Cutoff of relationship between surface finish and fatigue strength? 1

[canty]

I think it is fair to say that there is a well established relationship between the quality of the surface finish and the fatigue strength of a part. A smoother surface has fewer points in which a crack can begin to propogate and affect the fatigue life of the part. Unfortunately, this general statement doesn't do much to guide design other than to make qualitative statements about the relative fatigue life expectations between two specified finishes.

A few years ago, somehow I came to the conclusion that there was a point in terms of decreasing the quality of a surface finish where there were diminishing returns. In other words, in terms of its impact on fatigue life, there wasn't a huge difference between, say, a 250 and a 125 finish (Ra microinch) - or even a 125 and a 62. At some point though, a fine enough surface finish starts to improve the predicted fatigue life of a part. I can't recall if this was in the neighborhood of 32, 16, 8, etc. though.

Unfortunately, I can't remember how I arrived at this conclusion. I've done a quick search through a number of papers on the subject, and generally they examine 2 or 2 different finishes. Not enough to derive the sort of conclusion I recall making. Does anyone else know of something similar? Perhaps a rule of thumb you use in your company for highly stressed parts?

 

[BrianE22]

My old machine design book has a graph of endurance limit vs. tensile strength as a function of the surface finish but the finishes are listed as forged, hot rolled, machined, ground, polished. You could then look up the typical surface finishes for each of those processes and combine the results.

The graph shows a significant difference between ground and polished as well as machined and ground but the differences get smaller as the tensile strength gets lower.

 

[Tmoose]

Bad Geometry can overwhelm heroic measures to improve surface finish.

A properly shot peened surface is often "better" than a polished surface.

If the highly finished surface is generated by grinding, grinding "checks" (cracks) can clobber the fatigue strength.

 

[dgallup]

Three different factors taking place there Tmoose.

Bad geometry introduces detrimental stress risers

Shot peening introduces favorable compressive stresses on the surface

Highly polished surface reduces cracks for fracture initiation sites. Grinding is not the same as polishing or superfinishing.

Sorry, I don't have a graph of fatigue life versus surface roughness. You might want to search in the spring literature as they go to some pretty great extremes on fatigue testing of compression springs.

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[dhengr]

Canty:
We do so much in the way of poor design, introducing stress raisers into the product, not properly accounting for the real world conditions which our poor detailing forces on the system, not paying proper attention to the orientation of high stresses, etc., that at some point surface finish becomes a secondary consideration, as long as it isn’t rough as a cob. I’m not trying to completely downplay the importance of surface finish on a high cycled detail test, but polishing the base metal quickly takes a back seat to good detailing and attention to good design. One of the early reasons for setting some surface finish criteria in fatigue testing was to kinda take that variable out of the picture when doing high cycle, S-N curve plotting and testing. We expected these tests to be fairly repeatable, except for the variables that our design and detailing introduced. This latter was what we wanted to be testing.

 

[EdStainless]

I have worked on projects with high strength stainless components where the geometry and loads had been taken care of.
They cared about surface finish differences between 20 microinch and 10 microing Ra.
And they could measure the impact.
Yes, it gets smaller as you go finer. But as stresses go up it gets more important.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[Tmoose]

Generally speaking Grinding can theoretically create a finish down in the range suitable for a crankshaft bearing journal. 10 uin Ra.
Not quite polished, but pretty danged smooth.

https://www.cnccookbook.com/wp-content/uploads/2017/08/SurfaceFinishRoughnessByProcess.jpg

As a practical matter polishing is usually required as a last step, after crankshaft grinding, to remove grinding fuzz and hard nodules on the journal surface.

Grinding is also quite capable of creating grinding cracks/heat checks, which if in the wrong location and orientation could clobber the fatigue life prediction.

http://www.douglasmotorcycles.net/aa-files/images/doug/articles/crank_tour/cracks.jpg

 

[mfgenggear]

Hi Canty

I would say surface finish is more important on surfaces like bearing journals, high precision gear teeth. and bearings. or any contact surface. in general it is more important ,( depending the application ) to start with material that is free of defects. if it is important to be defect free, then non destructive inspection is required.

Harden or case harden steels is prone to cracks or burns because of improper procedure.
there are many factors,

 

[canty]

All,

Thanks for the responses. After more searching, I can't seem to find the whatever original resource it was that led me to this line of thought. BrianE22, I found a similar (but not quite like) the plot you were describing in my Machine Design book.

In response to some of the other discussion, about defects or geometry vs. surface finish, I agree. Defects and geometry (design) probably do have more weight in the "calculus" of fatigue life. However, more so I was trying to approach it from the "all other things being equal" perspective and asking, "Let's say we're concerned about the fatigue life of a highly stressed part, how good of a finish do we have to make it to where surface finish isn't so much of a concern?" Qualitatively, a finer finish is better. But how much better does it have to be?

