Tackling Turbulence Biological Flows

Elias Balaras
Department of Mechanical Engineering University of Maryland, College Park MD 20742, USA

Supported by NSF (Grant CTS-0347011) AFOSR (Grant 050421-7793) ONR (Grant 072789-8812)

Outline

Introduction
Simulating biological flows

Fluid-structure interactions
Embedded Boundary Method

Applications

Treatment of aortic valve stenosis

Summary and Future Research

Introduction
cm

What are biological flows?

Internal flows: The human body, where fluids play a critical role, i.e.
Respiratory system
Circulatory system
. A variety of flow phenomena at multiple scales:
Organ level (Re<8000)
Cellular level
Molecular level

Flow patterns in the human aorta

nm

Blood elements

Molecules on the cell surface

Introduction
What are biological flows?
External flows: Other organisms that move and feed in the water and air, i.e.
Micro organisms
Birds, insects,
Fish
Wide Re number range: 0.01<Re<106

mayfly

dragonfly

Yellow fin tuna

Introduction
Is turbulence important in biological flows?
Example 1: Turbulence is the exception in the circulation. It appears in pathologic situations and triggers some unique biological responses:
Atherosclerosis
Medical implants can trigger turbulence
Medical devices Turbulence is NOT desirable in blood circulation and there is a need to better understand and control (avoid) it:
Disease research
Surgical Planning
Devise Design

Blocked artery treated with stent

Introduction
Is turbulence important in biological flows?
Example 2: In external flows Re number can be higher. Turbulent wakes can be observed in:
Insect and bird flight
Fish swimming
Man-made devices (AVs, UAVs, etc.) Impact on:
Unsteady aerodynamics
Devise Design

Wake of a thrush nightingale in free flight (G. Spedding)

Introduction
Simulations of biological flows?
Basic characteristics:
Unsteady fully three-dimensional flows
Transitional, non-equilibrium flows
RANS closures not appropriate Eddy resolving approaches like Direct Numerical Simulations (DNS) or Large-Eddy Simulations (LES) are ideal Feasible: low and moderate Re numbers

What are the challenges in LES/DNS?

Geometric complexity
Fluid Structure Interactions
Complex Fluids
Unsteady boundary conditions

Very little work has been done on FSI in LES
Priority in order to cover a wide area of applications

Methodologies
Fluid-Structure interactions

Fluid-Structure interaction simulations are among the most challenging problems in computational mechanics. In grid based methods two are the main challenges:
Boundary motion
Coupling scheme

Boundary velocity & position Hydrodynamic loads

Fluid model

DNS/LES also requires optimal mass, momentum, and energy conservation properties to avoid contamination of the smallest resolved scales

Structural model

Methodologies
Fluid-Structure interactions

Boundary conforming methods (BCM)

Grid deformation is required to satisfy the conformation constrain Equations need to be modified to account for relative motion to the grid Boundary conditions are imposed as with stationary bodies Flexible in clustering grid points For large deformations grid quality is an issue for stability and efficiency

Non-Boundary conforming methods (nBCM)

A fixed Eulerian grid is used at all times Equations of motion remain unchanged Imposition of boundary conditions is not trivial Inflexible in clustering grid points Quality of the solution does not depend on how large deformations are

NBCM can be cost/efficient for DNS/LES at moderate Re
Imposition of B.C. on a grid not aligned to body
Coupling with structural model
Adaptive mesh refinement

Methodologies
resolution requirements
Boundary-Conforming Methods (BCM) Non-Boundary-Conforming Methods (NBCM)

As Re
, total number of grid points grows faster for NBCM than BCM
For laminar boundary layers, number of points of NBCM / BCM  Ren

Methodologies
The embedded boundary approach

Practically the solution is reconstructed locally to satisfy boundary conditions

(Fadlun et. al. 2000)
This is equivalent to the use of a forcing function

Methodologies
Embedded boundary method: implementation
Step 1: Establishment of the grid/interface relation
Step 2: Reconstruction of the solution near the immersed boundary
Step 4: Treatment of points that change phase

