Está en la página 1de 52
x UNIVERSITY “or ABERDEEN School of Engineering, University of Aberdeen 11.05.2012 Numerical Simulation of Seaweed and Flow Interaction Author: Michael Bruce Supervisor: Dr. Yakun Guo Student ID: 50907653 This thesis was submitted as part of the requirement for MEng Degree in Engineering Abstract This project looks at the hydrodynamic effects and velocity fields of different seaweed patches and densities. A numerical simulation was set up to represent seaweed hanging in a flow. When analysing the model, velocities were selected to coincide with the conditions occurring in Loch Etive, Scotland where is being considered as a location for Scotland's first seaweed farm. The computational fluid dynamics programme FLUENT was used to simulate results for this project. Arrays of cylinders were set up to see the change in velocity profiles and the drag coefficient across a range of velocities. Unfortunately due to complications of the software, a full range of results were unable to be obtained. This project found the highest drag coefficient occurred when there were more objects to accelerate around. Some models experienced velocity increases of cover 200% when flowing in-between cylinders. As the outcome of this project, it is recommended further research should be carried out looking into flow patterns of patch size and densities before a decision can be made on the best approach to maximising growth rate. Table of Contents Abstract List of Figures... ‘Symbols and Notations 1 Introduction... 1.1 Background and Justification 1.2 Aims and Objectives. 2 Literature Review. 2.1 Hydrodynamics of Seaweed..... 2.2. Flow around Smooth Single Cylinders ........ 2.3 Velocity Field.... 2.4 Patch Size and Density....... 3 Set-up of Numerical Simulation. 3.1 Introduction ... 3.2 Governing Equations... 3.3 Numerical Model........ 3.4 Geometry 3.4.1 Introduction... 3.4.2 Single Cylinder Model...... 3.4.3 Row of Cylinders Model 3.4.4 Column of Cylinders Model. 3.4.5 Offset Cylinders Model 3.4.6 Triangular Formation of Cylinders Model. 3.4.7 Square Formation of Cylinders Model .. 3.5 Meshing Method... 3.6 Quality of Meshing, 4 ANSYS-FLUENT.. 4.1 Introduction 42 Pressure-Based Solver .... 4.3 Spatial Discretization Scheme..... 4.4 Boundary Conditions... 4.5 Solution Initialization. 4.6 Model Validation, 5 Results. 5.1 Single Cylinder model... 5.2 Cylinders in a row....... 5.3 Cylinders in a Column... 5.4 Cylinders at an offset formation... 5.5 Cylinders in a triangular formation 5.6 Cylinders in a square formation 5.7 Drag Coefficients. 6 Discussion. 61 6.2 6.3 64 65 66 66 67 68 7 Recommendation for future work... 8 Conclusion References. Introduction ... Single Cylinder Model ... Row of Cylinders Model .. Column of Cylinders Model..... Offset Cylinders Model Triangular Formation of Cylinders Model ‘Square Formation of Cylinders Model... Drag Coefficients. Other data... oe 24 oe 25 1 26 26 27 oe 32 2 35 1 36 38 39 44 42 2 43 2 43 oe 44 2 45 45 46 110 46 46 247 List of Figures Figure 1 - Harvested seaweed [1]....... 8 Figure 2 - Single cylinder model geometry created on ANSYS Design Modeler soe 1B Figure 3 - Row of cylinders model geometry created on ANSYS Design Modeler eveeneeneeee 19 Figure 4 - Column of cylinders mode! geometry created on ANSYS Design Modeler . . . soe 1D Figure 5 - Offset cylinders model geometry created on ANSYS Design Modeler 20 Figure 6 - Triangular formation of cylinders model geometry created on ANSYS Design Modeler... 20 Figure 7 - Square formation of cylinders model geometry created on ANSYS Design Modeler... 121 Figure 8 - Mesh on x-z cross-sectional plane..... 22 Figure 9 - Cylinder drag coefficients at various Reynolds numbers... 28 Figure 10 - Comparison of simulated and measured time-averaged pressure coefficients at cylinders surface. 28 Figure 11 - Angle on cylinder surface 29 Figure 12 - Comparison of simulated and measured time-averaged stream-wise velocities at the symmetry line x/D=0.... 29 Figure 13 - Comparison of simulated & measured time-averaged stream-wise velocities along z/D=3..... ... 30 Figure 14 - Comparison of simulated & measured time-averaged stream-wise velocities along z/D=1 .... ... 30 Figure 15 - Single cylinder velocity vectors at 0.2m/s. 32 Figure 16 - Single cylinder velocity contours at 0.2m/s 32 Figure 17 - Single cylinder velocity vectors at 0.3 m/s 33 Figure 18 - Single cylinder velocity contours at 0.3m/s ... 33 Figure 19 - Single cylinder velocity vectors at 0.4 mis. 134 Figure 20 - Single cylinder velocity contours at 0.4 mis .. 134 Figure 21 - Row of cylinders velocity vectors at 0.2 mis. 35 Figure 22 - Row of cylinders velocity vectors at 0.3 mis... 1.35 Figure 23 - Row of cylinders velocity vectors at 0.4 mis... 36 Figure 24 - Column of cylinders velocity profile at 0.2 m/s 36 Figure 25 - Column of cylinders velocity vectors at 0.3 mis ... 