Hi Mike,

I assume that the austenitic stainless steel exhibits the "normal" relaxation behavior if you perform tests on smooth specimens ?

If yes, then the load increase you observed may be due to stress redistribution around the notch root (relaxation on the surface and increase in the ligament). This may cause non-uniform deformation in the radial direction. As your test is displacement-controlled, a slight increase of load may be required to keep the deformation more or less uniform.

If you have creep data for this steel I suggest to run a finite element analysis to look on stress redistribution more precisely.

Dear sir,

as you are finding this phenomenon in some of the tests and not in all, I feel it may not be related to some microscopic phenomenon. There is a possibility of recovery, but I am not getting why that also is happening in some tests only. 

I think you should try to find out, what was special in those particular test which led to such behavior. One suggestion, if you used different notch geometries for different tests please check if you are getting this behavior for some particular geometrical ratios of notches (you may start with notch sharpness).

You may get some clue from this paper

http://www.osti.gov/scitech/servlets/purl/6410656/

Regards

Naveen

What kind of austenitic stainless steel are you using? If it is Metastable stainless steel then there can be some austenite-martensite transformations. Austenite has FCC structure and martensite has BCC structure. FCC and BCC structures are not the same in volume, so if transformation takes place there will be a volume change. You can easily measure if transformation is taking place or not by comparing the magnetic property of the sample before and after testing. If the magnetic property increases, it means martensite is increasing during the test.

Hi Konstantin

Yes we have uniaxial relaxation and many uniaxial constant load creep tests as well.  More information can be found on the uniaxial tests in the publication.  However, the notched bar have not been published yet and partly because I am not confident I understand the reasons for the specimen shrinkage, hence my question.  

We also have conduced Full FEA, With elastic-plastic-creep model.  In Abaqus using Non-linear geometry option.  

The creep behaviour is a short rapid transient primary creep followed by a very slow minimum creep rate, followed by a long extended tertiary (due to thermal softening).  The creep model ignores the tertiary as I consider it unnecessary for relaxation predictions.  Nevertheless, there is a threshold stress which is dependent on the prior plastic strain.  The uniaxial relaxation predictions are OK, see picture and publication.  The uniaxial relaxation data do show between 0.1 and 100h a period where no relaxation occurs (small increase in stress is observed but is this significant?)  The modified Garofalo model which we used for the primary-secondary creep seems to capture this adequately.  Albeit the stress in predicted to decrease always.  

But the notch relaxation tests show a definite and large stress increase.  I think there is a competition between stress relaxation and material shrinkage and in the notches where a small volume of highly stressed and strained material is relaxing, but where there is a much larger volume/length of material exposed to temperature the shrinkage dominates.  

The earlier picture in the question of the Unloaded-test is interesting as the shrinkage occured in the 1st and 4th dwells which were preceded by plastic strain but not in the 2nd and 3rd dwells which were elastically loaded.  This suggests that plastic strain is important, which is why I think this might be an anelastic/internal stress effect.  However, I can't yet discount a metallurgical effect.  Which is why I am asking this question.  

The FEA results show no such behaviour and should be adequately capturing the stress redistribution that occurs during both plasticity on loading (of which there is lots) and during creep.  What we don't have in the material models is any anelastic or metallurgical effects.  Albeit as the threshold stress is  dependent on plastic strain (I will check how this is implemented in Abaqus) the threshold stress should vary across the notch due to the plastic strain varying.  

Article Creep deformation, rupture and ductility of Esshete 1250

Thank you Muhammad

The stainless is definitely metastable, see the publications for the compositions.   

The FCC to BCC austenite to alpha' Martensite/ferrite transformation is certainly something we are researching at the moment and you are right it does occur with thermal ageing and is affected by plastic strain. Also, I will follow your suggestion and check magnetic response of the materials.  However, FCC to BCC results in a volume increase, which in my strain controlled tests should have led to a stress drop and not to a stress increase.   

Conference Paper Microstructural Evolution in Esshete 1250 Austenitic Creep R...

Article Analysis of ferrite formed in 321 grade austenitic stainless steel

Dear Michael,

You might already know that these materials also transform during plastic deformation i.e. strain/stress induced transformations.

Muhammad

See Figure 11 in the Link

http://www.ommi.co.uk/PDF/Articles/208.pdf

So there can be martensitic transformation during cold working. What can happen is that during relaxation the transformed martensite transforms back to austenite and you see a shrinkage. When you measure the magnetism, I would suggest to measure it after each stage i.e. on the virgin specimen, before unloading, after unloading and after relaxation.

Have a nice weekend!

