Nylon is a commonly used polymer in industrial applications due to its good mechanical properties and high strength when exposed to heat and common chemical exposures. Nylon is used both as fibers and as a traditional resin. In this article I will discuss one particular type of Nylon called PA (polyamide) 66. The goals of this study is to examine 9 different candidate material models for PA66, in an effort to determine which material model is the most accurate at predicting the large-strain viscoplastic response of the material.
Experimental Data Used for the PA66 Material Model Study
In this study I used uniaxial tension data at different strain rates. The data that I used comes with MCalibration, so you can easily reproduce my study if you are interested. The following sections compares the experimental data to predictions from 9 different material models, in order from worst to best!
Figure 1. Experimental data for PA66 used in the study.
Results #9: Ansys MISO Plasticity with Creep
Results #8: Johnson-Cook Plasticity
Comparison between the experimental data and predictions from the (Abaqus) Johnson-Cook plasticity model. The model predictions look pretty much the same as for the MISO plasticity model. The reason for that is that they are both isotropic plasticity models. The results are still not good enough.
Results #7: Bergstrom-Boyce
The Bergstrom-Boyce (BB) model is a cool model – for elastomers. PA66 is a thermoplastic (at room temperature), so it is not a surprise that this model does not work better in this case. The average error in this case is 15.6%, which significantly better than the plasticity models. I still don’t recommend the BB model in this case.
Results #6: Abaqus Elastic-Plastic with Combined Hardening
This model is sometimes called the Chaboche model and the MCalibration implementation is using 5 back stress networks. The Abaqus commands are:
*Elastic *Plastic, hardening=combined, data type=parameters, number backstresses=5
Since the model is based on kinematic hardening plasticity the unloading predictions are significantly better than for isotropic hardening plasticity models. In this case the average error in the model predictions is 14.6%.
Results #5: Abaqus PRF with 2 Networks
The Abaqus Parallel Rheological Framework (PRF) model can be accurate for some polymers, but a 2 network PRF model is not sufficient for a glassy thermoplastic material. In this case the average error of the model predictions is 13.3%.
Results #4: Abaqus PRF with 3 Networks
Switching to a 3-network PRF model is usually a much better for thermoplastics. In this case the average error of a PRF model with Yeoh hyperelasticity and Powerlaw flow is 11.3%.
Results #3: Abaqus PRF with 4 Networks
I rarely use a PRF model with 4 or more networks. In this case a 4 network PRF model (with Yeoh networks and Power flow) has an error of 10.7%. This is slightly better than the 3 network version, and perhaps in some cases could be worth the trouble.
Results #2: PolyUMod TN Model
The PolyUMod TN model (which in Ansys is called TNM) was developed for thermoplastics and is usually very good at predicting the large strain response of this class of materials. In this case the average error in the model predictions is 9.4%. Not too bad! But we can still do better!