How to obtain material properties of plastics for non-linear analysis?

[bongirs]

Hello,

I am a design engineer and want to start a trend of using FEA during the design process in my group.
As of now, for some months I have been doing push button analysis using Creo Simulate for linear behaviour.
It has been a progressive journey from "relative" comparison of concepts to getting "absolute" values for a concept.
Now I want to move on to non-linear analysis.
I am facing difficulties in getting the material properties for plastics (mostly PBT GF20). Here are my questions:

1. In the material data sheet from the vendors are the stress and strain values, engineering values or true values?
2. Are these values for the raw material or for the product obtained after injection moulding? i.e. should I use these values as it is or apply some safety factor?
3. Usually only UTS, elongation at break and Young's modulus are mentioned. How do I find out Poisson's ratio, Yield strength, etc.
4. Are the other mechanical properties mentioned in the data sheet (like Charpy impact, ball indentation, etc.) useful to me? If yes how?
5. What are the standard steps taken by an FEA analyst after he receives a project and has no previous experience in it? (i.e. no ready made material library, etc.)

Forgive me if I am asking very basic doubts as I am not a FEA specialist but would like to use this tool to create good designs.

 

[rickfischer51]

1. In my experience, the data sheet values are usually engineering.
2. They are taken after injection molding in that the test specimens are usually molded. But this molding is usually done with ideal molding conditions. If some aspect of your part geometry or tool design requires unusual molding parameters, it would be best to get specimens molded with those parameters.
3. Yield strength is usually given, but it's definition varies depending on the response of the material. Yield strength can be an offset value or the stress at the maximum load. See ASTM D-638. Poissons ratio is typically between .32 and .45 for solid polymers. this is not always reported. Google it.
4. Generally, no. Impact values are relative values and do not usually relate to FE results. Ball indent is a measure of harness, which is a function of yield strength.
5. I usually make a linear elastic run just to see where the stresses are high and get an idea of my strain range. Then I add nonlinear material if needed. Check your calculated strains and make sure they areless than any limiting assumptions you made when getting your material data. The real challange with analysis of polymers is including visco elastic effects, which vary with temperature and strain rate. You might want to get a copy of Structural Analysis of Thermoplastic Components by Trantina and Nimmer and give it a look.

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[bongirs]

Hello Rick,

Thank you for clarifying some of my doubts.
I have some more

1. I believe I should use true values for my FEA. But I may be wrong in my assumption. Am I correct?

2. Should I take these values as it is for the analysis or should I apply some FOS (Factor of Safety)?

3. As I do not have exact properties of materials, I should apply some FOS for the results. But how do I decide how much FOS I should apply?

4. For determining whether the component is breaking (crack development) or not should I use the breaking strain as the parameter or breaking stress?

5. Does the FEA solver compute the stress on the elements and then determine the strain using the stress values or is it the other way round? This will help me in decision making for my previous doubt.

6. Consider material such as PBT. We generally do not use pure PBT but rather glass fiber filled. So PBT GF20 or PBT GF30 is what we use most of the times. What considerations should I take due to the glass fiber for my analysis?

 

[rickfischer51]

1. If you are running a plasticity model, then you should use true stress/strain data, but keep in mind that the difference is small for small strain, typically less than 5%. Also, the usual conversion is only valid below necking.
2, 3. You have to judge this on a case by case basis, depending on the quality of your material data, boundary conditions, geometric model, modeling assumptions and failure mode. Also, what are the consequences of a failure? Young's modulus data is usually pretty good, so if you are doing a linear elastic analysis to check the force required for a snap fit for a toy, FOS could be low. If you are doing a highly nonlinear analysis of a composite hydrogen fuel tank to determine a rupture load, you need either very, very good material data or a high FOS. Experience and past history plays a big part in this.
4. It depends on the response of the material. Some use strain. Do not use published elongation values. There is too much scatter in that data. This is highly dependent on process and practice of the molder. You may want to tensile test samples molded by your supplier, or even cut from your actual production parts, if that is possible. With metals, the uniform strain is sometimes used after modification with the triaxiality factor, but I'm not sure that it is appropriate for polymers. You don't tell us what you are designing or the nature of the loads. Sustained loads often lead to creep rupture, with cracks initiating at flaws and growing slowly until a critical crack size is reached. It's probably best to use a fracture mechanics approach. See Principles of Polymer Engineering by McCrum, Buckley and Bucknall, or Mechanical Properties of Solid Polymers by Ward and Sweeney. The McCrum book is an easier read.
5. Its the other way around.
6. The stiffness and strength of fiber reinforced polymers is of course highly dependent on the direction of the fiber orientation. Fibers generally orient in the direction of the flow of the melt in the mold at the surface of the part and transverse to the flow direction in the center of the part. Ideally, you would have a mold filling simulation performed with a fiber orientation prediction that could be exported to your FEA program. In the FE model, you would need an anisotropic material model set up with good data. The model would likely be a midplane model with shell elements with multiple layers. Obviously, you need a properly designed part to make this work. A part that has been designed without draft, parting line and gate position is not a design, it is a cartoon. A mold filling simulation taken through fill and pack should be done early in the design cycle, whether you intend to do FEA or not. Ideally, however, usually doesn't happen, so you usually end up with an approximation. I would start with an existing part, analyze it and compare the results to test data. Fudge material properties until you are comfortable with the results. And adjust your FOS accordingly.




Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[bongirs]

Hello Rick,

Thanks for clarifying so many of my doubts in such detail!
Your advice is very helpful and I appreciate you for taking the time to write the answers!

For PBT GF20 the breakage strain is around 2.5 to 3.5% (depends on vendor). So by using the conversion formulae the true values are pretty much reliable.

Regarding FOS I understand that a lot depends on the criticality of the component.
But I meant to ask in the sense that on which parameter I should apply FOS.
For example, in one analysis a snap lock was breaking at 100N.
Due to a lot of assumptions in material data, I decided to give FOS as 2 and concluded that the lock will break at 50N.
I am not satisfied with this as I cannot justify why I chose 2 and not some other value.
Maybe this will come with experience as you said.

The loads are mostly spasmodic and not at all fatigue.
Mostly I try to analyse the behaviour of locks (various designs) for a variety of situations.
I tried to plot a graph of stress strain from the vendor's data and the curve was not at all brittle in nature.
To be precise, I compared the data from 2 vendors for the same material.
I concluded that relatively the material of one vendor was more brittle.
But in actual parts it was quite the opposite.
Also both materials absorb same amount of moisture.
I think the best practise would be to calibrate the material properties by actual testing of parts.

Our tool design team does the simulation using Autodesk Moldflow.
I will get the required inputs from them for setting up the anisotropic properties.

Thanks again for suggesting the books.
I will definitely take a look at them.