37 Figure 26 - Column of cylinders velocity vectors at 0.4 m/s 37 Figure 27 - Offset formation of cylinders velocity vectors at 0.2 MIS -avu-uee- 38 Figure 28 - Offset formation of cylinders velocity vectors at 0.3 m/s 38 Figure 29 - Offset formation of cylinders velocity vectors at 0.4 MIS sacs: 39 Figure 30 - Triangular formation of cylinders velocity vectors at 0.2 m/s. 39 Figure 31 - Triangular formation of cylinders velocity vectors at 0.3 m/S........ 40 Figure 32 - Triangular formation of cylinders velocity vectors at 0.4 m/s... 40 Figure 33 - Square formation of cylinders velocity vectors at 0.2 MIS -vsese- 41 Figure 34 - Square formation of cylinders velocity vectors at 0.3 m/s au Figure 35 - Square formation of cylinders velocity vectors at 0.4 MIS .-.cueee- 42 Figure 36 - Comparison of drag coefficients for all models .... 142 List of Tables Table 1 - Flow regimes for smooth circular cylinder in an undisturbed flow......13 Table 2 - Mesh quality in regards to value of maximum skewness........ Table 3 - Mesh Quality for the models created Symbols and Notations A- Measure of area (m*) Ca - Coefficient of drag D - Diameter (m) f - Frequency (Hz) F4- Drag force (N) KC - Keulegan-Carpenter Number p - Density (kg/m?) Re - Reynolds number S;- Strouhal Number t- Time (s) T - Period of oscillation (s) U, - Free-stream velocity (m/s) v - Kinematic Viscosity (m*/s) 1 Introduction 1.1. Background and Justification Seaweed is a loose umbrella term for many types of marine algae. Seaweed can be used widely for a number of uses including food, medicinal purposes and even by farmers as a fertilizer. It has been grown purposely for food in China and Japan for over 2000 years. This is due to its availability, nutritional values and it's relatively simple cultivation. Today seaweed cultivation is a major economy in both Japan and China with 1,000,000 tonnes of dry seaweed being sold a year at a total value of $3,000 million. One area of seaweed that has not been researched extensively is its use as a biofuel. Only in recent years has this been suggested as a use, due to an increasing demand for alternate energy sources. Seaweed can be cultivated extremely fast compared to other stocks therefore would be an economically preferential source of fuel. Due to its relative low maintenance and quick growth rate, seaweed can be seen as an economically viable way of producing a crop that can be converted into a fuel that can help to cope with the earths increasing demands for energy Figure 1 — Harvested seaweed [1] Seaweed growth rate depends on three vital components: Light intensity in order to absorb photons for photosynthesis, all the nutrients required must be 8 available in the flow of water that passes, and a regular flow of water. The water motion is essential for plant growth as it cleans the plant, brings fresh nutrients and applies hydraulic forces that stimulate plant growth. The flow of water passing seaweed is a new topic and has not been studied specifically before. Other similar topics such as mussel farming have some parallels and can be used as good background information when looking at similar details. 1.2 Aims and Objectives This flow of water around the seaweed is what is to be investigated within this project, looking at the details of the velocity flow field, the interaction of flow and seaweed and the drag forces on the seaweed. After initial details are found on a singular piece of seaweed the model will then be expanded to multiple pieces of seaweed in different shapes and densities to investigate which patches are most suitable for gaining maximum growth rate. All these factors will be investigated using three-dimensional numerical software FLUENT. 2 Literature Review 2.1 Hydrodynamics of Seaweed When flow comes into contact with seaweed there are several hydrodynamic effects which must be considered. Due to the drag forces and low Young's Modulus there will be a reconfiguration of the seaweed. The drag forces will also result in a change in the velocity of the flow around, and at the lee-side of the sea plant. Sand-Jensen [2] describes the reconfiguration of five types of sea plants and confirmed several predictions, One is that with increasing the flexibility of the plant, the reconfiguration of the plant increases which, in tum, means a lower drag coefficient. This means that the lower the Young's modulus of the plant, the easier the plant would move with the flow and therefore would have a lower drag coefficient. Conversely, Sand-Jensen comments that with increasing velocities more flexible plants have a steeper decline in drag coefficient Another point Sand-Jensen states is that