Inertial and non-inertial microfluidics of deformable particles, and red blood cell-resolved capillary network hemodynamics with machine-learning applications

A computational study and a machine learning approach are used for the numerical simulation of healthy and stiffer red blood cells and the prediction of the hemodynamics of blood flow in a microvascular network. We utilized the immersed boundary method to simulate blood flow in the complex microvasculature and continuous immersed boundary method to capture cell deformation. Two common machine learning approaches namely, regression and classification are employed for prediction of volume flow rate and hematocrit in the entire microvasculature. In the first part of this thesis cross stream migration of a capsule in a three-dimensional curved vessel is investigated in the presence and absence of inertia. For these two cases the equilibrium location of the capsule is determined. The effect of different parameters including vessel curvature, capsule deformability, and tube Reynolds number is considered. We also study lateral migration of capsule in a U-shaped vessel to consider the effect of change in the curvature of vessel. In the second part of this study a cross–stream migration of a deformable capsule in curved microchannel of square and rectangular section under inertial and non-inertial transport is studied. We discuss the different focusing behavior in these two regimes due to the interplay between inertia, deformation, shear gradient, streamline curvature and secondary flow. We also investigate the effect of channel width and height on the capsule equilibrium location. In the third part of this thesis, the influence of reduction in cell deformability on microvascular hemodynamics is established. Blood flow including thousands of red blood cells in a physiologically realistic microvascular network is simulated. We study how reduced RBCs deformability can influence RBC trafficking with significant and heterogeneous change in hematocrit. We quantify the change in RBC partitioning, perfusion, vascular resistance, and wall shear stress due to the change in RBCs deformability. Finally, we investigate the applicability of machine learning (ML) techniques to predict blood flow and RBC distributions in physiologically realistic vascular networks. We acquire data from high-fidelity simulations of deformable RBC suspension flowing in the networks. With the flow and hematocrit specified at an inlet of a vasculature, the ML models predict the time-averaged flow rate and RBC distributions in the entire network, time-dependent flow rate and hematocrit in each vessel and vascular bifurcation in isolation over a long time, and finally, simultaneous spatially and temporally evolving quantities through the vessel hierarchy in the networks.Ph.D.Includes bibliographical reference