You are on page 1of 35

Region I: Boundary Layer

Region II: Near Field

Region III: Far Wake

Two-Phase Flow Simulations, Analysis and Modeling in the Near Wake of a Transom Stern
Prof. D. K.-P. Yue (PI), Dr. K. Hendrickson (co-PI), Dr. G. Weymouth, Dr. S. Banerjee Massachusetts Institute of Technology Presented at ONR Ship Wavebreaking and Bubbly Wake Review Pasadena, CA This work is funded by ONR (N00014-01-1-0630) April 11-12, 2013 Dr. Patrick L. Purtell

Towards turbulent multiphase wake predictions


The ultimate design goal is to model the dependence of the far-wake bubble size distribution on vessel geometry and operations.
Scientific Issue: Disparate range of scales [O(km)O(mm)] precludes direct brute-force simulation. Quantification of the far wake depends on turbulent entrainment and bubbly flow models near the ship. These models need extensive data and physical insights not currently available for ship flows. Objective: Bridge the prediction capabilities for nearfield ship flow and far-field bubbly wake with physicsbased modeling and parameterizations relevant to design.

Objectives
High resolution iLES simulations using validated conservative schemes of the near-field flow of a canonical stern to: Obtain extensive datasets for canonical transom stern bubbly wakes that incorporates both speed and geometry effects. Provide understanding of the flow properties of the mixedphase region in the near wake. Characterize the near-field air entrainment and cavity/bubble size distribution. Develop framework for analysis of incompressible highly variable-density turbulence (IHVDT) for RANS. Quantify modeling implications for VDT in the near-wake of the stern. Develop IHVDT closure models for RANS

Accomplishments
Developed a database of high resolution iLES simulations of a canonical stern over a range of speeds and geometries. Identified large scale flow features in mixed region
New jet flow structure New vortex structure with Strouhal frequency ~0.2 Quantified convective velocity of large scale coherent structures

Characterized surface wake features


Convergent corner waves follow ballistic scaling Rooster tail height and length scale with speed and geometry

Characterized air entrainment


Developed Lagrangian method for tracking air entrainment Convergence and wake evolution of bubble density spectrum Identified regions of strong entrainment Quantified surface entrainment rate and its scaling with geometry and speed

Characterized the IHVDT in the wake. Developed isotropic and higher-order anisotropic closure models for turbulence mass fluxes (TMF) and Reynolds stresses.

ILES Simulations of Transom Sterns


Parameters considered: Multiple speeds U and geometry (beam-to-draft ratios B/D) Establish the dependencies on speed FrD=U/ (gD) and aspect ratio B/D. The methods employed in this study have been documented and validated for canonical turbulent air entraining flows.

FrD=2.38, B/D=1

FrD=2.38, B/D=1

Average Velocity Field Reveals Jet Flow Pattern in Mixed Region


Centerplane

Play

Jet structure on centerline after convergent corner wave collision Weak roller structure in leading edge of mixed region

mixed region

Identify New Vortex Structure in Mixed Region of Breaking Waves


Centerplane

Play

Classic potential flow Roller-Bore structure

Jet flow pattern acts as wake profile and sets up paired vortex pattern in mixed region Jet flow and paired vortex pattern also present in divergent breaking waves Mixed region flow structure

Wake Vorticities Exhibit Characteristic Strouhal Frequency

fD/U

fD/U

Average wake surface features scale with speed & geometry Convergent Corner Waves
Define: Angle in y-z plane of particle leaving bottom corner of stern Angle in x-y plane formed by convergent corner waves Location of convergent corner wave collision Collapse to ballistic scaling with speed and geometry for B/D >= 1

measured

Average wake surface features scale with speed & geometry


Rooster tail & Divergent waves Location of rooster tail f(FrD, D/B) Wake angle constant for B/D >=1

Using Lagrangian Techniques to Establish Entrainment Characteristics

average volume fraction

<fbulk>

High void fraction not completely entrained

average spray fraction

average fraction of entrained air

Identifying regions of strong entrainment and entrainment distribution along wake


Volume (planar) of entrained air distribution

Strong air entrainment confined to narrow region in wake Location of breaking divergent waves identified Entrainment essentially constant in rooster tail

Surface entrainment rate scales with geometry and speed, peak at xc

xc

xd

Initial peak surface entrainment rate collapses at xc Diverging breaking waves has larger Q