plants on a horizontally flat substratum bend over in the flow to attain a position of low drag For this project however it has been decided that the seaweed will be hanging from a structure on the surface of the sea. This means that in terms of Sand- Jensen findings, the plant would be inverted and the flow would pass under and around the plant as opposed to over and around. Sand-Jensen witnesses that with enough velocity the plant would reconfigure to be aligned with the horizontal plane; he proves this by carrying out experiments on 5 sea plants. As the water velocity increased, the bending angle always decreased. Denny and Gaylord [3] also describe their investigations into the mechanics of wave-swept algae. The authors go into more detail describing how as the seaweed bends not only because of its low stiffness but also due to the seaweed having a lower cross sectional area and also a long length of stipe (in ‘comparison to the length of the blade). These factors of low stiffness, low cross- sectional area and long length of stipe result in easy plant reconfiguration. The journal points to the stipe of the giant bull kelp Nereocystis leutkeana as an 10 example of how with a stiffness of SOMPa and small diameter (<1cm) acts more like a rope as opposed to functioning as a branch of a tree. Boller and Carrington [4] argue that streamlining is caused mainly due to frontal area change and there is only a little reduction in drag coefficient. The journal goes on to explain that the reconfiguration of macro algae is essential to the survival of the plant. Interaction of farms and hydrodynamics are important as culture installations affect passing waves and therefore there is the transfer of energy [5] from the flow to the structure. This transfer of energy is made up of several interacting forces, one being the total drag force acting on the object in the flow. This drag force can be assessed by the quadratic relationship: Fa = 5Capau,? (1) Where C,is a steady drag coefficient, p is the fluid density, A is a measure of Area and U, is the mean streamline velocity. Stevens et al [6] states that the choice of area is arbitrary but must be consistent with the choice of drag coefficient. Rearranging this formula we can gather that the Cy can be calculated by C= @) "AUS?" This shows that the drag coefficient will be influenced by the area which the force is projected on, the drag force of the body and the velocity at which the flow travels at. The total force on the structure will be a combination of the wave forces, buoyancy and connections between each patch. Buck and Buchholz [7] confirm that the length of the crop is the greatest surface area in terms of wet mass (in contact with fluid) therefore the crop will sustain the majority of the loading Demes et al [8] agree with the selection of the previous equations set for determining the drag forces of solid objects in a moving fluid. Demes et al [8] reverberates the concept that increasing the thickness of the blade will coincide 1 with the increasing flexural stiffness and will also, in turn, increase the breaking force of the crop Applying the above knowledge to this project would mean that with using thicker seaweed would result in the seaweed being able to resist higher hydrodynamic forces occurring from the waves and currents, 2.2 Flow around Smooth Single Cylinders When flow takes place around a solid body the boundary layer always separates from the body [9], the place at which this takes place is known as the separation point. Downstream of this point the flow is disturbed by large-scale eddys caused by the unsteady separation of flow over the body. This disturbance is known as the wake. The highly turbulent motion causes energy dissipation and this results in a reduction in pressure around the body thus causing an increase of pressure drag. The magnitude of this pressure drag depends on the size of the wake behind the body. Therefore the point of separation is pivotal in influencing the pressure drag around the body. The point of separation is dictated by the non dimensional, Reynolds number Re = (3) where D is the characteristic length of the cylinder (the diameter) and v is the kinematic viscosity of the fluid. For very low Reynolds numbers, the flow is of a laminar regime and does not apply to the current project. Fully turbulent flow occurs at extremely high values of Reynolds number, therefore the transition between laminar and turbulent consist of a series of transitional flow regimes. Zdravkovich [10] describes the regimes of flows and how they are characterised is briefly summarised below; In the first region after laminar flow, as the Reynolds number is increasing turbulence spreads upstream but the free-shear layers in the wake of the circular cylinder remain laminar. 