Clear all

Hyperelastic Model Test Data Calibration Critrion

3 Posts
2 Users
Hossein Sahhaf
Posts: 5
Topic starter
Active Member
Joined: 3 years ago

Hello Dr. Jorgen

I'm happy to find your forum that I can ask my questions about polymer materials. I'm new in this field and I have many questions in different aspect of Hyperelastic polymer material simulation. I asked a few questions past, and now I have 2 new questions.

In the process of calibrating a hyperelastic model to experiment test data (e.g. uni-axial test data of Polyurethane) which criteria must be paid attention to get the best fit model? stability, less error in fitting (e.g. Root mean square of fitting) or anything else? which one has priority?

When is a material model unstable and how does it effect on simulation?

Thanks in advance.

Best Regards, Hossein Sahhaf.

2 Replies
Posts: 21
Eminent Member
Joined: 4 years ago


It is generally the analyst's responsibility to decide what is most important to achieve the best model. In the context of hyperelastic models, stability usually refers to the Drucker stability criterion which, simplistically, states that positive energy is required to strain (deform) a material. This is often checked by ensuring positive slopes in true stress - strain curves. From a modeling perspective, it is important to ensure that your hyperelastic model meets this criterion in different loading modes (for instance, equi-biaxial extension, planar tension, and uniaxial tension) in the range of strain that is being modeled. So, an instability at, say 5%, strain is generally unacceptable, but one at, say, 100% strain, may be acceptable if the component being modeled has strains less than 10%. When a material model is unstable and you deform a component through that instability, you will get non-physical behavior which can also cause nonconvergence (often due to excessive element distortion).

Regarding the fitting error, there are different ways of calculating that, and whichever you choose is not so important as long as you use the same metric for comparing different hyperelastic models. Bearing in mind that the stress-strain response of hyperelastic materials is loading mode dependent, you want to bias (minimize) fitting errors to the the most important loading mode(s) for your component. That is, if you have a component that is predominantly loaded in compression, minimize model fit errors in that loading mode.

As you'll see, the answer to both of your questions touches on the dominant loading mode(s) in your component. If you do not know the dominant loading mode, you can run a simple FE analysis with a Neo-Hookean material model to figure that out with a biaxiality analysis. The concept is explained here:

There are different biaxiality definitions you can use, and my personal favorite is:

Biaxiality = log(lamda_min)/log(lamda_max)

because it bounds all biaxiality values between -0.5 (uniaxial tension) and -2 (equibiaxial tension). Planar tension (pure shear) falls in between these two with a value of -1.


Posts: 21
Eminent Member
Joined: 4 years ago

In case that link I shared ever goes away, the source is:

Wadham-Gagnon, M., Hubert, P., Semler, C., Paidoussis, M.P., Vezina, M., and Lavoie, D. (2006). Hyperelastic modeling of rubber in commercial finite element software (ANSYS). Proceedings of the SAMPE '06: Creating New Opportunities for the World Economy.

You can also find this discussed in the Master's thesis of Wadham-Gagnon, also available online.