Dr. Chin:

Notwithstanding all the details in the various ASTM or equivalent standards, I think that the choice of strain- vs. load-control testing depending critically on what type of fatigue test that you're doing.  The temperature is less important, apart from the fact of course that at 600C you'll need a high-temperature extensometer, if you elect to use strain-control.

For stress- or strain-life testing generally with tensile type coupons, the best approach in my opinion is to always use strain-control; indeed, it is mandatory if your applied stresses exceed yield.  You can get away with simpler load-control testing if your global applied stresses are below the yield strength - this would often be the case for high-cycle fatigue testing, but for low-cycle fatigue experiments, at any temperature, you will collect much better data using strain-control.

However, if you're performing fatigue-crack propagation tests, now with some form of precracked sample, load-control is definitely the preferable way to go to collect reliable data.  (Note that with a crack present, the global stresses are invariably well below yield).

The various standards will give you the specific details but the approach that I describe above is the best general approach to this issue.

ROR

the preference is strain controlled, but depend the equipment 

Like Laura has rightly mentioned, it may also depend on the microstructure, specimen type and what you are trying to get out of it, especially when it comes to modelling, you may discover strain-control is preferred....

Thanks Mr. B Krishna for the reading on the standards.

Lauren and Rudolph, would you mind to clarify why is it more preferred? Does that mean that the result from high temp load controlled test will not be valid?

Dr. Chin:

Notwithstanding all the details in the various ASTM or equivalent standards, I think that the choice of strain- vs. load-control testing depending critically on what type of fatigue test that you're doing.  The temperature is less important, apart from the fact of course that at 600C you'll need a high-temperature extensometer, if you elect to use strain-control.

For stress- or strain-life testing generally with tensile type coupons, the best approach in my opinion is to always use strain-control; indeed, it is mandatory if your applied stresses exceed yield.  You can get away with simpler load-control testing if your global applied stresses are below the yield strength - this would often be the case for high-cycle fatigue testing, but for low-cycle fatigue experiments, at any temperature, you will collect much better data using strain-control.

However, if you're performing fatigue-crack propagation tests, now with some form of precracked sample, load-control is definitely the preferable way to go to collect reliable data.  (Note that with a crack present, the global stresses are invariably well below yield).

The various standards will give you the specific details but the approach that I describe above is the best general approach to this issue.

ROR

The justified answer should consider,

1. Loading history

2. Is there any dwell time in the interested range of application?

3. Creep activated temperature

4. Expected stress in the component for instance in the critical local (vicinity of the notch) it's possible to reach beyond in such case it's highly recommended to use strain control as load control does not take into consideration plasticity and in the real component has identical strain behavior (for confined plasticity ahead of notch).

Considering these factors and interested application, further scope of your analysis you should make appropriate choice.

Thanks for the detailed explanation and the suggestion of best approach under different circumstances Robert O Ritchie . How will the difference between load controlled and strain controlled fatigue testing be manifested, say if they were to be conducted at similar magnitude of strain/stress level? Will the difference be reflected in a way such that a larger scatter in fatigue life or etc?

Thank you for the step by step consideration as well, Bhavik Patel .

Dr. Chin:

Again the answer to your question depends on the type of fatigue test that you're contemplating. In many respects, it is very important with regards to testing that involves crack initiation and growth. Up to the point of crack initiation, the difference is less important, apart from the fact that the material may be cyclically hardening or softening - load vs. strain control will clearly affect how the material reacts to this. However, once a crack initiates, the effect of load vs. strain control will be far more profound. With the growth of the crack, the compliance of the specimen will tend to increase. As under load control, the load will be maintained, subsequent crack extension will likely accelerate. Conversely, under strain (or displacement) control, the load will drop, which will tend to stabilize or even arrest subsequent cracking. Exactly what happens will naturally depend on the specific specimen geometry and how the sample compliance changes with crack length, but these are the general trends.

ROR

I totally agree with Robert O Ritchie . To paraphrase and put his point in a general way - use strain-control approach if the experimentally applied stress would exceed the yield strength of the test material. Use load-control if the applied stress is reasonably below the yield strength of the material. In the first case, if the crack initiates and propagates under stress above the yield strength, it will generate increasing high stress intensity factor at the crack tip causing the material to fracture in the shortest possible time, hence low-cycle fatigue. Under this condition very small amount of data will be collected. In the second case, because the stress is reasonably below the yield stress of the material far more number of cycles (high-cycle fatigue) will be obtained giving you the opportunity to collect more data before the material fractures. Note: for Nickel based superalloys, temperature of 600 deg may not be an issue.

Prof. Ritchie has answered the question. In terms of testing, it is easier to do the load controlled test than strain controlled tests. In the strain controlled tests, one can control the total strain or plastic strain. The constraints in the two cases will be different. At high temperature, one is also concerned about the time dependent effects - Hence for the test environment and the material sensitivity to environment become important. The creep component also becomes important depending on the ratio of the test temperature to the material melting point. In addition the loading ramping rate also becomes important. Significant work has been done at high temperature fatigue. We did some work on Nickel base super alloys.

While I completely agree with the answers of Robert O Ritchie and Kuntimaddi Sadananda ; as an industrial engineer I would generally control a test according to the conditions experienced by the component in service in its intended application.

With regards to experimental issues: If the hysteresis cycle exhibits macroscopic plasticity then strain control. If the material cyclically hardens or softens then strain control. If the hysteresis cycle is purely elastic then load control and strain control are equivalent and it might be worth considering load control as test machines are generally more reliable in load control than strain control.

Nevertheless, the conditions experienced by the component in service trump all other considerations and I would consider load control testing for a load controlled service condition even if there is cyclic plasticity and even if the material cyclically softens. I would nevertheless put an extensometer on the specimen to measure the strain.

Your question mentioned Ni-based super-alloys and 600°C and although I don't know what plant component/application you are researching I would hazard a guess that that component experiences conditions that include a mix of primary and secondary stresses. In addition, I would guess that some of those stresses are static and some are cyclic. My expectation that your component experiences secondary cyclic stresses leads me to believe that "strain control" might be correct for your tests. However, this is all guess work on my behalf and I would strongly suggest that you talk to the industrial end user to ask them about the conditions in service before making your decision.