Efficient Parallel Non-Newtonian Blood Flow Simulations in Trilinos - Towards Incorporating Microscopic Information with a Data-Driven Approach
Please login to view abstract download link
Blood flow plays an essential role in the functioning of the healthy and diseased human organism. Understanding and predicting the flow in vessels of varying sizes is challenging due the tight coupling of processes at the molecular and cellular level with the macroscopic flows. No single simulation technique is able to cover the involved range of length scales such that a combination of different approaches is urgently required. In this talk, we present simulations of blood flow at the macroscopic level, utilizing sophis- ticated generalized Newtonian constitutive equations[1]. These models are integrated into our in-house FEM solver FEDDLIB (Finite Element and Domain Decomposition Library), a C++ library designed for large-scale simulations and providing an interface to the highly scalable implicit domain decomposition solver FROSch (Fast and Robust Overlapping Schwarz) [2, 3]. We highlight the use of monolithic overlapping Schwarz precondition- ers with RGDWS/GDSW coarse spaces and showcase scalability results. Furthermore, we will discuss initial strategies for enhancing macroscopic simulations by data-driven approaches leveraging cell-level data from dissipative particle dynamics simulations.