Mike

I hope, you have the paper of M. Kawai "Creep and Plasticity of Austenitic Stainless Steel under Multi-Axial Non-Propotional Loadings..", published in 1989. Here effects of small plastic prestrain on the subsequent creep are discussed in detail based on uni-axial and multi-axial tests.

I my view your uni-axial test results show similar tendencies. During the loading beyond the elastic range a significant backstress (resistance against inelastic flow) is induced. The subsequent relaxation exhibits a low rate as the applied stress approaches the backstress leading to a decrease of inelastic strain rate (during the first relaxation stage). The same is related to creep: the primary creep stage after plastic prestrain is short. 

Now assume that after the plastic loading the specimen is unloaded such that the applied stress is lower than the backstress. Then the negative inelastic strain rate is expected and the load should be increased to keep the total strain constant during the "relaxation test". This might be the reason for phenomena you observed in uni-axial tests. I think, that coarsening of carbides (or similar phenomena) do not lead to a significant shrinkage, but may affect the decrease of backstress during the relaxation (or creep) test and influence the rate of processes.

For notched specimen the effect of negative inelastic strain rate may be enhanced as the  backstress field after the plastic pre-loading  is heterogeneous. 

Deformation causes martensitic transformation, Besides when sxposed to elevated temperature the microstructure change, you can see the Shaeffler diagram. 

Dear Konstantin and Mahummad

Both ideas (i) internal stresses effects and (ii) strain induced martensite are plausible and the good point is that I can discount the precipitation.  

We will try to improve our material, plasticity and creep models and I will investigate the martensite using eddy current probes and dilatometer.  

Many thanks to both of you Mike

With regards to strain induced Martensite I have tested a large range of short term tensile test specimens from wide range of temperatures (20 to 650°C), with a Ferrite meter.  

There is zero measurement of magnetic phases in all specimens.  Even room temperature tests taken all the way to failure.   

I am therefore going to discount the strain induced martensite possibility and concentrate my investigations on internal/back stress effects, which I know are significant in this steel.  

Hi All

I am still plugging away with this issue of stress increase (material shrinkage) during creep and stress relaxation.  Of course the experiments and material modelling I need to do will take some time more.  

However, I remembered that the test with relaxation had included a diametral extensometer.  Therefore, this could go some way to helping explain the phenomenon.  

In the attachment the load increase can be seen in the plot of stress versus Δdiameter. However, in the plot of  Δdiameter versus time, there is absolutely no perturbation during the first creep dwell (compare this plot with the attachment in the original question Test-result.bmp).  The diameter continues to decrease very smoothly even though the stress is actually going up for the first 150hours of the first dwell.   

I am of course trying to wrap my mind around what this means.  I think it is telling me that Konstantin's suggestions should be followed up very seriously, i.e. we have a plasticity creep interaction, which is causing an anomalous stress redistribution.  My logic is:

Does any of this make sense?  

Have you seen any evidence of sigma or Laves phases. I think both phases have higher densities.

Jerry

The material Esshete 1250 has only 15%Cr and is generally much less susceptible to inter metallic formation than 300 series stainless steels.  Nevertheless, a very small amount can form after a very long time at >650°C.  This shrinkage I observed was in the first few hours at 550-600°C.    

A very wild hypothesis: 

Is there a possibility that a residual stress pattern already exist in specimen (e.g. shot peening type - high compression value at surface and low tensile value in larger zone towards center). On this residual stress pattern if a typical stress pattern of notch (high tensile value at surface and low at center) is superimposed, there may come a possibility of shrinkage (axial compression) of notch zone in initial stage of testing. I feel a FEA with this assumption may also reply no increase in diametrical contraction.

Regards

Hi Naveen 

It's a good thought, however, once all the plastic strain that is put in on loading (see first post, see Test-result.bmp) any residual stresses that might have been there before testing would be wiped out.  

I suggest you to evaluate a good measure of experimental uncertainty in your tests and the data you use to correct the data (elastic modulus, poisson`s ratio). Elastic properties and CTE are very hard to determine by simple tests at high temperature.

Hi Caner

I agree, to generalise +-10% is typical and sometimes as much as +-30% for such physical data!  

Hi Mike,

in the attached file there are some equations to support your assumptions. Of course these equations are only valid for smooth specimen. Here the change in diameter is proportional to the stress, Eq. 2, and the strass rate is proportional to inelastic strain rate, Eq. 3. Do you have data on change in delta d for smooth specimen ? If the stress increases (a skight increase is observed in your data) the transverse normal strain would increase. For the notched specimen the decrease in delta d for increasing load might be possible, but let us understand firrst the behavior of the smooth specimen.

