Finite Element Simulation of Human Head under Frontal Impact with Uncertainties in Constitutive Modeling and Material Parameters
Temfack Fogang, Therence Aymard
MetadataShow full item record
Traumatic brain injury (TBI) is a leading cause of death and disability, and poses a serious health, economical, and social problem worldwide. Given the tremendous advancement of computing capability and cutting edge imaging techniques, modeling and simulation of brain mechanical response to head impact can provide insightful understanding about TBI. However, the modeling and simulation are often plagued by limited experimental information (both in terms of measurement resolution as well as limited samples/specimens) and modeling idealization. The goal of the current work is to characterize such limited information and modeling idealization within a probabilistic framework. A three-dimensional (3D) brain finite element (FE) model consists of geometrical features, anatomical features (e.g., skull, white and gray matter, blood vessels, cerebrospinal fluid, etc.), constitutive material parameters, loading, boundary condition, and interaction mechanism between different regions. The geometrical and anatomical features are typically gleaned from a database of two-dimensional (2D) images, while loading and material parameters are estimated from controlled mechanical testing. The boundary conditions and interaction mechanism, on the other hand, are often idealized. Significant variations in experimentally characterized material parameters can be found in the existing literature. The quality and credibility of a deterministic 3D FE model is evidently restricted after a certain extent by the finite measurement resolution, lack of sufficient experimental data and modeling approximations. To understand some of these difficulties, the effects of variations in constitutive modeling as well as in material parameters on brain response are investigated within a probabilistic framework in this work. The proposed stochastic model will be useful to cross-validate the probabilistic model predictions against available experimental data.