A D V A N C E D M A T E R I A L S & P R O C E S S E S | O C T O B E R 2 0 2 2 3 8 QUANTIFYINGAND PROPAGATINGUNCERTAINTY IN SUPERELASTICITY SIMULATION INPUTS Digital image correlation data and Bayesian inference used together facilitate rigorous quantification of the uncertainty in material input parameters for finite element simulations of superelastic deformation. Harshad M. Paranjape,* Confluent Medical Technologies Inc., Fremont, California Kenneth I. Aycock, Jason D. Weaver, and Brent A. Craven, U.S. Food and Drug Administration, Center for Devices and Radiological Health, Office of Science and Engineering Laboratories, Division of Applied Mechanics, Silver Spring, Maryland Craig Bonsignore,* First Article Services LLC, Phoenix, Arizona Thomas W. Duerig, FASM,* Starlight Cardiovascular, San Diego, California Computer simulations using methods such as finite element analysis (FEA) play an important role in the design of implantable medical devices that are manufactured from superelastic materials like nickel-titanium (NiTi) shape memory alloys (SMA). The simulations are typically performed in a specific context of use, for example, durability assessment of a device under a particular anatomical boundary condition. A topic of emerging importance to NiTi simulation is the assessment and reporting of the credibility of a computational model for its context of use[1,2]. This *Member of ASM International credibility assessment generates evidence supporting the use of a computational model for decision making. Moreover, higher model credibility enables medical device manufacturers to use modeling for higher risk and higher impact contexts of use. As part of credibility activities, quantification and propagation of the uncertainty in material parameter inputs increases overall model credibility by providing conservative bounds on the uncertainty in model predictions. A recent work by the authors implemented a method to determine the material parameter inputs and their uncertainty for a computational model of the superelastic deformation of NiTi[3]. The material property determination is colloquially referred to as model calibration. This method for superelastic model calibration is unique in that it has uncertainty quantification built in, it uses full-field surface strain data together with global load data as inputs, and it is able to furnish both tensile and compressive plateau stress material properties from a single test. CALIBRATION FRAMEWORK A flowchart summarizing the material property determination method is shown in Fig. 1. The method essentially has three components: (1) a standard tensile test specimen geometry and a test protocol to obtain the surface strain fields in the test specimen using digital image correlation (DIC); (2) a library of simulations of the tensile test specimen loading protocol using a range of material parameter input values; and (3) a data-processing method using Bayesian Inference (BI) to minimize a cost function based on the local strain and global load measured experimentally and simulated in the simulation library. The calibration scheme is demonstrated on a Ti-50.8 at.%Ni superelastic NiTi sheet material and the superelastic constitutive model implemented in the Abaqus finite element framework[4]. The typical stress-strain response for the FEATURE Fig. 1 — A flowchart summarizing the material property determination method. MAP stands for maximum a posteriori. Reproduced from Paranjape et al.[3]. 1 90 1 0
RkJQdWJsaXNoZXIy MTYyMzk3NQ==