Material model for plastics is ANSYS 1

[assafwei]

Hi,

The company I work for migrated to ANSYS lately.
I was using FEMAP as my pre/post with the NX advanced non linear code (ADINA) for implicit non linear analysis.
Our products are mainly made from polymers (POM, PEHD, and PA with and without GF) and I was using a simple plasitc stress-strain curve non linear material (data was based on tension test of injected dog bone samples). Althoug I know this is not the best model for analysis of plastic parts, this is waht I had at the time (from the point of view of material models in my FEA package, and from the point of view of the material testing data I had).
Now that we have migrated to ANSYS, I was wondering what is the best meatrial model ANSYS has to offer for such analysis types (I am talking mostly about nonlinear contact, nonlinear materials and geomteric nonlinearities such as large deflections and strains).
The main issue I am using the FEA tool is for validating assemblies (snaps, and undercuts) in plastics, failure of connections when for example inserting a tube on a nipple and so on.
For the purpose of the discussion lets say I have to scenarios:

1. I will have a talk with the R&D manager - and he will tell me no matter how much it will cost, do all the tests you need to get the best material model for all the resins we are using
2. I will talk with the R&D manager - and he will tell me - look, we just invested in this new software, use the equipment that you have (tensile test machine) and do your best without any work done by external labs.

What path should I take, what tests should I do, and what material models should I use for any of the above?

Thanks.

 

[rickfischer51]

First, most FE codes have the same basic core of material models, and then each code adds or doesn't add additional models based on their willingness to clutter their software with additional models that will see limited use because most users don't understand them and wont use them because they require specialized material data that isn't on the data sheet and is difficult to test for and expensive to buy. There is no magic answer in Ansys.

The thing to keep in mind when analyzing polymers is that they are visco elastic/plastic. Mechanical response changes with duration of the load and the rate of loading. Plastic flow takes place due to conformational changes in the molecular structure, via various mechanisms (cooperative motion, bond transfer, etc). These mechanisms are sensitive to changes in free volume, so temperature effects are huge. Accounting for all of this in a single material model is problematic.

Another issue is that the material in the finished part may exhibit anisotropic behavior. This is especially true for injection molded parts, and really true for glass reinforced parts. Orientation is different at the surface and center of the section, there are residual stresses from differential pressure, differential cooling, etc. If your polymer is semicrystaline (PA), you can have different properties in different locations on the same part due to variations in cooling caused by thick sections, hot spots in the mold, etc. You PA is readily plasticized by moisture, so predicting Youngs modulus, elongation, impact properties, etc at any point in time is a crap-shoot. Then you have the guy molding the parts. He can change your material response with the twist of a knob. What all this means is that there are so many unknowns that trying to put a fine point on a material model is wasted effort, and your benefit will likely be lost in the noise. Accounting for all this is like hearding cats. Mold filling simulations can predict some of this, and this data can be exported for use in a mechanical simulation.

In my mind, the issue is dependent on where you are in the design cycle. An 80% solution with some chicken factor applied early in design is worth way more than a 100% solution too late to be be useful. Can the designers wait for your analysis while you generate material data? Mine usually cant.

Ansys has a visco model that uses a Prony series. I think in theory this model can provide a pretty complete visco response. But, data for the Prony series usually isnt available, so you have to generate it yourself or pay a lab to do it. I think you can get useable results by running running BISO or MISO plasticity with a creep model. Check the manual for allowable material combinations. Also, visco elastic theory allows for curve shifting. You can take a material curve and kludge it for temperature and strain effects. See Chapter 3 by Shiro Matsuoka in Failure of Plastics by Browstow and Cornelius. Another good read is Structural Anslysis of Thermoplastic components by Trantina and Nimmer.

You haven't told us how your parts are made, how critical is their performance or how tragic their failure, or the size and budget of your employer. Obviously this has to be considered. If you repeatedly use just a few resins, an investment in testing may be justified.

There was an article in one of the trade magazines (Design News or Machine Design) back in the 90's expounding the use of the extended Drucker-Prager model to predict yield in polymers. The author (Peter Tuschak?) tested Delrin rings and analyzed them with Abacus. Peter was a DuPont employee. So I called the guy at Dupont that did FEA work for customers and asked how often he used extended DP and he said never. No one wanted to wait for the material testing.



Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[assafwei]

Thanks for the reply.
Lets put things in order...
First lets take out some of the noise:
Lets put aside creep and temperature dependancy (this will come in the far future for me...)
We will also put aside fiber reinforced materials (they are a minority in our protfolio).

For now lets consider the following:
materials in use - PEHD and POM (about two resin manufacturers from each).
Injection orientation - I am doing a MoldFlow analysis for every mold we order before the mold is ordered, so I know what is the general flow behavior and I can try and take it into considaration (orientation, weld lines tec). Maybe in the future I will incorporate the MoldFlow data into the ANSYS analysis. Furthermore we have very strict design rules and thick sections are largely prohibited.
What I do need to take into consideration in the analysis is different strain rates (we are taliing about assemblies so slow assemly speed versus fast etc.). Since I am not familiar with the model I am wondering if this is a model fit for this purpose?

Regarding failure - failure is most of the times not a safety issue, but mostly an economical issues that costs alot to handle (car industry, recalls, mold modifications - all these are very costly).

Thanks.

 

[assafwei]

I re-read my last post - regarding the model I am talking about the visco elastic / visco plastic models

Thanks.

 

[rickfischer51]

I don't have any experience with the visco models. I think the Prony series data is best derived from DMA. See the attached.

Any time I had to analyze a plastic part, there was no time or budget for testing. I used the curve shifting techniques from Natsuoka to adjust for strain rate effects. The problem here is that the entire model gets treated as if it is at a high strain rate. I wrote a macro that assigned different materials to different areas of the part based on a prediction of the strain rate. A real kludge job, but all I could do with the restrictions I had. Ansys now has Perzyna and Pierce Rate Dependent Plasticity. See 3.5.1 in the Material Reference. It might be possible to generate stress strain curves at different strain rates using the Matsuoka reference, then fit that data to get viscosity and strain rate parameters. This is somewhat less kludgy than what I was doing before. If your inhouse tensile testing capability can produce high strain rates, you can skip the curve shifting and fit the constants to real data.

Rick Fischer
Principal Engineer
Argonne National Laboratory