12 The second region is when the free-shear layers in the wake become turbulent as the Reynolds number increases again The final region to become a turbulent flow is the cylinder boundary layer. The transition between laminar and turbulent first starts with turbulence at the separation point before then moving upstream then finally evoking the full cylinder, after this regime the flow is fully turbulent. Zdravkovich experimented with ranges of Reynolds numbers to see at which number each regime is achieved for a smooth circular cylinder, the results are shown in the table below: State_| Description Re Range L Laminar 0 to 180-200 Transition in the wake, elsewhere TW | laminar 180-200 to 350-400 Transition in the shear-layers, turbulent TrSL_ | wake 350-400 to 10°- 2 x 10° 105-2 x 10° to 6 x 10°-8 x TrBL | Transition in the circular boundary layers | 10° T Fully Turbulent Flow >Bx 10° ‘Table 2 -Flow regimes for smooth circular cylinder in an undisturbed flow, Zaravkovich [10] In the current project, the viscosity and the diameter will remain constant, only the change in velocity will have an effect on the Reynolds number. One factor that must also be mentioned is the characteristic shedding of eddies. This frequency of eddy shedding is denoted by the Strouhal number which is given by (4) where f is the frequency of vortex shedding. 13 2.3 Velocity Field A review by Sumer et al [11] discusses how placing a structure in a flow can have several effects. One of these effects is the resulting horseshoe vortex that can be created on the near side of the cylinder due to the rotation of the incoming flow. Vortices are also created behind the cylinder; these are known as lee-wake vortices. These vortices are created by the rotation of the boundary layer on the surface of the circular cylinder. The journal also noted that if the Keulegan-Carpenter number was below six then no horseshoe vortex would be present. Increasing Reynolds number means escalating the turbulence therefore the effect of vortices on the lee-side of the object becomes much greater. Sumer et al concludes that the horseshoe vortices exposed to waves are mainly governed by the Keulegan-Carpenter formula, vot Kc =", (6) where Up, T and D indicate the amplitude of the wave velocity, the wave time period and the cylinder diameter. In terms of velocity, Plew [12] investigates the effects of long-line mussel droppers on flows. One conclusion the report drew was that mussel farms caused a water velocity reduction between 47% and 67%. It should also be noted that the water was diverted horizontally around the structure as oppose to passing under the crop. Telfer [13] describes a hydrographical report on Loch Etive. Tefler shows how there is a change in flows at different depths in the Loch, with the highest velocities experienced near the surface. This report was comissioned for a client interested in starting up a fish farm. The regular flow of water through a fish farm is essential to bring nutrients to the farm and get rid of waste; the ‘same applies to a patch of seaweed. At the surface of the water Tefler states that the flow speed is rather slow and only varies from slightly above 0 mis to a rare maximum of 0.4 mis. Based on this report, these will be the range of 14 velocities that will be investigated in the current project. These velocities will give the flow a range of Reynolds numbers from 4000 to 8000, which according to Zdravkovich [10] will cause a transition in the shear layers meaning they are becoming turbulent. 2.4 Patch Size and Density Plew [12] also investigated the effects of multiple cylinders of different roughnesses to monitor what effect they had on the flow. It was found that drag increased with the spacing between the cylinders decreasing. Plew also shows that the drag coefficient of rougher cylinders increase at a higher rate than those of smooth cylinders, Peak turbulent Kinetic energy was also 49% higher ‘on rough cylinders than those on similarly spaced smooth cylinders. After the spacing experiments were carried out Plew moved on to testing different angles of placing the cylinders and configuring them in different orders. For double rows of cylinders it is calculated that the drag would be significantly higher. Also the direction of the line of mussels was investigated to see if that would have had an effect on the flow. 15 3 Set-up of Numerical Simulation 3.1 Introduction Guo, Zhang and Zhang [14] set up a numerical simulation looking at overlapping cylinders to see the interaction between turbulent flow and deep- sea marine structures. The structure was set up by placing a thinner diameter cylinder on top of a larger one to see what affects these shapes had on the velocity field. This study model used similar conditions as those required for this project will be therefore the information can be used as suggestion for the parameters to be set in ANSYS-FLUENT. The purpose of the numerical simulation is to show the interaction between seaweed and different velocities and patch shapes. 3.2 Governing Equations In order to separate the large scale and small scale motions the three dimensional, for incompressible flow the Navier-Stokes equation must be filtered. This low-pass filtering allows only non-small scales to pass, reducing the computational cost of the simulation. The governing equations are given by Rogallo and Moin [15] 5a 0, (6) ty = teil) — Ti, (8) Where, iis the velocity component of the resolved scales and Re is the Reynolds number. The filtered incompressible continuity equation (6) and the filtered Navier-Stokes equations of the resolved values is the result of filtering the continuity equation and Navier-Stokes equation respectively. The low-pass 16 filtered momentum equation (8) has a value 1,, for the non resolvable subgrid scale (SGS) code stresses. 14, describes the motion of smaller scaled structures on the Larger Eddy's. These SGS stresses are not known and therefore approximations are required. Using the Boussinesq hypothesis, the SGS turbulent stresses can be calculated by (9) where i, is the SGS eddy viscosity calculated as a function of the strain tensor for the resolved scale. Represented by; +) (10) The governing equations mentioned previously are solved using the proposed numerical model. 3.3. Numerical Model Countless numerical models are used for the simulation of turbulent flows. For this project it is difficult to find a comparable model of seaweed hanging from the surface of a flow. A cylinder shape is chosen to represent the seaweed as it reflects the vortex effects that will occur in the wake of the structure. A large eddy simulation (LES) is chosen due its low-pass filter which eliminates small scales in Navier-Scales equation mentioned previously. The commercial finite-volume CFD code, FLUENT 12.0, is selected to study the flow features around the formations of hanging circular cylinders. The incompressible Navier-Stokes equations are used along with the SGS Smogorinsky-Lilly in the LES. [16] 17 3.4 Geometry 3.4.1 Introduction Bodies were set up in way that the cylinders would hang from the surface of a channel with flow passing the body. The geometries were carried out in ANSYS. Design Modeler. A three dimensional model was created in each of the cases. All channel widths were of 0.1m and water depths of 0.3m. Each cylinder had a length of 0.25m so that calculations underneath the structure could also be taken into consideration. The diameter of the cylinder used was 002m which represents the thickness of a stipe of seaweed. The length of the tank modelled was 0.2m with the first cylinder placed proportionally closer to the velocity inlet allowing full velocity flow field to be seen behind the structure. A distance of 5mm is used between cylinders on the same axis. 3.4.2 Single Cylinder Model Figure 2 - Single cylinder model geometry created on ANSYS Design Mod! 18 3.4.3 Row of Cylinders Model Figure 3 - Row of cylinders model geometry created on ANSYS Design Modeler 3.4.4 Column of Cylinders Model Figure 4 - Column of cylinders model geometry created on ANSYS Design Modeler 19 3.4.5. Offset Cylinders Model Figure 5 - Offset cylinders model goometry created on ANSYS Design Modeler 3.4.6 Triangular Formation of Cylinders Model Figure 6 — Triangular formation of eylinders model geometry created on ANSYS Design Modeler 20 3.4.7. Square Formation of Cylinders Model Figure 7 - Square formation of cylinders model geometry created on ANSYS Design Modeler 3.5 Meshing Method Meshing was carried out with ANSYS meshing software. A triangular mesh was chosen as a quadrilateral mesh of acceptable quality was not available. A frozen body around the cylinder of 0.05m by 0.05m was chosen as a body sizing for a refined mesh. In this body sizing the element size used was 0.004m, Grids were also refined in the z-axis and x-axis of the body around the cylinder. This is done in order to achieve a more accurate view of the velocity field around and behind the cylinder. This outer refinement used was of element size 0.0075m. On the remaining body an element size used was of 0.009m. These elements sizes were chosen to ensure that the mesh would provide sufficiently accurate results. In order to improve the quality of results on the wall of the fluid, in contact with the cylinder, had a mapped face to ensure that readings could be taking from 360° around the cylinder at the same he ight. The mesh used can be seen below in figure 8, An unstructured mesh is used as a