h1

TY  - THES
AU  - Hillgärtner, Markus
TI  - Modeling of soft biological tissues by micro-mechanically motivated and data-driven approaches
VL  - 13
PB  - RWTH Aachen University
VL  - Dissertation
CY  - Aachen
M1  - RWTH-2021-11875
SN  - 978-3-9821703-2-9
T2  - Bericht / RWTH Aachen University, Lehr- und Forschungsgebiet Kontinuumsmechanik
SP  - xxiii, 222 Seiten : Illustrationen, Diagramme
PY  - 2021
N1  - Dissertation, RWTH Aachen University, 2021
AB  - The present work deals with the question of how constitutive modeling of soft biological tissues and other hyperelastic materials can be extended and improved by utilizing additional material data. Classically, measured experimental curves, such as tensile tests, are fitted to existing material models that are mostly phenomenological. While this kind of modeling can lead to a good agreement with the considered material's mechanical behavior, approaches of this kind fail to predict the behavior of previously untested materials. Phenomenological modeling proves to be particularly impracticable in the field of biomechanics, where the material behavior is highly patient-specific, and experiments on the living material under consideration are often not possible. Hence, there is a need for modeling concepts that can deliver accurate predictions of the mechanical behavior of materials based on additional information. In this thesis, two different approaches of this kind are presented. The first approach deals with micro-mechanically motivated modeling of soft biological tissues, where material information from minimally invasive testing can be incorporated. The presented multi-scale model builds up from single tropocollagen molecules with different lengths and preferred directions within a statistical framework. A damage mechanism that depends on force and time, based on the dissociation of adhesive bonds, is implemented, which enables modeling of both fatigue phenomena and inelasticity. This is achieved by a load-dependent degradation of the collagen molecules and the interfibrillar matrix. The second concept uses artificial neural networks for an invariant-based approach of modeling of hyperelastic materials. The proposed method can capture anisotropy by utilizing generalized structure tensors and processing any amount of additional material information. In the network architecture, general parts of the mathematical formulation are strictly separated from material-specific parts to achieve high flexibility in conjunction with a mechanically consistent framework. This general approach also allows predicting the behavior of unknown materials if the network has been previously trained with a sufficient amount of similar materials. Both approaches were validated with different experimental and artificial data sets and showed a high degree of agreement.
LB  - PUB:(DE-HGF)11 ; PUB:(DE-HGF)3
UR  - https://publications.rwth-aachen.de/record/837241
ER  -