Hi Mike

I was thinking that this anomalous stress relaxation or this negative creep relaxation is geometry induced or material induced...If you can do a compressive stress relaxatiuon test by holding your notched specimen at the same negative strain what can be the behavior...if there is same anomaly repeated, effect of geometry may be dominating factor. Like this strange relaxation, the negative creep shrinkage due to tensile straining is reported phenomena..but I saw people done relaxation on a superalloy which is prone to negative creep Waspalloy and they saw no anomaly in it...I think this can be geometry induced factor..I am attaching a link to an old report ..please see

Is there any instability in your strain time signal both axial and diametral....if not you also can think of a FEM simulation with similar eta factor (stress triaxiality factor ...mean hydrostatic component divided by Von Mises component) under compression...this can give you same Mises stress contour but negative hydrostatic stress values...then the radial and axial stresses can be mapped and looked into..

http://contrails.iit.edu/DigitalCollection/1954/WADCTR54-175part02.pdf

Bonjour Mike

What type of austenitic stainless steel ? Cast or laminated ? Which heat treatment? Is it possible to have a recrystallisation at 600°C?

Salut Joël

It is wrought austenitic stainless steel, forged into billet, pierced and then hot extruded into pipe 65mm thick.  Then solution treated at 1080°C for 45 minutes.  

The steel itself is a UK steel Esshete 1250 (15Cr, 10Ni, 6Mn 1Mo), which also contains 1%Nb and 0.3%V to form fine MX carbides, also M23C6 are formed.  It usually has a fine grain size ASTM 7.    

To be honest I have not conducted any metallography on the test specimens 

I have finally got around to looking at the results of some elastic-plastic-creep finite element analysis results for these tests.  As suggested by Konstantin at certain points in the structure the axial stress is predicted to increase during stress relaxation.  This is as suggested due to stress redistribution.  The FEA results show only a small effect and I have not yet determined if this effect is big enough to make the net section stress increase.  I have the results but need to dig deeper into them.  

However, while the stress redistribution effect is realistic my judgement is still that it is not a big enough effect to explain my test results.  

Michael William Spindler - is this, by chance, a stabilized grade? If so, I would point to the work of Li and Messler in the Welding Journal who show volumetric contraction under constant load, correlated with Nb(C,N) precipitiation (June 2000). I expect to publish on this in the near future.

Is there a chance you have some delta ferrite present in the original sample, and it is being soaked out at the higher temperature? Metallography would show this, but a Feritscope or just use a magnet could be used to see if the ferrite is decreasing in your tests? You could run Thermo-Calc with your composition to see if ferrite might be present from the cast structure and see the temperatures where it might be dissolving.

Dear George A. Young and Jerry Arnold

Thank you both for resurrecting this old question, I had forgotten about it until I saw your comments:

George A. Young I have seen this effect in both stabilised and unstabilised austenitic stainless steels (forgive my UK English spellings). I certainly agree precipitation of higher order phases such as Nb(C,N) will lead to a contraction but the contraction we have observed is bigger than can be explained by Nb(C,N). I will have to go back 20 years to find the calculations. I will add them if I can find the calcs.

Jerry Arnold I agree delta-ferrite transforming to a mix of carbides and austenite will also lead to a contraction. The steels we have observed this in are are wrought (not cast). One is definitely zero% delta-ferrite the other is

Interesting item Mike. It brings three thoughts to my mind:

- This has also been observed in some of the 316H SUS grades under plain round bar very-long-term stress relaxation tests, see works by Ohba et al 1986 (the paper is in Japanese, but I can share translated copy). Their increases were occuring around ~ 500-1000 hrs into the test, prob ~ 10% increase in load. The authors attributed this response to effects of decomposition kinetics of delta ferrite, M23C6 and carbon which alter the Type III and respectively Type II residual stresses in the material. In essence, they provide a somewhat similar description to your response to Jerry. The observations of this are at 550C in the 316SUS grade, and are sensitive to initial strains which as we know affects precipitation of both main and higher order phases. In Esshete, similar changes can happen.

- The anelasticity is a another avenue to explore indeed, and it fits perfectly in the case of notched samples where depending on the geometry and loading conditions, redistribution of stress (last point) combined with any micromechanical processes of unloading may affect the load registered on the cell noticeably. I am going back to the work by Ashwin Rao who you well remember - the noted anelastic strains were actually pretty notable at ~0.03 - 0.04%, and the longer the dwell, the higher the anelastic strain grew during the first dwell....

- The mechanics of the notched sample is also critical. I fully support Konstantin's and Kaustav's feedback. With a suitable constitutive model, you could capture aspects of this. But a model is as good as you make it so it may qualitatively give you some info but not the full picture.