There are many ways to model Nitinol using Finite Element Analysis. Over the years we have studied them all, from rigorous analytical implementations based on micromechanical descriptions to phenomenological models based on thermodynamics. As with all simulations, the level of detail and approach taken to implement the constitutive behavior of a material is driven by the specific purpose of the analysis and the available information.
The picture above illustrates a particularly creative approach that was developed by my very good friend and colleague Paul Labossiere. In this work, he utilized our thermodynamic user subroutine (UMAT) for ABAQUS but implemented it with a very clever script that generated finite elements with random properties for our UMAT. With this model, we were able to explore numerous aspects of modeling the constitutive properties of Nitinol and the limitations of different assumptions.
The image above compares a continuum model with a polycrystalline model. All of our analysis results were compared to experimental measurements and as a result of this research we were able to definitively identify the critical attributes of modeling superelastic Nitinol.