cui force calculation

Upload: foad-farahani

Post on 04-Jun-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 Cui Force Calculation

    1/50

    Martin-Luther-UniversittHalle-Wittenberg

    Force calculation when particle nearcontact using lattice Boltzmann method

    M.Sc. Yan Cui

    Prof. Martin Sommerfeld

  • 8/13/2019 Cui Force Calculation

    2/50

    Martin-Luther-UniversittHalle-Wittenberg

    Outline

    Project description

    Lattice Boltzmann Method

    Particle near contact Force calculation

    Separate scheme

    Cut-back method

    Lubrication force

    Validations Numerical-Numerical

    Numerical-Experiment

    Simulation of my project

    Grid resolution

    Plug flow

    Shear flow

    Particle rotation under plug & shear flow

    What will do next?

  • 8/13/2019 Cui Force Calculation

    3/50

    Martin-Luther-UniversittHalle-Wittenberg

    Introduction to dry powder inhalator

    Nowadays different drugs are appliedthough inhalation.

    The drugs are in form of fine liquiddroplets or solid particles(powders).

    The size of the particles needs to berather smallin order to ensure thatthey are transported up to the alveoliof the lung.

    One solution: coating of larger carrierparticles with the fine agent particles.

    Adhesion forcebetween large andsmall particles can be adjusted bysurface treatment.

    Using LBMfor analyze the detachmentof the agent particles.

  • 8/13/2019 Cui Force Calculation

    4/50

    Martin-Luther-UniversittHalle-Wittenberg

    Sketch map of detachment

    dispersion

  • 8/13/2019 Cui Force Calculation

    5/50

    Martin-Luther-UniversittHalle-Wittenberg

    Objectives of the simulations

    Application of a high resolution Lattice-

    Boltzmann-Method with multi-level grid

    refinement

    Determination of the fluid dynamic forces acting

    on small agent particles (~ 5 m) attached to the

    surface of a larger carrier particle (~ 100 m)

    Identification of the detachment of agent particlesin dependence of adhesion forces, flow conditions and

    location on the surface

    Different flow conditions for fixed carrier particle:

    laminar and turbulent plug flow and shear flow

    Homogeneous isotropic turbulence

    Determination of the maximum adhesion force

    to guaranty agent particle detachment

    Flow conditions

    similar to those

    expected in an

    inhaler

    Results used for defined

    surface modification

  • 8/13/2019 Cui Force Calculation

    6/50

    Martin-Luther-UniversittHalle-Wittenberg

    Lattice Boltzmann Method

    The Lattice-Boltzmann-Method simulates the motion of a fluid on a microscopic levelsolving the discretised distribution function f(x, v, t), describing the number of fluid

    elements at a given location and time having the velocity v.

    macroscopic flow system

    Discretised Lattice-Boltzmann-Equation (single relaxation time, BGK):

    ),(),(),(),( )0( txftxfttxftttvxf iiiii

    Collision termPropagation term

    0

    2

    2

    2

    2

    2

    2

    2

    2

    2

    2

    1

    1

    1

    1

    1

    Discrete velocity directions in the D3Q19-model

    Bounce-back boundary

    condition for curved walls

    Momentum exchange yields

    forces on the particle

    equilibrium Maxwell distribution)t,x(f )0(i

  • 8/13/2019 Cui Force Calculation

    7/50

    Martin-Luther-UniversittHalle-Wittenberg

    Particle force calculation

    Particle force = Force(fluid nodes near particle surface)

    Solid nodes

    Fluid nodes

    Fluid nodes near particle surface

  • 8/13/2019 Cui Force Calculation

    8/50

    Martin-Luther-UniversittHalle-Wittenberg

    Separate scheme

    When two particles are near contactless than one grid spacing, the codeitself will recognize two particles as

    one particle. Because there are notenough fluid nodes surround oneach particle.

    One solution: not only accumulatingthe fluid nodes near particle surface,

    but also accumulates those solidnodes near another particle surface.

    Defect: those solid nodes do nothave any fluid property, during theaccumulation, those nodes will

    become a vacuum which may causetremendous force.

  • 8/13/2019 Cui Force Calculation

    9/50

    Martin-Luther-UniversittHalle-Wittenberg

    Separate scheme

    A modified approach: we give thosesolid nodes near another particlesurface the equivalent equilibrium

    equations, and set the velocityconstant(equal to the particlevelocity).

    After that, the very large forcedisappeared, we can get a more

    reasonable force for each particle.

    Is it enough?

    Solid nodes near another particle surface

    Fluid nodes near particle surfaceFluid nodes

    Solid nodes

  • 8/13/2019 Cui Force Calculation

    10/50

    Martin-Luther-UniversittHalle-Wittenberg

    Surfaces near contact

    N.-Q. Nguyen and A. J. C. Ladd in 2002 [1] find that whentwo particle surfaces come within one grid spacing, fluidnodes are excluded from regions between the solid surfaces,

    leading to a loss of mass conservation.

    =

    = 0

    Although the sums and are zero for any

    closed surface, when two particles are close to contact some

    of the boundary nodes are missing and the surfaces are nolonger closed. In this case 0and mass conservation isno longer ensured.

    If the two particles move as a rigid body, mass conservationis restored. But for soft matter systems, the leak of mass

    conservation should be taken into consideration. Solution: Enforce mass conservation, particle-by-particle, by

    redistributing the excess mass among the boundary nodes.[1] N.-Q. Nguyen and A. J. C. Ladd, Lubrication corrections for lattice-Boltzmannsimulations of particle suspensions, Physical Review, E 66, 046708 (2002).

  • 8/13/2019 Cui Force Calculation

    11/50

    Martin-Luther-UniversittHalle-Wittenberg

    Cut-back-method

    Later Burkhard Dnweg and A. J. C. Ladd [2] write that analternative idea is to ensure that there is always at least onefluid node in the gap between the particle surfaces.

    E-Jiang Ding and Cyrus K.Aidun in 2003 [3] introduce a cutback method for dealing with particle near contact. They turnthe blue nodes (solid nodes) into fluid nodes, so the shape ofparticle is maintained by the red line.

    [2] B. Dnweg and A. J. C. Ladd, Lattice Boltzmann simulations of soft matter systems,Advances in Polymer Science, 221:89-166 (2009).

    [3] E-Jiang Ding and Cyrus K. Aidum, Extension of the lattice-Boltzmann method for directsimulation of suspended particles near contact, J.S.P, Vol. 112, Nos. 314 (2003).

  • 8/13/2019 Cui Force Calculation

    12/50

    Martin-Luther-UniversittHalle-Wittenberg

    Modified cut-back-method

    An modified approach is:

    Cut only one particle, retain the full geometry of the other one.

    For the cut-particle, we use an extrapolation bounce back

    boundary condition to give back some part of the geometry.

    Firstly, turn those blue nodes into fluid nodes. Secondly, measure the distance between the green nodes

    and the original surface, set it as q (0 1 grid spacing).

    Thirdly, travel to the reversed direction, and only travel 1-q

    grid spacing, so now the geometry should be on the yellowline

  • 8/13/2019 Cui Force Calculation

    13/50

  • 8/13/2019 Cui Force Calculation

    14/50

  • 8/13/2019 Cui Force Calculation

    15/50

    Martin-Luther-UniversittHalle-Wittenberg

    Numerical-numerical validation

    Robyn Larsen, Dmitry Eskin, Jos Derksen, Liftand drag onagglomerates attached to walls,ICMF 2010.

    In the first part of this paper, they study the lift and dragforce acting on the particle attached to the wall, with theshear flow and the immersed boundary condition in LBM.

    We use the variable-distance-bounce-back boundarycondition, so its a good validation between these twonumerical methods.

    Immersed boundary LBM:

    Treat the boundary as deformable with high stiffness.

    Small distortion yield a force to restore the boundary intooriginal shape.

    The body force term is used to mimic the presence of boundary.

  • 8/13/2019 Cui Force Calculation

    16/50

    Martin-Luther-UniversittHalle-Wittenberg

    Problem definition

    Variable to get: dimensionless lift and drag force.

    Changed variable: domain size, length, width and

    height(comparable to the diameter of particle). Fixed Re number: 0.012.

    Analytical solution of Leighton & Acrivos: L*=9.22, F*=32.1.

    PE

    R

    I

    O

    D

    I

    C

    Moving Wall

    Non-Slip

    PE

    R

    I

    O

    D

    I

    C

  • 8/13/2019 Cui Force Calculation

    17/50

    Martin-Luther-UniversittHalle-Wittenberg

    Results

  • 8/13/2019 Cui Force Calculation

    18/50

    Martin-Luther-UniversittHalle-Wittenberg

    Numerical-Experiment validation

    Stationary particle in a wall-bounded linear shear flow.

    The problem of lift force on a particle sitting on the wall in aboundary layer has been considered experimentally.

    Hall and Mollinger and Nieuwstadt measured lift force on asmall stationary particle in contact with the wall. In all casesconsidered the particle was sufficiently small to be entirely inthe viscous sublayer of the turbulent boundary layer. Owingto turbulence, the lift force was fluctuating in time and they

    obtained time-averaged lift coefficient

    for varying particle size.

    Later Muthanna obtained lift force on

    a stationary spherical particle attached

    to a wall in a laminar linear shear flow. The present results for the stationary

    particle touching the wall in a linear

    shear flow can be compared with these experimental results.

  • 8/13/2019 Cui Force Calculation

    19/50

    Martin-Luther-UniversittHalle-Wittenberg

    Simulation-Experiment validation

    In the beginning, I introduce non-dimensional lift force, +,

    and particle radius, +, as follows:

    + =

    + =

    , is the shear velocity

    Initial condition:

    Two-level-grid-refinement

    20 cells along the particle diameter

  • 8/13/2019 Cui Force Calculation

    20/50

    Martin-Luther-UniversittHalle-Wittenberg

    i l i i lid i

  • 8/13/2019 Cui Force Calculation

    21/50

    Martin-Luther-UniversittHalle-Wittenberg

    Simulation-Experiment validation

    The lift coefficients obtained by Hall and Mollinger areconsiderably higher than those obtained in our simulation.

    It should be noted that the measurements of Hall and

    Mollinger are for a particle sitting on a wall in a turbulentboundary layer. Nevertheless the orders of magnitude higherlift force measured in their experiments cannot be explainedby the present simulations.

    Experiments of Muthanna in a linear shear flow are more

    relevant to the present simulations. Their lift coefficients areconsiderably smaller than those measured in a turbulentboundary layer, but they are still larger than the computedlift coefficient. As pointed out by these authors, accuratemeasurement of lift force on small particles in wall-bounded

    flows is a challenging problem.

  • 8/13/2019 Cui Force Calculation

    22/50

    G id l ti

  • 8/13/2019 Cui Force Calculation

    23/50

    Martin-Luther-UniversittHalle-Wittenberg

    Grid resolution

    Change the grid resolution on the agent particles => 6 gridspacing along the diameter

    G id l ti

  • 8/13/2019 Cui Force Calculation

    24/50

    Martin-Luther-UniversittHalle-Wittenberg

    Grid resolution

    Change the domain length x => 60 grid spacing

    G id l ti

  • 8/13/2019 Cui Force Calculation

    25/50

    Martin-Luther-UniversittHalle-Wittenberg

    Grid resolution

    Change the domain length y/z => 50 grid spacing

    G id l ti

  • 8/13/2019 Cui Force Calculation

    26/50

    Martin-Luther-UniversittHalle-Wittenberg

    Grid resolution

    Pa amete st d

  • 8/13/2019 Cui Force Calculation

    27/50

    Martin-Luther-UniversittHalle-Wittenberg

    Parameter study

    Force vector (plug flow)

  • 8/13/2019 Cui Force Calculation

    28/50

    Martin-Luther-UniversittHalle-Wittenberg

    Force vector (plug flow)

    Plug flow

  • 8/13/2019 Cui Force Calculation

    29/50

    Martin-Luther-UniversittHalle-Wittenberg

    Plug flow

    Plug flow

  • 8/13/2019 Cui Force Calculation

    30/50

    Martin-Luther-UniversittHalle-Wittenberg

    Plug flow

    Plug flow

  • 8/13/2019 Cui Force Calculation

    31/50

    Martin-Luther-UniversittHalle-Wittenberg

    Plug flow

    Plug flow

  • 8/13/2019 Cui Force Calculation

    32/50

    Martin-Luther-UniversittHalle-Wittenberg

    Plug flow

    Plug flow

  • 8/13/2019 Cui Force Calculation

    33/50

    Martin-Luther-UniversittHalle-Wittenberg

    Plug flow

    Plug flow

  • 8/13/2019 Cui Force Calculation

    34/50

    Martin-Luther-UniversittHalle-Wittenberg

    Comparing different Re:

    Plug flow

    Plug flow

  • 8/13/2019 Cui Force Calculation

    35/50

    Martin-Luther-UniversittHalle-Wittenberg

    Comparing different size of agent particles:

    Plug flow

    Diameter_of_Agent_Particle => Feffect. Because 1 N

    = 1 kgm/s^2, the force is relate to the mass of particles.

    Plug flow

  • 8/13/2019 Cui Force Calculation

    36/50

    Martin-Luther-UniversittHalle-Wittenberg

    Comparing different coverage rate:

    Plug flow

    Plug flow

  • 8/13/2019 Cui Force Calculation

    37/50

    Martin-Luther-UniversittHalle-Wittenberg

    Comparing different coverage rate:

    Plug flow

    Coverage_Rate => Band_of_Feffect. This is because of the

    gap between agent particles. If the gap is too large, the flowdisturbance between particles will increase; if the gap is tiny, the

    flow will mostly influence on the top of agent particles, but less

    impact in the gap, because there are less space for the flow.

    Force vector (shear flow)

  • 8/13/2019 Cui Force Calculation

    38/50

    Martin-Luther-UniversittHalle-Wittenberg

    Force vector (shear flow)

    Shear flow

  • 8/13/2019 Cui Force Calculation

    39/50

    Martin-Luther-UniversittHalle-Wittenberg

    Shear flow

    Shear flow in 3D and its projection on

    Feffect-x and Feffect-z:

    Shear flow

  • 8/13/2019 Cui Force Calculation

    40/50

    Martin-Luther-UniversittHalle-Wittenberg

    Shear flow

    Shear flow in Feffect-x projection:

    Shear flow

  • 8/13/2019 Cui Force Calculation

    41/50

    Martin-Luther-UniversittHalle-Wittenberg

    Shear flow

    Comparing plug flow and shear flow in

    Feffect-x projection(same Re):

    Shear flow

  • 8/13/2019 Cui Force Calculation

    42/50

    Martin-Luther-UniversittHalle-Wittenberg

    Shear flow

    Shear flow in Feffect-z projection:

    Shear flow

  • 8/13/2019 Cui Force Calculation

    43/50

    Martin-Luther-UniversittHalle-Wittenberg

    Shear flow

    Comparing different Re with shear flow in

    Feffect-z projection:

    Shear flow

  • 8/13/2019 Cui Force Calculation

    44/50

    Martin-Luther-UniversittHalle-Wittenberg

    Shear flow

    Comparing different Re of Shear flow 3D:

    Particle rotation

  • 8/13/2019 Cui Force Calculation

    45/50

    Martin-Luther-UniversittHalle-Wittenberg

    Particle rotation

    1

    2 3

    4

    8 5

    7 6

    1

    2 3

    4

    Particle rotation

  • 8/13/2019 Cui Force Calculation

    46/50

    Martin-Luther-UniversittHalle-Wittenberg

    Comparing different Re under

    Rotation_Plug flow:

    1

    2 3

    4

    Particle rotation

  • 8/13/2019 Cui Force Calculation

    47/50

    Martin-Luther-UniversittHalle-Wittenberg

    1

    2 3

    4

    8 5

    7 6

    Rotation_Shear flow:

    Particle rotation

  • 8/13/2019 Cui Force Calculation

    48/50

    Martin-Luther-UniversittHalle-Wittenberg

    1

    2 3

    4

    8 5

    7 6

    Comparing Rotation_Plug flow and

    Rotation_Shear flow:

    What will do next?

  • 8/13/2019 Cui Force Calculation

    49/50

    Martin-Luther-UniversittHalle-Wittenberg

    Form of multi-layer agent particles

    Problem: the form of agglomerate using the tree structure. Soevery particle has only one father. However, in multi-layer,

    the particle from second layer may have 2 or 3 contact to thefirst layer particles, which means it has 2 or 3 fathers. This isnot allowed during modeling.

    Turbulent effect

    Theoretically in LBM, the velocity should be

  • 8/13/2019 Cui Force Calculation

    50/50

    Thanks for attention!