SNIC SUPR
A Turbulent Flow Simulation of the Carotid Artery
Dnr:

SNIC 2019/3-363

Type:

SNIC Medium Compute

Principal Investigator:

Christian Gasser

Affiliation:

Kungliga Tekniska högskolan

Start Date:

2019-06-24

End Date:

2020-01-01

Primary Classification:

10105: Computational Mathematics

Secondary Classification:

10399: Other Physics Topics

Tertiary Classification:

10799: Other Natural Sciences not elsewhere specified

Webpage:

Allocation

Abstract

Background Carotid artery disease is a major cause of strokes in many industrialised countries. Plaque builds up in the carotid arteries, it eventually ruptures risking build up and plugging of brain arteries, causing strokes. Simulations based on CT-images and some patient-specific parameters could be used to predict complications. Modelling of these systems are characterised by non-newtonian fluidal properties, soft walls, pulsatile flow and turbulent flow. Turbulent flows has shown to affect the future of cells physiological activity and possibly be pathological [1, 3, 5]. A high degree of stenosis in the carotid artery has been suggested to produce turbulent flows [2] and shown in simulations [6], this was also comfirmed in the initial test simulations. Several technical problems arise such as, choosing and implementation of fitting boundary conditions, creation of a usable/optimised mesh, choosing a method to accurately produce turbulence. Most importantly rigorous verification of the simulated results. Earlier work at this group have found that the pressure outflow needs to be refined and suggests a model based on a Windkessel theory, furthermore the lack of computational power was evident [7]. In light of this we propose the use of the unified continuum mechanics solver Unicorn. A High Performance branch of the FEniCS project to do finite element simulations of turbulent flow, it has been successful with earlier biofluidal simulations [4]. This would be done on PDC’s Beskow where Unicorn is easily set up. If successful, trying to scale down to normal FEniCS at the local cluster would simplify the pipeline for assessment and study of carotid artery stenosis in future research. If this approach renders sufficiently realistic results a clinical study is proposed to determine inherited pathology by simulations on several CT-generated meshes. Research questions/goals Main questions that will be answered: • Optimise stabilisation terms in Galerkin Least Square stabilisation(GLS)/Discuss other turbulence models (already found in OpenFOAM?). • Refine BC’s: Is the current discrete inflow solution bad (shocks)? Would a sinusoidal or a fitted polynomial be a better solution? Investigate outflow conditions (pressure). Discuss other solutions than "velocity in pressure out". • Investigate performance implications from adaptive mesh refinement (vital to GLS). • Numerical stability over a range of Re numbers over several Cardiac cycles. • Verification of results. • Scale down to non-HPC FEniCS.