Monday  |  Milwaukee  |  11:00 AM–11:20 AM

#19148, Generalized Constitutive Models for Hyperelastic Materials Using System Identification

Computational studies of solids and structures synthesized from hyperelastic materials are often a formidable challenge, requiring well-calibrated constitutive models ascertained from experimental data. The general procedure for developing the phenomenological material models is to first test the material under uniaxial, pure shear, and biaxial loading conditions and then to curve-fit a hyperelastic material model to the experimental data. We propose a computational framework to identify hyperelastic models with a nearly arbitrary mathematical structure using Sparse Identification of Nonlinear Dynamical Systems (SINDy). The key observation enabling the approach is that the strain-energy functions are characterized by a smooth, strictly monotonically increasing response that does not necessitate searching for hybrid models. As a demonstrative example, we tested a silicone rubber, under all three loading conditions. Considering experimental data as a dynamical system response, we select candidate functions and search the space of functions for which candidates most actively contribute to the observed material response. The goodness-of-fit with test data and generalization over an extended strain range is analyzed and compared to well-established hyperelastic modeling approaches, including Neo-Hookean, Mooney-Rivlin, Ogden, and Yeoh. It is shown that the proposed methodology can be used to describe the material data inside the characterized strain ranges and over extended strain ranges without the limiting assumptions of conventional hyperelasticity models.

Nicholas Pagliocca   Rowan University
Mitja Trkov   Rowan University
George Youssef   San Diego State University
Behrad Koohbor   Rowan University