Methodologies
(Basic Fluid Solver)

Cartesian/Cylindrical coordinates
Semi-implicit Crank-Nicolson/Adams Bashforth fractional step method
Second order central difference on a staggered grid
The Lagrangian dynamic eddy viscosity model is used for the parameterization of the SGS
Solver is parallelized using domain a decomposition approach

Methodologies
Embedded boundary method: Steps 1-3

Tagging

Local Reconstruction

Boundary Points;

Solid Points;

Fluid Points

B.C. for interface applied on boundary points

Methodologies
Embedded boundary method: Step 4 Time Step k -1 Time Step k Time Step k -1

: Old Boundary Points  New Fluid Points Field Extension at Time Step k -1: Physical Solution at Time Step k -1 Extrapolate Solution near the Interface Non-physical Derivatives at Time Step k -1 Both Solution and Its derivatives are orrect

Balaras Comput. & Fluids 2004 Yang & Balaras J. Comput. Phys. 2006

Methodologies
Coupling scheme
Two general categories of coupling schemes: Fluid model
Weak coupling: Equations for fluid and structure are advanced sequentially using the latest info available.
Strong coupling: Equations for fluid and structure are advanced simultaneously

Boundary velocity & position

Hydrodynamic loads

Structural model Which one?
Weak coupling schemes are unstable for low density ratios
Strong coupling computationally expensive

Methodologies
Strong Coupling scheme

Methodologies
Coupling scheme: stability Re = UD/ = 200 Damping Ratio
=0.004 Mass Ratio n = 0.89 Reduced Velocity Ured = 4

Methodologies
Coupling scheme: stability Mass Ratio n = 0.89 Trajectory Mass Ratio n = 0.88

Ignorable Differences Strong Coupling: Niter
4

Weak Coupling: Diverges Strong Coupling: Stable

Methodologies
Coupling scheme: robustness
Re = UbulkD/ = 200
Mass Ratio n = 10
Damping Ratio =0.03
Reduced Velocity Ured = 5 2 x 4 DoFs
Yang et. al. J. Fluids & Structures 2007

2 x 9 DoFs

2 x 1 DoFs

Niter/timestep
1.7 Niter/timestep
2

Niter/timestep
2

Methodologies
code performance: flow around a golf ball

Grid resolution:
Marginal grid: 61 million points
316 x 127 x 1502 (64 proc)
Coarse grid: 172 million points
536 x 127 x 2502 (125 proc)
Intermediate grid: 575 million points
760 x 252 x 3002 (250 proc)
Fine grid: 1.14 billion points
760 x 502 x 3002 (500 proc)

surface mesh

Methodologies
code performance: flow around a golf ball

Methodologies
code performance: flow around a golf ball

Re = 10k

Re = 115k

Methodologies
code performance: flow around a golf ball

Methodologies
Adaptive mesh refinement

Flexibility in distributing grid nodes is important in moving boundary problems
Local refinement of a sub-grid block is performed by bisection in each coordinate direction.
Each sub-grid block has a structured Cartesian topology and utilizes the single block solver described before

Cross-section of locally refined grid around a sphere

Adaptive Mesh Refinement

topology

Level 1

Divide the domain in sub-blocks. Each sub-grid block has a structured Cartesian topology, and is part of a tree data structure that covers the entire computational domain. Local refinement of a sub-grid block is performed by bisection in each coordinate direction. Number of nodes in each sub-block remains constant

Level 2

Level 3

Level 4

Adaptive mesh refinement: overview

We use a projection method, where advective and diffusive terms are advanced explicitly
We use the Paramesh toolkit (developed by MacNeice and Olson) for the implementation of the AMR process. The package creates and maintains the hierarchy of sub-grid blocks, with each block containing a fixed number of grid points.
A single-block Cartesian grid solver is employed in each sub-grid block:
standard staggered grid in each subblock
second-order central finite-differences
A multigrid solver is used for the Poisson equation (adapted from FLASH)
Guard cells are used to discretize equations at the interior coarse-fine interfaces

standard staggered grid coarse-fine interface

Adaptive mesh refinement: accuracy

Validation: Taylor Green Vortex



Compare numerical solution to analytical solution of 2D Navier-Stokes equations Domain: [
/2, 5
/2]x [
/2, 5
/2] Homogeneous Dirichlet/Neumann velocity boundary conditions and Neumann pressure boundary condition

