Dear Sir.

Is very important to analize the relation between the porcentual variability range and the porcentual standard deviation.

Please see this paper:

https://www.researchgate.net/publication/237666670_EVALUACION_DE_ACEITES_DE_CORTE_EMULSIONABLES_EN_OPERACIONES_DE_AGUJEREADO

Excuse that the paper is in spanish.

Article EVALUACION DE ACEITES DE CORTE EMULSIONABLES EN OPERACIONES ...

The first important question for you to answer is: "Do you have the ability to measure crack length during the experiment?" For example, by electrical potential drop or compliance techniques?  Or more primitively, a long focal length microscope showing the crack on the sample during the test?  Such methods improve the amount of data you can get from the samples.  

Deat Dr. Jamie,

Which technique is the best one? We want to buy it? Which one is preferred?

Thanks,

Here is an answer I gave someone else recently:

Both methods can be quite accurate if done correctly. Based on my own experience I would say:

Compliance Advantages (with front face clip gauge or back face strain gauge): 1) Easier to implement with fewer potential sources of error. 2) Can also provide measurement of crack closure loads. 3) Works well on both conducting and non-conducting materials.

Compliance Disadvantages: 1) Limited to the temperature capabilities of the gauge 2) Does not work well for static loading crack growth as it relies on the compliance curve.

EPD Advantages: 1) Can go to very high temperatures (I've tested at 1300°C). 2) Works equally well for static and cyclic loading.

EPD disadvantages: 1) Non-conducting materials require conductive surface coatings. 2) Setup is more complex with more sources of error. 3) Cannot measure crack closure loads.

Both methods have potential errors if the material has crack bridging.

I would have two main remarks regarding the design of your fatigue experiments.

First, be aware that the variability of fatigue results is important and if only a small variation of the performance is expected from a given treatment, it is important to ensure that a large number of specimen is used for each condition. This will allow for sufficient statistical power in the interpretation of results. Nelson's ''Accelereted testing'' book has many precious information on designing and interpreting such experiments.

Second, if I'm correct, the compact tension specimen's notch is there to model a preexisting crack. Therefore, would the nucleation of the crack not be considered as previous to the start of the test in your particular case?

Best regards.

Dear laurent,

That is my concern. I do not know if I can measure the number of cycles needed for nucleation of a crack using CT specimens that do not have preexisting cracks.

I am not sure if I am understanding you correctly, but if your CT specimen has not been previously notched, the stress/strain distribution will be relatively uniform and thus,you will not know the location of the crack nucleation/initiation beforehand.  At worst, the main strain concentration will be around the fixture holes and you won't even have an adequate failure mode.

If you are to have a blunt notch, again you won't know where the crack will be heading and it might be hard to monitor its formation. may be a sketch of your specimen would be helpfull.

ASTM E647 will help you defining the specimen size and correct procedure so you'll get valide results.

I think somebody  has to help Aslan.

Aslan,  please tell me if your interest is to study the crack nucleation, why do you go for a CT geometry or any notched geometry for that matter? Please note crack nucleation and crack extension from a pre-existing flaw are entirely different phenomena.

If your interest is on "nucleation" you should use a smooth geometry. You may prefer an "hour-glass" specimen rather than a one with a uniform gage section for this purpose. You may perform fatigue cycles with DCPD attached to the specimen to detect the crack initiation. Alternatively you can use replica technique (an age old method, yet well proven) to detect the initiation cycles. But remember, as someone pointed out initiation cycles is highly statistical in nature. You require statistically meaningful data to support your results.

 I do not think you will get any valid information using any cracked (notched) geometry. The reasons are: (i) the state of stress ahead of the notch is going to be very high and would not reveal the "actual" initiation cycle for a nominal remote stress value (ii) The root radius of the notch may not be exact in all specimens and this will impose a variability in the state of stress even among different specimens (iii) if you are employing EDM for machining the notch (to have a fine control on root radius), the local metallurgical changes will alter the material property and what you have measured (!!??) as the initiation cycle is not that of the material that you are looking for, (iv) the crack will usually extend more in the specimen centre and less on the sides; compliance or DCPD gives an average value of the crack extension over the crack front and NOT the local value. I do not see any logic in using a notched geometry for crack nucleation studies.

Go for a smooth cylindrical hour-glass specimen. I prefer a circular geometry over a rectangular geometry for crack nucleation studies. Refer ASTM E466 for a specimen design. 

Thank you all for your comments. I think using a notched specimen for measuring the crack initiation is not very justified.

I would second S. Sivaporasad suggestion of using a cylindrical dogbone specimen in order to constrain the initiation of the fatigue crack while limiting the number of geometric features that can influence the stress distribution.

I would further advise on being very careful in selecting the manufacturing method of your specimens because surface finish can also influence your results and might hide the grain size effects that you are looking for. You might want to consider polishing the gauge section to a very low roughness so that the surface finish has minimal effects on the nucleation process.

Crack initiation and propagation behavior can be quite different for a material. You can use a dogbone specimen with a reduced cross section for the crack initiation study. What I would recommand is that a flat surface should be provided on the gauge section. This design can be found in some my published papers. Since the just initiated crack is on the scale of micrometer, a microscope must be used to observe the initiation process.

Thank you all. Actually I have noticed the surface finish effect. It has very strong influences on the results that might totally cover the grain size effects.

better you electro-polish the specimen prior to test.