Bubble density spectrum converges and evolves down the wake

10/3

Turbulent Mass Flux is an Essential Feature of IHVDT


Continuity Equation : Speed effects
xc xr xd

Turbulent mass flux (TMF) Geometry effects


B/D=1.25 B/D=1.00 B/D=0.75 B/D=0.50

B/D=1 Fr=2.38 Fr=2.53 Fr=3.00

Fr=2.38

TMP must be modeled to obtain IHVDT closure <M1> collapses for different speeds and geometries with Fr and B/D scaling Peak in <M1> identifies the region of intense breaking near x~xB (xB<~xc) (and smaller secondary breaking downstream) Results <M2> and <M3> are similar with generally M1>M3> M2

IHVDT Departure from Isotropy is Associated with Wave Breaking


Turbulent Intensities (TI)
i=3, vertical i=2, cross-stream i=1, stream-wise Fr=2.38; B/D=1 Anisotropic

xc

xr

xd

Fr=2.38 Fr=2.53 Fr=3.00

Isotropic

TI collapses with Fr and B/D scaling similar to TMF Generally TI and anisotropy peak at peak breaking location xB<xc Departure from isotropy can be better quantified by Lumley map of IHVDT invariants

IHVDT Anisotropy/Isotropy in Wake is Characterized by Invariant Parameters


Turbulence anisotropy tensor xc xr xd
Points will only fall within this bounded region Turbulent anisotropy tensor 1 component

2 component

Isotropic

isotropic

III = bij bjkbki


Increasing (x/D)/Fr

Fr=2.38 Fr=2.53 Fr=3.00

Breaking @ x~xB

Increasing (x/D)/Fr, flow becomes more isotropic Anisotropic

<J> (also) collapses with Fr and B/D scaling Initial anisotropy is reduced by breaking at ~xB (and at secondary breaking downstream) Turbulence becomes increasingly isotropic (increasing <J> 1) with distance (x/D)/Fr IHVDT closure must capture varying anisotropy/isotropic characteristics along the wake

Comparison of TMF and Density Gradient Shows Limitation of Models Using (Only) Isotropic Eddy Diffusivity Closure
xc xr xd

(x/D)/Fr=0.71

(x/D)/Fr=1.60

(x/D)/Fr=2.80

(x/D)/Fr

(x/D)/Fr=0.71

(x/D)/Fr=1.60

(x/D)/Fr=2.80

Correlation between TMF and gradient diffusion are generally weak especially in more anisotropic regions TMF exhibits counter gradient behavior not captured by density gradients Similar features for other components (not shown)

(x/D)/Fr

For 3 Fr cases

Fr=2.38, B/D=1

Performance of Anisotropic vs Isotropic TMF Closure Models Importance of Incorporating Anisotropy


xc
Isotropic model Error = DNS - Isotropic model

xr

xd

DNS Data

(x/D)/Fr=0.71

(x/D)/Fr=0.71

(x/D)/Fr=0.71

y/D
(x/D)/Fr

DNS Data

Anisotropic model

Error = DNS - Anisotropic model

(x/D)/Fr=0.71

(x/D)/Fr=0.71

(x/D)/Fr=0.71

y/D

Anisotropic closure model obtains salient features in TMF not captured by isotopic model Performance of anisotropic model is remarkably good for x>~xB Modeling errors are larger in mixed region away from centerline

Performance of Anisotropic vs Isotropic TMF Closure Models Importance of Incorporating Anisotropy Fr=2.38, B/D=1

xc

xr

xd

DNS Data

Isotropic model

Error = DNS - Isotropic model

(x/D)/Fr=1.60

(x/D)/Fr=1.60

(x/D)/Fr=1.60

(x/D)/Fr

DNS Data

Anisotropic model

Error = DNS - Anisotropic model

(x/D)/Fr=1.60

(x/D)/Fr=1.60

(x/D)/Fr=1.60

Anisotropic closure model obtains salient features in TMF not captured by isotopic model Performance of anisotropic model is remarkably good for x>~xB Modeling errors are larger in mixed region away from centerline

Performance of Anisotropic vs Isotropic TMF Closure Models Importance of Incorporating Anisotropy


Fr=2.38, B/D=1
DNS Data Isotropic model Error = DNS - Isotropic model

xc

xr

xd

(x/D)/Fr=2.80

(x/D)/Fr=2.80

(x/D)/Fr=2.80

(x/D)/Fr

DNS Data

Anisotropic model

Error = DNS - Anisotropic model

(x/D)/Fr=2.80

(x/D)/Fr=2.80

(x/D)/Fr=2.80

Anisotropic closure model obtains salient features in TMF not captured by isotopic model Performance of anisotropic model is remarkably good for x>~xB Modeling errors are larger in mixed region away from centerline

DNS Data

Performance of Anisotropic vs Isotropic TMF Closure Models Importance of Incorporating Anisotropy

