Design of Experiment for obtaining hardening material properties 1

[prz235]

Hello! It's my first topic here, so warm welcome to everybody willing to help me
I have been reading this forum ocasionaly for over 1,5 year, but (until now) I didn't need to register (due to the fact I'm not experienced engineer - I couldn't give much support to anyone).
However, my problem is quite serious, and I can't handle it on my own, so that's why I'm posting:
There is a need in my company to establish shape of part after stamping process, and residual stress distribution (also PEEQ).
Case is quite simple - 3mm S420MC steel sheet is stamped into "U"-shape in outer company. We don't know real yield stress of this part (I mean after stamping process), so our calculations of whole component (made of several stamped parts) are a bit uncertain.
My idea is to simulate this stamping process using real material properties, but those properties are the main issue - even though I know YS, UTS and A5, there is a problem with isotropic and kinematic hardening material models, which are not precise enough.
I decided to make experiment on our Instron Tensile/Compression testing machine of 3-point cyclic bending to establish combined hardening properties:
1. Positive bending (to achieve stresses over YS)
2. Unload
3. Negative bending (to achieve YS in opposite direction)
4. Unload
My problem is to how (using this Force-displacement data) establish "combined" hardening parameters (C and gamma). I will use Abaqus and Isight to corelate parameters, but I need values to start my corelation process.
As far as I know, parameter C is slope of hardening line, (strain hardening modulus), but what about gamma? Or is there any basic mistake in my thinking?
Thanks in advance for any support!
Best regards!

 

[realtime2]

Hello prz235,
Your steel appears to be an HSLa 420, or in the old system an SAE HSLA 960X. Those
steels do not harden or soften much when cyclically strained, and you can probably
find a cyclic stress-strain curve on-line somewhere, but it will be limited in strain.

Using a bending specimen will not help much at all. The behavior is like a bunch of
layers; outside layer extreme fiber and plastic, inside layers less plasticity and even
elastic. What you observe in a bending test is an integration of all these "layers" and one does
not really know which layer is cyclically softening/hardening and by how much.

If you want the reversed properties as in a tension compression straining then you should
really test an axial sample (short gauge length to prevent buckling) with grips that also prevent
buckling. Probably your strains are going to be very large in a forming simulation,
which suggests that you should use a diametral strain extensometer and an hourglass specimen.
You can then convert the diametral strains to longitudinal for use in your model. The specimen
grips will need to be rigid and well aligned; the usual Instron swivel grips will not be
sufficiently rigid.

 

[prz235]

Hello realtime2, thanks for quick and helpful answer.

Even though HSLa420 does not harden much under cyclic loading, I would like to establish hardening type (kinematic or isotropic or mixed) for only one backstress - to know the real spring back after stamping process.

When it comes to testing - unfortunately there is no option to test specimens for tensile/compression loads. Most of tooling we have are the ones to perform bending tests - for 3 point bending and for "fixed cantiliver" bending That's why I wanted to establish initial hardening parameters, to have them "reverse engineered" from Isight bendning test.

I hope I spoke more clearly than last time

Thanks for your support and I hope to have it once again,
Best regards!

 

[prz235]

I would like to establish hardening type (kinematic or isotropic or mixed) for only one backstress

- sorry, now I see my mistake. Point of this sentence was to indicate, that our U-shaped part works in really narrow range of strains - so there is no need for performing analysis for each 1% of peeq (that's what I assume, maybe wrong?). I hope now it's more clear.

 

[realtime2]

The problem will be that the first half cycle will be different than the
following half cycles. I don't have a figure on hand for HSLA 420 but the
same holds true for most materials. You can find some examples in
J.F.Martin, "Fatigue Damage Analysis for Irregular Shaped Structures
Subjected to Representative Loads," Univ. of Illinois Fracture Control
Program FCP Report No. 10 Dec. 1973
(I think this report is on-line now, otherwise it will be difficult to get)

The yield surface rules that apply to these situations are mostly
described by Kinematic "Hardening". It is similar to Masing's hypothesis
and if applied along a line boils down to Rainflow cycle counting.
G.Masing, Proc. of 2nd Intern. Congress of Applied Mechanics, Zurich 1926.

I guess what I would try is to model the monotonic tensile curve for the
first half-cycle, and then switch to the doubled cyclic stress-strain
curve for the unloading and spring-back estimate. -not sure if your FEA
program can do this. Best solution of course, would be to do a test using
and axial specimen subjected to similar strains as expected.

Try the paper by P.Watson and T.H. Topper, "An Evaluation of the Fatigue
Performance of Automotive Steels," SAE paper 710597, June 1971.
It has the initial portion of a monotonic curve for HSLA 950 and a
cyclic stress-strain curve. Make sure that at least your material's tensile curve
is similar in shape.