This discussion did eventually lead me to an interesting finding though; one that supports some of the earlier suggestions of a 10 or so µin finish. It's not the one I'm thinking of, because it's from 2017 (so there's another one out there as well!), but I think it supports this line of thinking.

Fatigue Life Estimation of Medium-Carbon Steel with Different Surface Roughness. They took 75 samples of three different finishes (note they're using Ra in micrometer. 0.4, 0.8 and 1.6 would correspond to roughly 16, 32, and 64 microinch, respectively), and ended up with a plot like this:

Looking at the above plot, say we had a Stress Amplitude of 300 MPa (~43 ksi). The 64 and 32 finished specimens tended to fail at around 1.25 million cycles. The 16 finish, on the other hand, tended to last until about 2 million cycles.

Granted, the above plot is a reduction of the actual data, which is a bit more messy:

Additionally, there were a fair number of samples, but only 3 different finishes. So there's not a lot of granularity in where that cutoff is. All in all, it might be a specious assumption, but I suppose it can't hurt to say you need good finishes on things you need good fatigue life on. And this gives a rough place to start from. At the very least to start with testing to verify the fatigue life of the design. Also, probably with all fatigue questions the probabilistic nature of it means that it's difficult to make such a hard and fast rule.

For anyone that is still thinking about this, on further introspection, I think it was actually a semi-log plot, so that when the number of cycles were converted to a standard axis, the difference was even more stark. If you find it, feel free to share.

Thanks again for everyone's help!

 

[WKTaylor]

Canty... is this question in-line with aerospace structures and/or mechanical components... or generic industrial/mechanical structures?

There are forums for aerospace and aircraft where we could discuss this in relationship to aerospace materials/parts...

There are a lot of other critical factors to consider other than just ASME B46.1 Surface Texture (Surface Roughness, Waviness, and Lay)... although this is the starting point for fatigue discussions.

a few examples, affecting fatigue crack initiation/growth...

Alloy/temper and raw stock 'form' [grain-flow/orientation]?: sheet, plate, extrusion, tubing, bar, wire, die-forging, hand-forging, forged-block, casting, etc

How is the basic surface finish attained?: machining, grinding, turning, drilling/reaming, grit-blasting, peening, as-cast, as-forged... etc.

Mechanical surface-finish quality' of a part is generally less critical than surface-finish quality specifically at holes/hole-edges and along all sharp edges/features.

Pre-stressing [peening, cold-working, etc] or unintended residual/embedded stresses or effects due to manufacturing/forming practices can be beneficial or severely detrimental

Fastener-to-hole 'fit' is a biggie.

Protective coatings can have a benign or detrimental effect... especially 'on-top of everything else'.

Regards, Wil Taylor

o Trust - But Verify!
o We believe to be true what we prefer to be true. [Unknown]
o For those who believe, no proof is required; for those who cannot believe, no proof is possible. [variation,Stuart Chase]
o Unfortunately, in science what You 'believe' is irrelevant. ["Orion", Homebuiltairplanes.com forum]

 

[LiftDivergence]

In addition to all the comments above, I think although there is definitely a correlation between surface finish and the fatigue life, the original question seems to be of the mindset that improvements in the surface finish say from 125 μ in. Ra to 63 μ in. Ra will give an improvement in cycles to failure, but will have asymptotic effect. Not arguing with the correlation, I just think that might not be the best way to think of it.

Remember that in the end, our knowledge of fatigue is overwhelmingly based on test data and correlation to it. There are a lot of complexities and great engineering in LEFM, EPFM, exact solutions for geometric effects on crack propagation, etc. But ultimately, even our crack growth models are correlation to test data. For example, even the NASGRO equation, with the fancy crack closure function and all the other improvements over previous equations (Walker etc.) is still just trying to curve fit to known behavior from tests.

For most rotating beam tests or whatever is done, the specimen is polished to a mirror finish. So that is one baseline of our knowledge. For that finish, we can predict a fatigue strength or endurance limit. Then, we apply special factors, one of which is for differences in surface finish, to knock this down.

So rather than thinking we have some baseline, and improving the surface condition will make things better and better at a decreasing rate, it may be beneficial to think that we have some known ideal baseline for which the surface factor is 1.0, and discrepancies in the condition will make the fatigue strength worse and worse.

You can get an idea of how the type of knockdown on the fatigue strength you get from different surface finishes as shown below:

Keep em' Flying
//Fight Corrosion!

 

[dgallup]

LPS LiftDivergence. What publication is the above taken from?

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[metengr]

Looks to be similar from Chapter 6 - Fatigue Failure Resulting
from Variable Loading

http://highered.mheducation.com/sites/dl/free/0073121932/365766/chapter06.pdf

 

[LiftDivergence]

The publication I referenced is:

Norton, Robert L. Machine Design: an Integrated Approach. Prentice Hall, 2011.

Specifically the fourth edition. One of my favorites, the way Norton writes is great.

Keep em' Flying
//Fight Corrosion!