(x/D)/Fr=0.71

(x/D)/Fr=1.60

(x/D)/Fr=2.80

Isotropic model

(x/D)/Fr=0.71

(x/D)/Fr=1.60

(x/D)/Fr=2.80

Anisotropic model

(x/D)/Fr=0.71

(x/D)/Fr=1.60

(x/D)/Fr=2.80

Understanding the Turbulent Kinetic Energy Budget for IHVDT Modeling


Transport Production Gravity Dissipation TMF effects

Fr=2.38, B/D=1

Results for different geometries and speeds (not shown) also collapse with Fr and B/D scaling TKE production and dissipation peak near x~xB (and near secondary breaking downstream) but are not balanced due to increased transport Enhanced TKE transport in strong breaking region(s) suggests that gradient diffusion closure models will be inadequate

TKE Model Incorporating TMF Effect Captures Features of IHVDT

DNS Data

TKE model

Error = DNS - TKE model = 0.12

(x/D)/Fr=0.71

(x/D)/Fr=0.71

(x/D)/Fr=0.71

DNS Data

TKE model

Error = DNS - TKE model = 0.06

(x/D)/Fr=1.60

(x/D)/Fr=1.60

(x/D)/Fr=1.60

DNS Data

TKE model

Error = DNS - TKE model = 0.04

(x/D)/Fr=2.80

(x/D)/Fr=2.80

(x/D)/Fr=2.80

Accomplishments
Developed a database of high resolution iLES simulations of a canonical stern over a range of speeds and geometries. Identified large scale flow features in mixed region
New jet flow structure New vortex structure with Strouhal frequency ~0.2 Quantified convective velocity of large scale coherent structures

Characterized surface wake features


Convergent corner waves follow ballistic scaling Rooster tail height and length scale with speed and geometry

Characterized air entrainment


Developed Lagrangian method for tracking air entrainment Convergence and wake evolution of bubble density spectrum Identified regions of strong entrainment Quantified surface entrainment rate and its scaling with geometry and speed

Characterized the IHVDT in the wake. Developed isotropic and higher-order anisotropic closure models for turbulence mass fluxes (TMF) and Reynolds stresses.

Future Work
Develop high resolution DNS/iLES simulations for canonical air entraining turbulent flows. For different air entrainment scenarios/flow conditions:
Obtain characterizations, quantifications and models for air entrainment, and scaling with flow parameters Characterize the volume, sizes, locations of entrained air Extend Lagrangian method for tracking air entrainment to allow tracking of large numbers of small bubbles

Refine IHVDT closure modeling for turbulence mass fluxes (TMF) and Reynolds stresses
Further evaluation with different IHVDT canonical flows to understand dependency of model parameters A-posteriori validations

In addition to understanding the mixed region flow structure, we need to make sure the following is very clear in the summary:
X_c will tell you where the peak air entrainment is and we understand the scaling and limits of this finding (dry stern, waves collide before breaking) Q* tells you how much air is entrained at x_c and we have a basic (although not complete) understanding of this value

I.e. we have told them where and how much air is entrained within certain geometry conditions

Newly Identified Vortex Structure in Mixed Region of Breaking Waves


Instantaneous velocity vectors
Possible comparison to Juans paper.

Average velocity vector cuts at x locations


Motivates the jet structure and the physical reason behind the vortex structure

Average vorticity field Frequency analysis Consistent for both diverging breaking waves and centerline
This slide is the outline slide for the big point.

Surface features of the wake scale with geometry and speed


Confirmed ballistic scaling for sharp corner, dry transom sterns Define region of applicability

This slide is the outline slide for the big point.

Air entrainment in the wakes


Lagrangian cavity identification algorithm identifies entrained air and spray formation Volume of air entrained in the wakes
Something about max void fractions Something about entrainment fluctuations

Surface entrainment rate


Scales with velocity and geometry

This slide is the outline slide for the big point.

Formulas for the TMF and TKE models

First order eddy-diffusivity model

Second order eddy-diffusivity model

Turbulent kinetic energy model

DNS Data

Performance of Anisotropic vs Isotropic TMF Closure Models Importance of Incorporating Anisotropy


xc xr xd

(x/D)/Fr=0.71

(x/D)/Fr=1.60

(x/D)/Fr=2.80

Isotropic model

(x/D)/Fr

Model Error
Isotropic model

(x/D)/Fr=0.71

(x/D)/Fr=1.60

(x/D)/Fr=2.80

Anisotropic model

Anisotropic model

(x/D)/Fr

(x/D)/Fr=0.71

(x/D)/Fr=1.60

(x/D)/Fr=2.80

You might also like