Comparison of numerical implementations for modelling flow through arterial stenoses

Abstract

The fluid dynamics of occluded arteries depends on many physiological factors and the resulting complexity of the post-stenotic flow field present a challenge for numerical solvers. The primary goal of this paper is to assess the performance of non-commercial and academic CFD solvers that have not been compared previously, for the purpose of stenosed arterial flow modelling. We evaluated CHeart, Crimson, Nektar++ and OpenFOAM which altogether employ a wide range of algorithms, including finite element and finite volume methods, coupled and segregated pressure-velocity solves, implicit and explicit time stepping, and varying spatial discretization orders. Furthermore, improved realism in test scenarios was achieved, in which three representative artery geometries of increasing complexity were employed with time-dependent boundary conditions. In the simplest geometry, codes that employed explicit stepping and segregated solve yielded an equivalent accuracy within shorter computational times regardless of whether FEM or FVM was used. However, the disagreements between solver predictions grew in more complex geometries with increasing mesh refinement (by 50%+ with 10× grid points), reaching a level (7% in Δp) which is unsuitable for clinical purposes. We conclude that while accelerated schemes are safe to use in the lower Re simpler stenotic arteries, the choice of a CFD solver is likely to affect the clinical recommendation in the general diseased arteries, and should be made with application-specific considerations.