Countless applications in science and engineering such as blood flow, dam or ship construction, insect flight, car design or wind interacting with buildings and bridges besides many other problems involve incompressible flows and their interaction with elastic bodies. Whenever the flow field triggers corresponding forces acting on wetted structures large enough to cause finite deformations, this can, in turn, significantly alter the flow domain and thereby the velocity and pressure present in the fluid. As a consequence, any reliable model of such physical processes must capture the interaction of fluids and solids.
As a prime example from the biomedical context, complex clinical applications in the cardiovascular system continue to challenge numerical methods and well-established algorithms. Here, incorporating standard mathematical methods can impair the solution scheme's performance drastically. However, there is vast potential for computational science and engineering to assist in clinic, e.g., in training medical personnel or evaluating treatment options in-silico on digital twins. Parameter studies on virtual cohorts or patient-specific modelling of cardiovascular diseases to foster understanding of the underlying principles at work additionally continue to motivate research efforts to develop tailored numerical methods.
Within this work, we first devise novel schemes to tackle generalised Newtonian fluid flow, which incorporate shear-thinning effects contributing to the complex behaviour of blood. Such fluids show a dependence of the viscosity on the local shear rate, such that subtle differences in the flow field's gradients might impact the shear resistance and hence the overall problem drastically. In this context, we extend well-established concepts towards generalised Newtonian fluid flow, devising a stabilised coupled velocity-pressure formulation and an iteration-free split-step scheme decoupling velocity and pressure spaces. Both approaches allow for equal-order interpolation with standard continuous finite elements, which might be considered advantageous in practical biomedical applications as spatial discretisations derived from segmented medical image data are often lower-order accurate.
The stabilised and split-step flow solvers build upon a pressure Poisson equation as a fundamental ingredient, which is derived from the fluid's momentum balance equation, implicitly enforcing mass balance. Together with fully consistent boundary conditions, this yields increased accuracy compared to standard inf-sup stabilisation schemes for the coupled approach or even allows recovering the pressure given velocity. Furthermore, the design of efficient solution schemes decoupling the involved fields and linearising the governing equations based on higher-order accurate extrapolation combined with adaptive timestepping results in a substantial overall speed-up. In this way, the proposed novel techniques facilitate practical applications in science, industry and medicine, since the time to solution can be reduced significantly.
Targeting the haemodynamic regime, we couple the split-step flow solver to a nearly incompressible, hyperelastic, fibre-reinforced continuum modelling the arterial tissue. In this regard, major contributions of this work are (i) combining added-mass stable, semi-implicit coupling of fluid and structure, Robin interface conditions and acceleration schemes such as Aitken's relaxation or the Interface Quasi-Newton Inverse Least-Squares method and (ii) incorporating modelling aspects and numerical techniques necessary for clinical application. The latter aspect involves lumped parameter models to account for the neglected downstream vasculature, construction of suitable inflow profiles on general inflow sections given volumetric flow rates only, stabilising re-entrant flow and dominant convection, incorporating stress states present in the structure at the time of image acquisition and automatic construction of proper material orientation for complex geometries as encountered, e.g., in aortic dissection.
The final fluid-structure interaction scheme thus merely couples the fluid's pressure and the structural solver iteratively, while the remaining subproblems are solved only once per time step, all of which consist of standard problems frequently encountered in science and engineering such as Poisson, mass-matrix or advection-diffusion-reaction equations. Thus, we heavily profit from standard algebraic multigrid preconditioners and linear solvers, which enable us to simulate realistic scenarios using physiological problem parameters.
We carefully investigate the proposed flow and fluid-structure interaction schemes with respect to their accuracy, robustness and reliability via several numerical tests ranging from academic setups over benchmark scenarios all the way to patient-specific simulations, were we demonstrate favourable mathematical properties such as expected convergence rates and required robustness. Thus, the numerical examples presented herein aim to bridge the gap to real-world problems, considering state-of-the-art modelling aspects, physiological parameters and showcasing the framework's versatility and thereby substantially contribute to the field of computational (bio-)mechanics.