Thermoplastic FEA with ANSYS, flexural test correlation

[aarriaga]

Hello to everybody, I am trying to correlate flexural tests in ABS and PP20%Talc (rectangular specimens) by means of simulating the same test in ANSYS feeded with uniaxial tensile data. Both flexural and tensile tests are performed at a speed of 10 mm/min. I have seen that uniaxial test data rigidity is very low compared to flexural modulus and so when I compare force-displacement results of the simulated flexural test and the experimental one, the last one can go a 20% up from the simulation curve. What does it mean, uniaxial data is not valid for simulating plastic components (most of them subjected to flexural loads)? Am I doing any error maybe in tensile stress-strain transformation (assuming incompresibility in plastic region)? Does it have any sense to perform uniaxial compression tests for FEA data generation with plastics? Including friction does not improve results. Thank you in advance!

[Jorgen]

Hello aarriaga:

Let's see if I understand:
1) You performed a uniaxial tensile test on a polymer (either ABS or particle filled PP)
2) You fit a constitutive model to the uniaxial test data
3) You then simulated a flexural (bending) test using ANSYS
4) Finally, when you compared the results from the flexural simulation you got different results than direct experimental data from a flexural test.

Is my understanding correct? If so, I have a few questions/comments:

* How different were the simulated results from the experimental results in the flexural tests? From your message I am guessing it was 20%.
* The response of polymers is strain-rate dependent, and an applied deformation rate of 10 mm/min in uniaxial tension and 10 mm/min in flexural loading are not imposing the same strain rate in the sample.
* What material model did you use to represent the polymer behavior? If you used a traditional elastic-plastic model then you should not be suprised to get an error of 20% in your simulations. There are other and better models for polymers that can give higher accuracy.
* Uniaxial data is perfectly OK for calibrating material models for plastic components. In fact, it is typically better to use uniaxial data than flexural data since uniaxial data is easier to interpret.
* Both uniaxial tension and compression experiments are useful for plastics. In compression, friction can sometimes be a complicating factor, and in tension necking can be a complicating factor. Both of these factors can be addressed.

Best of luck,
Jorgen

[aarriaga]

Hello Jorgen and thank you for your reply!

Yes, you have understood well everything from my message:

* The difference of a 20% is in the peak force of the curve but the main problem was in the difference in rigidities; flexural modulus doubles the tensile one.

* I have seen that the strain rate does not influence excesively the stress-strain behaviour at speeds up to 100mm/min, and also I have tried to introduce in ANSYS a strain rate-curve similar to the rate that is reached in the flexural test. Anyway, the results were not much better.

* I have used a traditional elasto-plastic model in ANSYS (Multilinear elastic, MISO, with Von Mises yielding criteria). Which model would you use when working with plastics?

Thanks again!

Aitor Arriaga

[Jorgen]

Hello Aitor,

A few comments:

:arrow: The Young's modulus and the Flexural modulus are both related to the stiffness of the material, but they have different names for a reason: they test the stiffness behavior of the material in different ways. They are likely to have similar magnitude but they don't have to be the same, as you noticed in your experiments.

:arrow: The best material model for to used in FE software to simulate plastics will depend on the type of plastic, the type of loading that is applied (small strain, large strain), and the desired accuracy of the model predictions.

- Jorgen