Computational Modeling of Hollow-Fiber Hemodialyzers (Artificial Kidneys)

Abstract

Hollow-fiber hemodialyzers are widely used medical devices for removing uremic toxins from blood. Despite their introduction in the 1960s, their design has seen limited advancement, and several performance limitations remain unresolved. Computational modeling offers a powerful alternative to experimental approaches by enabling rapid exploration of device design, dialysis dose, and operating conditions while reducing time and cost. This dissertation first reviews existing models of dialyzer hydrodynamics, broadly categorizing them into heterogeneous (single-fiber resolved) and pseudo-homogeneous (porous media) approaches, and highlights restrictive assumptions that may limit predictive accuracy. Existing approaches for estimating dialysate-side permeability rely on single-fiber theories, semi-empirical packed-bed correlations, or hydraulic resistance experiments. In this work, permeability is computed directly by resolving flow through the complex shell-side geometry containing 10⁴ fibers, using three-dimensional Navier–Stokes simulations. Analysis across twenty dialyzer modules yields a generalized relationship between pressure drop and flow rate, formulated here as a Modified Ergun Equation. Building on this framework, a sustainability-oriented performance metric—clearance per pumping power—is introduced to quantify the trade-off between solute removal and hydraulic energy consumption and to identify water- and energy-efficient operating conditions. In addition, a multiscale modeling framework is developed to predict solute clearance by coupling porous media descriptions with resolved free-flow regions around fibers. Finally, a multifiber model is proposed, enabling prediction of dialyzer clearance using only a limited number of representative fibers.