A methodology for the experimental characterisation and computational modelling of Nitinol wires used in stent-graft devices

  • Alexandros BOUKIS

Student thesis: Doctoral Thesis


Stent-grafts are medical devices designed to treat abdominal aortic aneurysms (AAAs). These devices are usually composed of a Nitinol wireframe and a fabric graft. The design of next generation stent-grafts is directed towards a very low-profile when compacted. Such devices can treat narrow access vessels and tortuous anatomies while inducing less trauma to patients. Therefore, Nitinol stents are required to undergo greater deformations during manufacture without compromising the performance and durability of the medical device. Experimental characterisation of the material is required to demonstrate its mechanical behavioural characteristics. Since Nitinol wires are subjected to multi-mode loading conditions, investigation should incorporate all loading modes and thermomechanical conditions that a device encounters during its life cycle. The resulting data can then enable the derivation of the material’s constitutive properties which are necessary for computational analyses. Finite element analysis (FEA) has become an integral part of the design of medical devices and computational analyses reports are used nowadays as scientific evidence to support medical device submissions. The Nitinol constitutive model that is implemented in the finite element software Abaqus is considered the industry standard. However, its capabilities and limitations are not fully explored since there are no experimental data available for this purpose. In the present work, a methodology was developed to characterise medical grade Nitinol wires in tension, compression, bending and torsion. The mechanical behaviour of the material was examined under high-strain tensile deformation. The relevant constitutive parameters were identified from the experimental results and the Abaqus Nitinol model was assessed, including its superelastic-plastic modelling capabilities, by simulating the mechanical tests of all loading modes. The key findings from the present work are summarised below. High straining of the material during compaction results in decreased austenite stiffness, decreased load and unload plateaus and increased residual strain. These features are more pronounced when multiple compaction attempts take place. A recommendation is made, to limit the compaction strain to 10%, if possible. This is because if more than one compaction attempts take place at 10% strain, the effect on the unload plateau, which influences the radial strength of the stent, will be small. In addition, up to 10% compaction strain there is only a small, gradual increase of residual strain. The bending response of the material is load-rate sensitive. Therefore, compaction should take place slowly in order to avoid potential rate effects. Sterilization is not expected to have a negative impact on the stress levels experienced by Nitinol wires during this process, since temperature sensitivity was not observed within the post-transformation region. The Abaqus Nitinol model provided results that were in agreement with the experiments when modelling tension within the superelastic range, as only minor qualitative differences were identified between computational and experimental responses. However, the material model was incapable of forecasting the sensitivity of the unload plateau to the high strain loading. The present work shows that this limitation can be overcome by implementing Fortran subroutines that modify the transformation stresses during unloading as a function of the plastic strain reached during the simulation. Tension-compression asymmetry was also exhibited in the FEA results. The austenite stiffness and the start of the transformation during loading in compression agreed quantitatively with the experimental results, exhibiting negligible errors. However, the unloading transformation stresses in compression were underpredicted by up to 36.6% compared to the experimental data. The flexural stiffness of austenite was also underestimated in the computational bending responses by up to 28.8%, although the transformation during loading started at the same force levels for FEA and experimental results. Results show that the loading path of the bending response can be improved if the flexural modulus of austenite is used as input parameter. However, the unloading path of the bending response was not captured correctly since it took place at lower forces compared to the experimental curves. The computational response in torsion was also not in agreement with the experimental results, as the torsional stiffness was underestimated by 11.2% while the loading and unloading paths took place in lower load levels. The model however, was able to capture the qualitative features of the combined tension-torsion deformation.
Date of Award30 May 2022
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsUniversity of Strathclyde
SupervisorMarcus Wheel (Supervisor) & Andrew McLaren (Supervisor)

Cite this