Concrete Softening at higher strain rates
Moritz Hupfauf
I am comparing the KCC model and the RHT model for their ability to simulate the concrete behavior from a contact detonation. For this purpose, I have done some single element simulations to find the influence of the Dynamic Increase Factor (DIF) on the softening behavior and consequently the fracture energy of the concrete. The DIF is the same for the KCC model and the RHT model, but with a maximum factor of 10 for the KCC model. For the RHT model, I control the fracture energy with the epsmin parameter, and for the KCC model, I use a modified description of the lambda-eta curve.
The problem now is that for strain rates higher than 10^2 the softening curve shows some numerical problems. The only parameter that seems to affect this is the time scale factor tsfac. Attached is a plot of the tensile stress-strain relationship for both material models. The red lines show the shape of the curves I would expect.
Has anyone else encountered this problem and perhaps found a solution?
Best regards,
Moritz
l...@schwer.net
The use of single elements for strain rate sensitivity studies is less than ideal for frictional materials, e.g. concrete.
The problem arises due to the mass at the nodes and their delayed lateral response. Although the nodes at the top are prescribed to move, say vertically, and those on the bottom constrained, the lateral motion of these nodes is dictated by the Poisson induced stress and their inertia. The delayed lateral motion of these nodes causes additional pressure in compressive strain rate and lateral tensile-stress in tensile strain rates. Both of these cause the stress trajectory to deviate from that anticipated analytically, and hence deviation from expected responses.
This manuscript attempts to explain this phenomena
https://www.dynalook.com/conferences/copy_of_european-conf-2009/M-I-04.pdf/view
Also, tensile DIF is an illusion. The physics needed is fracture mechanics and not a simple engineering DIF model.
--len
--
You received this message because you are subscribed to the Google Groups "LS-DYNA2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ls-dyna2+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ls-dyna2/8c4669b9-ce75-4f6a-b460-bdf1029624bcn%40googlegroups.com.
Moritz Hupfauf
As far as I understand it, this is not only a problem for single elements, but also for whole structures subjected to high strain rates?
I agree that proper modeling of individual cracks based on fracture mechanics would probably give better results. But as far as I know, there is currently no option in LS-Dyna that can do this in the context of contact detonations and concrete.
So I think the RHT model and the KCC model are my best options at the moment. But especially the KCC model runs into numerical problems in the spalling region of the concrete opposite the contact charge, resulting in blowups of individual elements. Also, the velocity of the concrete does not agree with experimental measurements. I think this could very well be related to the behavior at higher strain rates.
Or do you have a better suggestion on how to avoid or work around this problem?
l...@schwer.net
Hello Moritz –
You are correct, simulating strain rate effects numerically, even with larger cylindrical sample sizes, is problematic. The same problem exist experimentally, that is once the strain rate exceeds a certain level, it is no longer a material characterization test, but a wave propagation test. To be a material characterization test, the tested sample must be in a uniform state of stress/strain, i.e. no wave propagation. NOTE: this leaves out most split Hopkin’s bar (SHB) experimental results.
My preference is to omit numerical strain rate effects. The frictional nature of the concrete shear failure surface “automatically” includes some strain rate effects; experts attribute a few percent of observed strain rate effect to the water remaining in the concrete.
My other reasons for omitting strain rate effects in concrete modeling, is there is no reason to think whatever rate effect there maybe are not stress path dependent. Most all the strain rate data (a) is not a material characterization and (b) either uniaxial stress or (c) uniaxial strain (most high rate data, e.g. SHB). Since it is unlikely any practical numerical simulation will involve either purely uniaxial stress or uniaxial strain, there is no data to characterize arbitrary stress paths.
So rather than added “garbage input” strain rate data, I argue it should be omitted; there is also the problem that adding numerical strain rate effects is additive to the nominal strain rate effects provided by the frictional material.
The KCC model requires an erosion criterion and associated value, e.g. *MAT_ADD_EROSION; again not a material characteristic, but an ad hoc criterion and nonunique numerical value artifact.
The RHT model has a build in erosion criterion of effective plastic strain with a default erosion value of 200%. Effective plastic strain is the worst possible erosion criterion, and 200% is an arbitrary number. The problem with effective plastic strain, as LS-DYNA defines it, is a positive quantity integrated over time. That is, when any (small) amount of plastic strain occurs, over time that amount will increase and eventually can reach the arbitrary erosion value.
I have recently been conducting some contact charge simulations using the RHT model with *MAT_ADD_EROSION using the maximum shear strain criterion; a value of EPSSH=0.6 seems to be reasonable with the RHT model EPSF=5 to keep it from eroding any elements. My rational is most materials, especially concrete, fail in shear and since the shear stress is limited by the shear failure surface, then shear strain is a good candidate for erosion criterion.
To view this discussion on the web visit https://groups.google.com/d/msgid/ls-dyna2/5e10ca12-910d-4e58-8858-483c9e877da3n%40googlegroups.com.
Moritz Hupfauf
My problem with unsing erosion in contact detonations is that a lot of the energy from the explosive is lost in the erosion, which greatly affects the results. This is especially relevant for my simulations, since I am not only comparing the damage to the concrete structure, but also the resulting velocity of the spalled concrete.
l...@schwer.net
Hello Moritz –
1\ I have never modeled concrete at the microscale. Since both KCC and RHT use a homogeneous representation of concrete what may or may not happen at the microscale is not relevant to strain rate effects when using these concrete models.
2\ Have you plotted the eroded initial and kinetic energies and compared them to the initial internal energy of the explosive – quantify “a lot of the energy from the explosive is lost in the erosion.”
3\ What erosion criterion and value are you using for KCC model?
4\ My guess is your lack of agreement with experimental results has more to do with your choice of concrete material model and erosion criterion with associated value, than strain rate effects.
To view this discussion on the web visit https://groups.google.com/d/msgid/ls-dyna2/3e58ed2c-3b6a-4e03-9c8d-cdccdf20a224n%40googlegroups.com.