Ansys FEA modeling of 15% glass reinforced ABS plastic

[Court]

Hello,

I am attempting to model a motor/gear housing composed of a 15% glass filled, ABS polymer. I am using Ansys v8 and am wondering what material model would be recommended. The ambient temperature will be approximately 26°C, and the loading is known and is a pulsating tension. I am concerned about fatigue failures occuring due to the cyclical loading with mean stress present. Customers have experienced failures along the anchoring points of the motor housing. Polymers are relatively new to our team, as we usually deal with 6061 series aluminum alloys, and use a stochiastic strain life method with good results. However, as plastics become more durable and cost effective, we are finding that plastic parts are helping us stay competitive in an industry that is slow and hesitant to react to change (avionics). Any help or guidance would be greatly appreciated. :D

Courtney

[Jorgen]

Hello Courtney,

You are asking great questions. In general, the mechanical response of glassy polymers (like ABS) is quite different than for metals. In fact, predicting the response of polymers subjected to mechanical loads is often more difficult due to the non-linear response and strong dependence on deformation rate, temperature, ageing time, etc. The situation is further made complicated due to the lack of general tools for predicting the material response. Fortunately, a lot of progress has been made the last few years, and by performing a careful analysis, it is now possible to predict most of the experimentally observed characteristics.

Your first question is related to what material model is most appropriate for a glass filled ABS. The answer to that is that it depends on your specific application. How large strains are you expecting? If the strains are small you might be able to use a simple viscoelastic model, if the strains are intermediate or large you will need a more advanced viscoplastic model. Note that there are great models available today for predicting the general deformation response of ABS, including large deformations. Unfortunately, these more advanced models are currently not built-in into ANSYS, but need to be provided by an external subroutine.

You also mention that you are concerned with fatigue failures. There are different approaches to predict the fatigue behavior of polymers. Some of the phenomenological low- and high-cycle fatigue models that are used for metals have also been used for polymer, sometimes even sucessfully. There are also new interesting damage models being developed. For example, I am currently working on integrating a damage concept into an advanced constitive model for glassy polymers. By integrating the damage model into the constitutive model it is possible to accurately track the evolution of damage in the material as a function of applied load history, and also create contour graphs, etc, of the damage state in general multiaxial deformation histories.

Jorgen

[Court]

Dr. Bergstrom,

I appreciate your response. I would expect that strains would be less than 5% in our case. This is an educated guess of course, as we have not performed a strain gauge study. As you are well aware, a significant portion of the challenge in calculating nodal states of stress lies in simply finding material property values based on the choice of modeling. Unfortunately, our management will not outsource our polymer material to a reputable testing facility that handles polymers (like axel) due to limitations in budget. This leaves me to either:


A. Find material properties of a similar ABS material and allow a liberal margin of safety to absorb the errors. This approach would be undesirable but probably the most feasible.

B. Make a case for having the injection molding vendor send the material off to a material testing facility specializing in polymer materials on his nickel. Unlikely that this would fly, however, this would be the ideal situation. However, I would need to heed to your advice on what model is most appropriate for the ABS material.

I am guessing that hysteretic heating (significant temperature rise) is unlikely due to the incredibly slow frequency of the cyclical loads, so an isothermal body is probably a reasonable assumption. The strain rate is less than 0.1%/minute in this case. Any suggestions you could give in terms of sources of information for polymer material properties would be immensely appreciated. Without consideration of a very advanced fatigue model, based on your breadth of experience is there a (thumb in the air) % value of some characteristic point in the stress-strain curve of the material that would indicate a good target upper bound to stay under in order to avoid high/intermediate-cycle fatigue failures?


I know a human tendency (myself usually) is to quickly assume a model based on limited help data in an FEA program, quickly mesh an FEA model and hope for the best. I feel that this is like driving at 110 mph during an ice storm and hoping you won't loose control and crash. I truly would like to understand the topic better, so if you could recommend one good hardcover book as a starting point to better understand polymer behavior, I would be most interested. For instance, if someone were to ask, "Should I consider a linear elastic, non-linear elastic, non-linear plastic, viscoelastic, or hyperelastic model?" I would expect the general answer to be "It depends on the material, loading conditions, temperature, strain rate, etc."


Finally, I would like to mention that I was thrilled to see that finally someone has stepped up to the plate and given numerical models of polymer materials due dillegence. Industry in general certainly needs a better understanding of polymers and I extend my thanks to you for creating a convenient and friendly hub to discuss relevant topics in polymers.

Sincerely,
Courtney McCauley

[Jorgen]

Hello Courney,

It is good to hear that the strains in your case are only 5% or so, that will simply the analysis. In order to calibrate any material model you will need experimental data. It sounds like this might be tricky for you. Unfortunately I don't think I have any data that I can share with you for this type of ABS, but it should be possible to find data for similar materials in the literature.

I agree with you that hysteretic heating will not likely be an issue. However, it is going to be difficult to perform an accurate fatigue prediction without an accurate material characterization. Acceptable strain/stress levels for long fatigue life will also depend on how many cycles that your product will need to endure during its intended lifetime. There are some rules of thumb available, for example, the stress/strain should be lower than the value at yield in order to avoid rapid fatigue. There are unfortunately not many good books about this type of polymer modeling, one of the best references that I have found is Ward's book "An introduction to the mechanical properties of solid polymers". I have a strong interest in writing a book on exactly this subject, but unfortunately I don't have the time for now...

Jorgen

[Court]

Dr. Bergstrom,

Thanks for your response. Lets say for the moment, even though not appropriate, you were to pick a material as close to ABS as possible that you would have valid material parameters for. Considering small strains, and low strain rates, could you point me in the direction of a visccoelastic model with a few parameters to get me started. At the very least, I could gather an idea of relative changes in Von-mises/first principle stresses as a function of geometric changes in the structure. I am looking into purchasing the book you have suggested.


Thanks,
Courtney McCauley

[Jorgen]

Courtney,

I don't know who's your supplier of ABS, but GE Plastics has on their website (http://www.geplastics.com/resins/designsolution/tools/) rather nice data for their different types of ABS. The data that is available include stress-strain curves, and fatigue data.

Regarding models, I propose that you use a simple linear viscoelastic model.
Note that this model is not suitable if your strains are as large as 5%, since that is likely larger than the yield strain.

Jorgen

[Court]

This was exactly what I was looking for. Thanks so much!!!