Fatigue testing

[CarlesFEA]

Dear experts,

I have access to a fatigue testing machine (tension-compression) and have some questions about how to properly define and analyze the data we can extract from.
I am interested in characterizing material behavior (mainly steel, cast iron, aluminum and magnesium) on the high cycle domain (N>1e7 cycles)
Test will be conducted at either R=-1, R=0 or R=infinity.
I have been reading several references in order to clarify as much as possible several of my doubts:
[1] MIL-HDBK-5H
[2] ISO 1099: 2006 Axial force controlled method
[3] ISO 12106: 2003 Axial strain controlled method
[4] Fatigue Testing and Analysis, Lee, Pan et al.
[5] ISO 1210: 2012 Metallic materials – Fatigue testing – Statistical planning and analysis of data

In the past we have sent magnesium and steel specimens to an external lab in order to get the characterizations. Strain controlled test were done and each specimen was tested at a different strain level.

Q1 > Should we go for Strain or Stress controlled tests?
Pros for Strain:
a) it would be possible to compare with results obtained by external lab in the past.
b) We can have a fatigue characterisation that covers not only high cycle domain but also low cycle which we may be interested in in the future
Pros for Stress:
a) as stated in [1] and [3], it seems that Stress controlled test can be conducted at higher frequencies (20-35Hz vs 0.2-3Hz with strain controlled). Both references discuss about the possibility of switching from Strain to Stress when high number of cycles are achieved. As our main goal is high cycle domain, I believe speeding up test is of outmost importance. I haven't double checked with our fatigue testing machine if there is any kind of frequency testing limitation though…
b) more conservative results (from [1]: "under load control max tensile load will remain constant but stress will increase as the crack grows, resulting in a shorter period of crack growth")

Q2 > How stress/strain levels should be defined?
a) Several (3,4) levels with replications as stated in [1] and [4] (from [4]: " Replicate test are required to estimate the variability and statistical distribution of fatigue life")
b) Evenly linear or double log10 distributed (=> this tends to locate more specimens in the higher life regimes) levels as stated in [5]

Q3> As I am interested in high cycle fatigue, how to deal with runouts when running high cycle fatigue?
a) By using the staircase method (as explained in [4]): this method assumes that there is a fatigue limit which is not necessarily true for all materials and that you estimate it before you launch the tests
b) No fatigue limit is assumed and runouts are considered using Weibull and Maximum Likehood Estimation techniques (or other techniques) => literature, software??


I look forward to hear any experiences, comments, hints or suggestions you might have concerning this topic.
Thank you in advance for your time!

Regards,
Carles

 

[EdStainless]

What data do you need? Are you trying to see if you meet (or exceed) the MIL or MMPDS values, or is there some other form of data that would help your design efforts?
Regrettably fatigue is a very complex subject, and each type of test tells you something a little different.
And you have to careful with accelerated testing. Significant changes in loads or frequency can produce radically different results. Initial test validation (to models and historical values) can take a lot of work.
When dealing with HCF we have always tried to estimate what our goal was for mean life, and min (mean - 3sigma), then we would terminate tests at mean + 1 sigma. In our applications the min was 400k and runout was 3.2 million cycles.

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

 

[metengr]

Q1 > Should we go for Strain or Stress controlled tests?

It depends on the actual service application. For low cycle fatigue applications, strain controlled is what you would select for prototype testing.

How stress/strain levels should be defined?

Again, by the expected or anticipated service conditions. You want to simulate loading in service to the fatigue test loading cycles.