← Back to Completed Projects

Project Overview

Developed a fast and scalable framework to enable patient-specific multi-scale CFD simulations of aortic hemodynamics by automatically estimating Windkessel model parameters.

This work bridges 3D high-fidelity CFD (LBM) and 0D reduced-order cardiovascular models, enabling efficient and realistic simulation of blood flow in complex vascular systems.

Problem & Motivation

Accurate cardiovascular CFD simulations require physiological outlet boundary conditions, typically modeled using Windkessel (WK3) models.

However:

• WK3 parameters (Rc, Rp, C) are case-specific and unknown
• Existing methods:
  • Require iterative CFD simulations (very expensive)
  • Ignore geometry-induced resistance
  • Or rely on over-simplified assumptions (e.g., Poiseuille flow)

Result: High computational cost or reduced accuracy in patient-specific simulations.

What I Built (Core Contribution)

Designed a fast parameter estimation pipeline with three key components:

1. Geometry-aware resistance extraction
   • Performed steady CFD simulation
   • Extracted branch-wise geometric resistance (Rg)
   • Captures real effects of:
     • complex geometry
     • bifurcations
     • stenosis
2. Reduced-order circuit modeling
   • Built a 0D circuit analog coupling:
     • geometric resistance (Rg)
     • Windkessel model (Rc, Rp, C)
3. Optimization-based parameter estimation
   • Implemented MATLAB global optimization (pattern search)
   • Solved WK3 ODE system to match:
     • systolic / diastolic pressure
     • flow distribution at outlets
   • Optimization runtime: < 1 minute

Technical Stack

• CFD Solver: Lattice Boltzmann Method (LBM)
• Framework: C++ + MPI (Palabos-based)
• Optimization: MATLAB Global Optimization Toolbox
• Scale: ~21 million grid points (HPC cluster)
• Physics: Multi-scale coupling (3D CFD + 0D lumped model)

Key Results

• Achieved accurate control of flow distribution and pressure waveform
• Pressure prediction error: as low as ~0.4 mmHg
• Successfully handled: normal geometry, stenosed aorta, non-physiological flow distributions
• Reduced computational cost by:
  • eliminating iterative CFD loops
  • using only one steady CFD simulation

Engineering Impact

• Enabled fast and reliable patient-specific simulations
• Improved robustness for:
  • complex geometries
  • pathological conditions (e.g., stenosis)
• Provides a scalable workflow for multi-scale modeling
• Applicable to:
  • biomedical simulation platforms
  • digital twins in healthcare
  • simulation-driven diagnosis tools

Result Gifs

Velocity Field

Vorticity Field

Surface Force

Paper Link

The whole paper can be found here.

← Back to Completed Projects