Adaptive mesh refinement: accuracy

Validation: Taylor Green Vortex
Domain with 2 refinement levels linear interpolation

Domain with 2 refinement levels quadratic interpolation

Uniform domain

interpolation strategy is critical in

maintaining the 2nd order accuracy of numerical scheme

Adaptive mesh refinement: validation

Vortex Ring impinging on a wall, Re
570

Compare AMR solution to numerical solution using a Single Block, Cartesian solver. Velocity Dirichlet BCs in top and Bottom Boundaries, periodic on side walls. Pressure Neumann BCs.

Q contour for vortex impinging normal to a wall, Re
570 (top view)

Adaptive mesh refinement: validation

Vortex Ring impinging on a wall, Re
570

Compare AMR solution to numerical solution using a Single Block, Cartesian solver. Velocity Dirichlet BCs in top and Bottom Boundaries, periodic on side walls. Pressure Neumann BCs.

vorticity isolines at a cross section, Re
570

Falling plates: results

Vorticity contours and Block boundaries for AMR calculation.

Falling plates: results

Vanella, Rabenold & Balaras, J. Comp. Phys. 2008 (under review)

Applications

Heart valve disease

4 chambers
2 atriums
2 ventricles
4 valves
2 atrioventricular
2 semilunar
Left side; high pressure
Right side: low pressure
Mitrial and Aortic valves are the most commonly affected valves

Heart valve disease

Replacement of defective heart valves with artificial prostheses is a safe and routine surgical procedure worldwide (180,000/year)
Several different types of prosthetic valves:
Mechanical HV
high durability, excellent biocompatibility, low level of transvalvular pressure drop
Hemolysis and thrombus formation are major complications
Bioprosthetic (tissue) HV
(better hemodynamics, long-term anticoagulants not required)
Mechanical bi-leaflet Bio-prosthetic

Heart valve disease

Valvular Heart Disease:
Not regarded as major public health problem
Common and Underdiagnosed?

Heart valve disease

Prevalence of heart valve disease (%)
Nkomo VT et al. Burden of valvular heart diseases: a population-based study Lancet 2006; 368:1005-11

Heart valve disease

Survival after detection of moderate or severe valvular heart disease

Nkomo VT et al. Burden of valvular heart diseases: a population-based study Lancet 2006; 368:1005-11

Treatment: Conventional AVR

Treatment: Conventional AVR

Current Status
Mean Age: Prior Operation: Cross-clamp time: Perfusion time: Operative Mortality: Major Complications: CVA: Renal Failure
Source: Society of Thoracic Surgeons Database

66 years 17 % 80 minutes 110 minutes 4% 18 % 2% 5%

Treatment: Conventional AVR

Perioperative mortality

Survival after aortic valve replacement by age. From Blackstone et al. 2003

Treatment: Conventional AVR

Long-term results with conventional AVR: Bad for the Brain

Treatment: Conventional AVR

Long-term results: Causes of Death (% of all Deaths)* Mechanical 37% 17% 36% 10% Bioprosthetic 41% 21 % 26% 12%

Prosthesis related Cardiac not prosthesis related Noncardiac Undetermined

At 15 years, 20 percent had suffered a stroke: Bad for the Brain

* (Hammermeister, Sethi et al. 2000) From The VA prospective valve replacement study follow-up = 15 years.

