| Abstract Scope |
The heterogeneous mechanical properties found in biological materials have profound implications for both engineering and medical applications. However, in traditional inverse finite element analysis for material characterization, material homogeneity is often assumed due to the difficulties in experimentally measuring local stress within the material. In this work, we proposed using physics-informed neural networks (PINNs) to identify the full-field elastic properties of highly nonlinear, hyperelastic materials. We applied our improved PINNs to three structurally complex materials and determined the accuracies of PINN-estimated full-field material parameters in three constitutive material models: Neo-Hookean, Mooney-Rivlin, and Gent. Our PINN model consistently produced highly accurate estimates of the full-field elastic properties, with L2 relative errors of less than 5% across all examples. These excellent results demonstrated the promising potential of using PINNs to advance our understanding of mechanical behaviors in biological materials. |