 90 degrees to the surface when measure either runout so move the indicator and base until one finds the high point (center). I have enclosed a picture and more info that might help.

Dave D.
www.qmsi.ca

 

[pmarc]

Not that I want to be picky, Dave, but do you mean that total runout always controls size distortion of toleranced feature? What if size tolerance of larger cylinder was for instance 0.1 and allowable total runout tolerance was 0.5 like in your example?

What is the reason that your chuck or collet is not covering whole length of datum feature D and why is the whole technique you described reserved for small coaxial products?

 

[dingy2]

pmarc:

If you cannot place the product in a chuck, then one cannot confirm either circular or total runout. That is the reason that I mentioned that it is reserved for small coaxial products. The crankshaft in question may be relatively large but can be placed in chucks at both ends.

Total runout combines off center, roundness and size distortion along the feature relative to the datum structure, probably datums A-B in the crankshaft case. Since the feature is a feature of size, it must meet the size tolerance first and then the total runout requirement.

In my example, the datum feature is inserted into the chuck full depth of the chuck which give a rigid datum and it is as far as it could go inside the chuck. If I were to have the crankshaft set up on chucks rather than V blocks, then I would only insert the datum feature just inside the chuck jaws on both chucks just in case the centers are not exactly the same.

Dave D.

http://www.qmsi.ca

 

[TheTick]

Thanks for great advice so far, but I'm surprised that there is no official recommended set-up re type (and number) of vee blocks, type of table, how to conduct measurements, etc.

Type: NONE
Number: ZERO

 

[DeanD3W]

I think the way the datum feature simulators are constructed should be driven by function... Some round shafts are supported in plastic v-bearings, so v-blocks as simulators with included angle, width, and spacing that match the mating assembly should lead to a valid runout result. Yes, the form, size, orientation, and coaxiality error of the datum features will affect the runout reading, but if the part is supported by the mating assembly in the same way then that's not all bad with regard to finding a meaningful runout value.

Plastic v-bearings can be found in printers.

If the assembly does not use v-bearings then I'm generally with the others who have brought up collets or chucks as datum feature simulators. On the other hand, if the cost is too high for such things, and if the size, orientation and form of the datum features is controlled well enough relative to the runout tolerance value, then the additional uncertainty added by supporting the part in v-blocks may be acceptable. It's no worse than the uncertainty introduced by approximating a hole's axis by using the center point of two circles within the hole, or approximating a hole's mating size by measuring a best-fit diameter... Those sorts of atrocities are committed thousands of time every day. In some cases the parts are nice enough, relative to the position or size tolerances applied, that those practices may be acceptable ways to save a bit of inspection cost. Doing these things without understanding the issues isn't OK, but as long as one understands and considers the issues as measurement set-up and process decisions are made then all should be OK, or possibly even optimal.

Dean

http://www.d3w-engineering.com

 

[gcubed888]

Thank you all so much for the help to date. Just to answer a few questions:

It's total indicated run-out (TIR) I'm trying to measure, not circular run-out (not entirely sure what circular run-out is though?).

If the number of vee-blocks (and noted the discussion on this) is supposed to mimic the number of points that the crankshaft is supported on when installed in the engine block, then ten vee blocks would apparently be needed when taking measurements on the 16 cylinder crankshaft out on the table.


I'm not a mechanical engineer - I just thought there would be a standardised method somewhere out there? Would prefer to use a "certified flat" table and just the two vee blocks for cost reasons: some people are saying this should be fine, whereas others are saying that the crankshaft will sag and this will affect TIR measurement (the crankshaft supplier method calls for five vee blocks, but no explanation as to why).

 

[TheTick]

The interesting thing about runout is that it is a GD&T spec that is defined around a specific measuring process. All other GD&T specs can be measured in many ways.

Measuring runout requires that the axis of rotation be held constant through rotation. Vee blocks all but guarantee that this will not be the case. If you are dead set on using vee blocks, you would have to first verify roundness of the datum features. Verifying roundness is a far more tedious, time-consuming (and thus expensive) proposition.

 

[dingy2]

If the Designer wanted datum targets as V blocks, then datum targets should be shown on the drawing as per the 2009 standard fig. 4-47. If the full diameter is shown as the datum feature with no datum targets, then the full diameter should be used rather than V blocks. Using V blocks does not comply with the requirement.



Dave D.

http://www.qmsi.ca

 

[pmarc]

Dave,
Assuming that full diameter is shown as the datum feature and there is nothing on the print specifying that only limited portion of that feature should be used to establish a datum axis, how should datum feature simulator A look like? I mean, what should be the length of that datum feature simulator in axial direction?

 

[dingy2]

pmarc:

The theoretical datum feature simulator should be round and full length of the feature excluding the inside rad in RMB. I guess we would try to fulfill this requirement as much as possible with the chuck closer to full compliance than the V blocks.

Dave D.

http://www.qmsi.ca

 

[ArtL]

Mr 888 I'm not really familar with large enfines, but smaller engine cranks are ground on centers previously machined into the billet or forging/casting.the journals, rod and bearing,are "crush ground" on the centers.In machine shops, this is checked, as it is ground, on centers.
I assume V blocks would be a reasonable alternative.V blocks are usually bought in matched sets, so using 3 V blocks will throw another variable in the whole mess.I feel a fixture is needed to keep the V blocks linear. A lathe bed type of arrangement would work.I've seen opposing " blade" type discs used as v block replacements, argument being the small linear surface will be less prone to pick up foreign matter between surfaces.The "factory" usually calls out specs and how to arrive to accurate measuring.I have found the main idea is to measure journal to journal, not journal to V block. So if journal #1 runs out .003",mark on journal high or low spot,one not both, Go to #2 journal, measure an see if the runout, if any, coincides with #1.Go down the row of journals, record the findings.I all sag the same way, then its probibly sag caused by weight.You could put crank on V blocks and check runout of machined centers on ends of cranks also.If checking cranks is fairly often, investing in perminent fixtureing may save a lot of time and confusion in the long run.

 

[ptruitt]

Perhaps it would be worthwhile to calculate the measurement uncertainty for each method that seems viable and give that inbformation to whoever needs to weigh cost of quality against risk.

Peter Truitt
Minnesota