Treatment: Conventional AVR

What is the role of LES/DNS? 1. Need to better correlate hemodynamic performance of current prosthetic valves to thromboembolic complications 2. Design better implants

Flow around prosthetic heart valves

The incompressible Navier-Stokes equations governing fluid motion are solved as a coupled system with the ODE governing the motion the leaflet

Flow around prosthetic heart valves

Used different grid types and sizes:

Cylindrical coordinates CY2: 329
141
246

Cartesian coordinates CT2: 640
200
200

<u>
Re = UbD/
= 4000 <u> at the inflow plane <urms>

Variation of the flow rate and opening angles during the cycle

Variation of the instantaneous streamwise velocity at y-z palne

Treatment: Conventional AVR

AVR Surgery: Denied to Many Patients ?

Surgical AVR

Not a surgical candidate -
Too old -
Too sick -
Wont tolerate operation

Nonoperative Therapy

Treatment: Conventional AVR

AVR Surgery: Denied to Many Patients ?

124 patients > 60 years* 
Symptomatic AS 
39 % Aortic valve replacement Surgery Age 60 69 77 % 70 79 60 % > 80 22 %
* Charlson E, et al. Decision-making and outcomes in severe symptomatic aortic stenosis J

Heart Valve Dis 2006;15(3):312-21.

The Alternative: Aortic Valve Bypass

Aortic Valve Bypass (or Apicoaortic Conduit)
Creates a new outflow from the apex of the left ventricle to descending aorta.

Conceived by Carrel in 1910
Performed experimentally by Sarnoff in 1955
Clinically by Templeton in 1962
First in man reported by J.W.Brown in 1974
More than 100 operations U. Maryland recently

The Alternative: Aortic Valve Bypass

The Alternative: Aortic Valve Bypass

AVB Components

Left Ventricle Connector (LV connector)
Prosthetic Valve
Vascular Graft (if not part of valve)

The Alternative: Aortic Valve Bypass

The Alternative: Aortic Valve Bypass

Advantages of Aortic Valve Bypass

Avoid sternotomy (patent grafts) 
No aortic cross-clamping 
No (or minimal) cardiopulmonary bypass (BEATING HEART operation! ) 
Patient-prosthesis mismatch impossible 
Brain Protective Current application: Very High-Risk Patients

The Alternative: Aortic Valve Bypass Open Questions:


What is the relative blood flow through the conduit and the native aortic valve? 
How much retrograde flow is there? How much stasis is there in the descending aorta? Is there a known "threshold" where thrombosis might occur? 
Is blood flow to brain and coronaries unchanged compared to normal anatomy? 
Can we predict the final left ventricular outflow gradient and the size of the conduit 
What would be the SMALLEST conduit we could use to achieve adequate relief?

The Alternative: Aortic Valve Bypass

Set-up of preliminary computations

AVB geometry

Normal geometry

The Alternative: Aortic Valve Bypass

The Alternative: Aortic Valve Bypass

The Alternative: Aortic Valve Bypass

The Alternative: Aortic Valve Bypass

Table 1: Summary of the computations Case I Normal II AS III AS + 20 mm AVB IV AS + 16 mm AVB V AS + 10 mm AVB Native Aortic Valve Stenosis (%) 0 80 80 80 80 Table 2: Native Aortic valve gradients. Clinical data Mean (mmHg) Case II (80% stenosis) Case III (AVB, D=20mm) Case IV (AVB, D=16mm) Case V (AVB, D=10mm) 43.07.0* 8.83.3* Mean Normalized 1 0.21 Numerical data (2-D) Mean (mmHg)** 43.0 12.5 13.8 17.6 Mean Normalized 1 0.29 0.32 0.41 Diameter of AVB Conduit



No conduit No conduit 20 16 10

The Alternative: Aortic Valve Bypass

Flowrate distribution

The Alternative: Aortic Valve Bypass

Velocity isolines at peak systole

The Alternative: Aortic Valve Bypass

Vorticity isolines at peak systole

Acknowledgements

Prof. B. Balachandran (University of Maryland)
Prof. S. Predikian (University of Cordoba, Argentina)
Prof. J. Gammie (University of Maryland Medicine)
Dr. Kyung Se Cha
Ph.D Students
Jianming Yang
Nikolaos Beratlis
Marcos Vanella
Antonio Cristallo