In this thesis we propose a model to numerically study the behavior of blood, which is considered as an incompressible Newtonian fluid, in the presence of a clot attached to a vessel wall. The purpose of this study is to find out whether different flow regimes may cause a clot to detach from a vessel wall or it would lead to a stable state. In Chapter 1, we give a literature review of previous studies modeling blood coagulation, blood clots in the vascular system, platelet adhesion and aggregation and pathological clot formation. Our work is mainly based upon some part of the mathematical model given by Bajd and Serša [3], which is presented in Chapter 1. Then, we will describe the mathematical modeling of the fluid presenting the blood and the solid representing the clot in Chapter 2. The third chapter will focus on the numerical approach consisting of a projection method and the immersed boundary method [36] for solving the Navier-Stokes equations. Finally, in Chapter 4, we will discuss the results and give conclusions about the influence of different flow regimes on the clot stability.

This thesis develops the work of Fatone, Funaro et al. on superconsistent discretizations. We first give a precise definition of superconsistency and how it relates to consistency. We then suggest an explicit approach to constructing a superconsistent scheme from a linear operator and we apply the method to the advection-diffusion operator which we find, for example, in advective diffusive problems and the Navier Stokes equations. Both an analytical study and some numerical tests are provided. The choice of problems is made so that the diffusive term is small compared to the advective term. Solutions to such problems typically contain a boundary layer and classical centered finite difference methods may display some spurious oscillations in and near that layer. In contrast, solutions obtained with the superconsistent method are far more stable. In this work, only problems in one and two dimensions are treated since three dimensional problems would be too demanding on computational resources.

The objective of this thesis is to introduce a new implicit purely lagrangian smoothed particle hydrodynamics (SPH) method, for the resolution of the two-dimensional incompressible Navier-Stokes equations in the presence of a free surface. Our discretization scheme is based on that of Kéou Noutcheuwa et Owens [19]. We have treated the free surface by combining Yildiz et al. [43] multiple boundary tangent (MBT) method and boundary conditions on the auxiliary fields of Yang et Prosperetti [42]. In this way, we obtain a discretization scheme of order $\mathcal{O}(\Delta t ^2)$ and $\mathcal{O}(\Delta x ^2)$, according to certain constraints on the smoothing length $h$. First, we tested our scheme with a two-dimensional Poiseuille flow by means of which we analyze the discretization error of the SPH method. Then, we tried to simulate a two-dimensional Newtonian extrusion problem. Unfortunately, although the behavior of the free surface is satisfactory, we have encountered numerical problems on the singularity at the output of the die.

This paper first discusses various attempts at modeling and simulating anguilliform swimming, then we develop a new technique, based on a method of generalized immersed boundary and the beam theory of Reissner-Simo. Subsequent to the derivation of the equations of polar fluids, the beam theory is derived from continuum mechanics and the resulting equations are then discretized, allowing a numerical solution. For the first time, the theory of additive Runge-Kutta schemes are combined with the Runge-Kutta-Munthe-Kaas method to generate schemes of arbitrarily high formal order of convergence. Moreover, the interpolation and spreading operations are handled from a new point of view that suggests the use of interpolatory nodal splines instead of spreading traditional functions. Finally, many numerical verifications are done before considering simulations of swimming.

While simulating blood flow in a microvascular network (in the absence of biological control), it is possible to observe the presence of oscillations in certain parameters such as blood flow, nodal pressure and hematocrit (red blood cell concentration in blood). This behaviour seems consistent with certain in vivo experiments. Despite this agreement, one has to wonder if the fluctuations observed in simulations are physical in nature, numerical or an artefact of unrealistic models since there are always differences between modelling and in vivo experiments. To settle this question satisfactorily, we will study and analyze blood flow and the nature of the fluctuations in different microvascular networks using a convective model and a well-mixed model to depict the governing equations for conservation of red blood cell mass. These models take into account two important rheological effects : the Fåhraeus-Lindqvist effect describing the apparent viscosity of blood flow in a vessel and the plasma skimming effect which describes the separation of red blood cells at diverging nodes. To describe the latter effect, we will implement two plasma skimming models (Pries et al. and Fenton et al.). In this thesis, we will present a description of the physiological problem (blood rheology). We will introduce the mathematical models used (well-mixed and convective) as well as the plasma skimming models (Pries et al. and Fenton et al.) accompanied by a detailed analysis of the numerical methods implemented. For the well-mixed model, we use the traditional explicit Euler method as well as a new implicit scheme that allows us to solve the problem in an efficient manner. This is done using a Newton-Krylov method with a preconditioned conjugate gradient and GMRES method for the inner iterations as well as a quasi- Newton method (Broyden’s method). The implicit method will include the vi backward Euler and trapezoidal methods. For the convective model, the explicit method of Kiani et al. will be implemented as well as a new numerical implicit approach. The stability of these numerical schemes will be explored. Using three different topologies, we will compare the results of the two mathematical models as well as the two plasma skimming models and the various numerical methods in order to ascertain to what extent the oscillations that have been observed using the traditional schemes may be attributable to the choice of the mathematical models or the choice of the numerical methods.

The Immersed Interface Method allows us to extend the scope of some numerical methods to discontinuous problems. Here we use it in the case of an incompressible fluid governed by the Navier-Stokes equations, in which a membrane is immersed, inducing a singular force. We use a projection method and staggered (MAC-type) finite difference approximations. A very complete derivation for the jump conditions is presented in the Appendix, for the case where the viscosity is continuous. Two numerical examples are shown : one without a membrane, and the other where the membrane is motionless. The general case of a moving membrane is also thoroughly studied.

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal.