gaskin chapter 3 ansys cfx

Upload: harley-souza-alencar

Post on 19-Oct-2015

119 views

Category:

Documents


0 download

DESCRIPTION

Quick Tutorial

TRANSCRIPT

  • 38

    3. CFD MODELING

    The geometry was modeled using the academically available 3D modeling software CATIA

    (V5R19). A 3D mesh was generated using the sweep and CFX mesh applications of ANSYS

    (12.0). The symmetry of the geometry was taken advantage of and only a quarter of the vehicle was

    modeled. This reduced the number of mesh elements and resulted in a lower computational time for

    the simulation.

    Figure 22 shows the geometry that was used for the simulation. The vertical thruster nozzle and

    horizontal thrusters were removed so an objective analysis could be performed on the

    hydrodynamic behaviour of the hull form.

    Figure 22: Underwater vehicle hull

    The CFD analysis was undertaken using ANSYS - CFX 12.0. The turbulence model used for the

    analysis was the standard k- and SST model. The convergence was restricted to within 10-4 and the

    computations were carried out on a HP Compaq dc7800 quad processor with 4.00 GB of RAM.

    3.1. MESH APPLICATION

    The 3-D hull was imported into the Design Modeler section of ANSYS. The domain that represents

    the water was created around the hull. It extends eight hull lengths behind, two lengths forward and

    eight widths on either side of the vehicle as shown by Figure 23. The size of the water domain has

    been chosen to reduce the effect the boundary conditions have on the underwater vehicle.

    All of the rectangular volumes of the geometry were meshed using the sweep method. ANSYS Inc

    (2009) states that a volume is not a sweepable volume if there is a completely contained internal

    void in the body, if the sweep application cannot find a source and target to sweep along, or if there

    is a hard sizing control used on the body that does not contain the same properties for the source and

  • 39

    the target. The sweep method produces a mesh containing mainly hexagons which are preferred by

    the CFX Solver.

    Figure 23: Mesh methods

    The volume containing the ROV hull, as illustrated by Figure 23, was meshed using the CFX

    method. This method uses a combination of tetrahedra, pyramids and prisms and is able to mesh

    irregular shapes and is most commonly used when meshing complex geometries.

    3.1.1. MESH PARTICULARS

    The mesh particulars for each volume, shown by Figure 23, are outlined by Table 5.

    Table 5: Mesh particulars

    Volume Vehicle hull Rectangular sections

    Method CFX (unstructured) Sweep (structured)

    Element form Tetrahedral Hexagonal

    Element size (mm) 4 - 25 25

    Inflation

    (boundary) layer

    20 layers and minimum

    height of 1 x 10-5

    mm for

    SST turbulence model

    N/A

    The interfaces between the sweep and CFX mesh have to be the same magnitude so ANSYS can

    transfer information accurately between the two mesh types. If there is a discontinuity between the

  • 40

    mesh interfaces then ANSYS will interpolate data between elements. This will cause errors to

    develop as the simulation is solved and will render the solution meaningless.

    Figure 24 shows the interfaces between the sweep and CFX mesh, for the simulation performed for

    this report. The connections line up so data can be transferred between nodes accurately with

    minimal amount of interpolation.

    Figure 24: Mesh connections

    The sweepable bodies have a bias on the mesh to increase the number of mesh elements closest to

    the vehicles hull.

  • 41

    3.2. CFX PRE

    Once the mesh was produced it was imported into CFX Pre where the settings for CFX Solve are

    applied. A run was performed using the k- turbulence model and a set of results was obtained that

    was used as initial conditions for subsequent runs. The turbulence model was then changed to SST

    to predict the turbulence around the hull more accurately.

    Figure 25 illustrates the boundaries applied to the water domain in CFX Pre. The walls that define

    the limit to the water domain are defined as being smooth walls. This reduces the drag effect they

    have on the simulation. The water enters through the inlet and exits through the outlet. The

    underwater vehicles hull is a non slip surface with a roughness of 1.5 x 10-2

    mm. This roughness

    represents the roughness of the fiberglass hull.

    Figure 25: CFX Pre

    The underwater vehicle is considered to be sufficiently submerged so each simulation was run in the

    steady state condition with no free surface effect. The inlet velocity of the water was varied from

    0.5 knots to 4 knots.

  • 42

    3.3. CFX SOLVER

    The mesh was partitioned into four, thus allowing a larger mesh to be produced. A definition file of

    about 4 x 106 elements was run using the AMC Research License and took up to 6 hours to

    converge using the SST turbulence model. All runs converged to within 10-4

    before the simulation

    stopped.

  • 43

    3.4. CFX POST

    Once the CFX Solver had produced a results file, it was imported into CFX Post to interpret the

    simulation. Images, data and other information relevant to specific areas of interest within the

    simulation were derived from CFX Post.

    3.4.1. PRESSURE DISTRIBUTION

    The pressure distribution, as shown in Figure 26, indicates that there exists a high pressure region at

    the forward section of the vehicle and a low pressure region at the aft section of the vehicle as

    predicted by Figure 13 in Chapter 2 of this report.

    Figure 26: Pressure contours

    3.4.2. VELOCITY DISTRIBUTION

    The velocity distribution, as shown in Figure 27, indicates that separation is occurring at the low

    pressure regions as illustrated by Figure 26. The separation is occurring because the fluid close to

    the surface of the hull has low momentum and is unable to overcome the pressure differential. This

    results in vortices forming in the areas indicated by Figure 27, which contribute to increasing the

    drag coefficient. This is consistent with Figure 11 of Chapter 2 of this report.

  • 44

    Figure 27: Velocity vectors

    3.4.3. BOUNDARY LAYER THICKNESS

    The boundary layer thickness was determined using the turbulent equation developed by Horner

    (1965) and is shown by Equation (3.1).

    (3.1)

    where

    = Total boundary layer thickness (m)

    x = Dimension is direction of flow (m)

    Rx = Reynolds number

    As Figure 28 shows, the boundary layer thickness is at a minimum before the bow of the vessel and

    increases as the flow travels towards the stern. The boundary layer is thicker for the lower speed,

    which is due to the lower momentum that causes the flow to separate earlier from the surface of the

    vehicle. At the operating speed of 2 knots, the boundary layer is 1 x 10-5

    mm thick at amidships.

    The following y+ section outlines the importance of predicting the boundary layer thickness.

  • 45

    Figure 28: Total boundary layer thickness

    3.4.4. Y-PLUS

    The SST turbulence model requires a mesh node close to the boundary layer to accurately predict

    the turbulence. Figure 29 shows the y+ distribution over the surface of the underwater vehicle. The

    y+ is below 2 as required when using the SST Turbulence model. It also shows the y+ decreases

    along the length of the vehicle which indicates that the boundary layer thickness is increasing as

    predicted by Figure 28.

    Figure 29: y+ distribution on surface of AMC II UV

  • 46

    3.4.5. GRID INDEPENDENCE

    Moinuddin and Thomas (2007) outline that when conducting a CFD analysis, it is essential to

    undertake a grid refinement process by gradually reducing the mesh size used for the analysis. It is

    usual to find that as the mesh size is reduced the results converge for a specific quantity of interest.

    Further reducing the mesh size has virtually no effect on the results. Once the result has converged

    it is known as the grid independent result. To be meaningful, any results obtained using a CFD

    package should be grid independent.

    The drag coefficient is useful to quantify and compare different hull forms and determine their

    hydrodynamic resistance. The force required for Equation (3.2) was derived from the CFD

    simulations.

    (3.2)

    where

    CD = Drag coefficient

    F = Force opposing direction of movement (N)

    = Density of fluid (kg/m3)

    = Velocity in direction of movement (m/s)

    A = Projected area of body (m2)

    The drag coefficient was calculated using Equation (3.2) and the convergence behaviour is shown

    by Figure 30. It shows the drag coefficient has converged with the 5,110,127 element mesh. The

    data used to develop Figure 30 is presented by Appendix 1.

    Figure 30: Grid independent result for the drag coefficient

  • 47

    3.4.6. VALIDATION

    Unfortunately, due to construction delays, model testing was not undertaken on the AMC II UV.

    Therefore, alternative methods for validating the CFD are presented.

    The AMC II UV, shown by Figure 31(b), is assumed to be that of an elongated cylinder. The

    Hoerner (1965) drag coefficient equation for an elongated cylinder is shown by Equation (3.3)

    where the diameter and length of the vehicle is 0.35 m and 0.856 m.

    (3.3)

    where

    d = diameter (m)

    l = length (m)

    The drag coefficient developed using the Hoerner equation for an elongated cylinder is similar to

    the 0.29 obtained using CFD on the AMC II UV.

    Experimental results for the AMC I UV, Figure 31(a), the NSTL AUV, Figure 32(a) and the

    REMUS AUV Figure 32(b) have also been used to validate the CFD for the AMC II UV. The drag

    coefficient at the Reynolds number of 7.62 x 105 is shown for each vehicle on Figure 31 and Figure

    32.

    Figure 31: AMC underwater vehicles

  • 48

    Figure 32: Other underwater vehicles

    The drag coefficient at different Reynolds number obtained by Saiju et al (2006) for the NSTL

    AUV, Prestero (2001) for the REMUS AUV and the AMC I and II UV are presented by Figure 33.

    As expected the drag coefficient for the AMC II UV is less than the AMC I UV for all Reynolds

    numbers. The AMC I UV has a flat side profile and cumbersome guardrails which contribute

    significantly to the drag of the vehicle. The drag of the NSTL and REMUS AUVs is expected to be

    less than the AMC II UV because they have a higher aspect ratio.

    Figure 33: Drag coefficient at different Reynolds numbers

  • 49

    3.5. DISCUSSION

    The NSTL and REMUS AUVs investigated by Saiju et al (2006) and Prestero (2001) have a high

    aspect ratio. Therefore, as shown by Figure 14 in Chapter 2, the dominant resistance force is due to

    skin friction. The main contributor to skin friction is the surface roughness of the body. As a result,

    the surface roughness of the underwater vehicles needs to be small.

    On the other hand, the low aspect ratio of the AMC II UV means that the dominant resistance force

    is due to form drag. The surface roughness still contributes to the drag coefficient as it creates a

    turbulent boundary layer and induces vortices. A material with a smooth surface finish will be

    suitable for construction of the hull. Fiberglass is a material that is cheap, readily available, easy to

    work with and has a smooth surface finish.

    A Reynolds number of 7.62 x 105 corresponds to a vehicle speed of 2 knots and a drag coefficient of

    0.29 for the AMC II UV. The cross sectional area of the vehicle is 0.1225 m2 and the density of the

    water is 1000 kg/m3. Thus, drag is calculated by Equation (3.4) as;

    (3.4)

    The two forward thrusters will be positioned in the horizontal direction developing 43.16 N which

    is sufficient to overcome the resistive force of 18.8 N at 2 knots.