Efficient Parallel Numerical Simulation Of Cerebral Blood Flow With The Consideration Of Blood Flow Distributions

Numerical simulation is widely used in the biomedical field and is the driving force behind the application and development of the hemodynamics.After decades of development,numerical simulations of blood flows in human arteries have achieved a large number of significant results.In this paper,based on the medical image of a patient,we construct the geometrical model of a full cerebral artery and generate the corresponding unstructured mesh to perform the patient-specific cerebral hemodynamic simulation.Firstly,a novel resistance split method is proposed to determine the model parameters in the outlet boundary conditions,which can effectively regulate the regional blood flow distribution.Secondly,a two-level additive Schwarz preconditioned Newton-Krylov algorithm is studied,aiming to improve the parallel efficiency and reduce the computational time of blood flow simulation.Results show that the obtained hemodynamic parameters,including the pressure,the velocity,and the wall shear stress,are consistent with the clinical measurements.In addition,comparing with the one-level Schwarz preconditioner,the two-level method can effectively reduce the number of iterations and the computational time while ensuring the stability and scalability of the algorithm.Scalability tests show a parallel efficiency of 50%for the two-level Schwarz method when using 2160 processor cores based on a mesh of about 20-million elements.