structured mesh is not available on the computers that are available for this project. The 21 unstructured mesh causes a greater number of cells to be chosen and therefore the quality of the mesh is sacrificed as the classroom computers can only process up to 1 milion cells. Figure 8 - Mesh on xz cross-sectional plane 3.6 Quality of Meshing The quality of meshing is essential to ensure an accurate analysis of the ‘simulation. A fine mesh will only produce more accurate data than a coarse mesh if the mesh elements are of the same or better quality. The quality of the mesh is defined by three important factors; the maximum cell squish, the maximum cell skewness and the maximum aspect ratio. The value of skewness can be compared to the following table to see the cell quality, 22 Value of Skewness Cell Quality 7 Degenerate 0.9-<4 Bad (sliver) 0.75-0.9 Poor 0.5-0.75 Fair 025-05 Good 30-025 Excellent 0 Equilateral Table Z- Mesh quality in regards to Value of maximum skewness [76] Maximum cell squish accounts for the compressed ratio of the cell. The worst cells have a value of close to 1, with the better quality cells tending to 0. The maximum cell squish is recommended to be under 0.95 for tetrahedral cells. Aspect ratio is the measure of how stretched the cell is. For flow simulation, like we have in the current project, it is important to keep the aspect ratio below 35:1 [16]. Number of | Maximum | Maximum ) Maximum | a aity of Model Elements | _ cel cell aspect | Mesh squish_| skewness ratio Single Cylinder | 477426 _| 0.7983 | 0.8358 18.68 Poor Row 318662 | 0.7617 | 0.8089 16.34 Poor Column 429583 | 0.7809 | 0.8072 18.71 Poor Offset 279780 | 0.8019 | 0.8289 16.93 Poor Triangle 350354 _| 0.7484 0.8168 15.62 Poor ‘Square 396973 | 0.7352 | 0.8146 16.17 Poor Table 3 - Mesh Quality or the models created Although the qualities of meshes above are classified as poor, the average skewness is below 0.75 and proves to be of much better quality. Ansys user guide states that is common for 3D models to experience some elements which have a skewness with a value of above 0.9 [17]. For this project this can be overlooked as a poor skewness value is only occurs in a minimal amount of elements. Using the average cell skewness value for each model the mesh is of fair quality according to table 3. 23 4 ANSYS-FLUENT 4.1 — Introduction ANSYS FLUENT is an adaptable Computational Fluid Dynamic (CFD) software which allows the simulation of heat transfer, model flow, turbulence and reactions, The software has high-performance computing capabilities and can model two-dimensional and three-dimensional structures capable of flows of turbulent, transient, laminar, incompressible, compressible, and steady behaviours. FLUENT is also capable of producing flows in a gas or liquid state while being able to edit properties of the fluid/solid. The ANSYS version used in this project is ANSYS 12 [16] ANSYS Workbench contains all the steps required for the simulation. The workbench houses individual components such as Design Modeler, Meshing and FLUENT. All of these components are used in the current project and linked to one another through the workbench interface. Having the Workbench interface means that the analyses can be kept in one place and makes different components easily accessible. ANSYS FLUENT is professional package which is used in industry as well as for research purposes. The software is supported by a users guide, theory manuals and technical assistance online [17]. 42 Pressure-Based Solver Although ANSYS FLUENT has two solvers, the pressure-based solver and the density-based solver, the pressure-based solver is chosen as the preferred solver for this project. In both solvers, the velocity field is derived from a pressure equation; this pressure equation is in turn derived from the continuity and momentum equations [17]. For the selected pressure-based coupled algorithm, the continuity and momentum equations are simultaneously solved Iterations will continue until the solution converges. A pressure-based coupled algorithm maybe used as oppose to a pressure-based segregated algorithm this 24 maybe preferable due to a faster rate of convergence, however this comes at a cost of computational expense. A pressure-velocity coupling method must be used in order to define the segregated algorithm used by FLUENT during its calculations. It is recommended that for a Large Eddy Simulation, pressure implicit with splitting of operator (PISO) is selected. The PISO pressure-velocity coupling is based on the higher order of approximation in terms of the correlation relationship between pressure and velocity. Limitations of the SIMPLE and SIMPLEC pressure-velocity coupling methods [18] are that the new velocities with their corresponding fluxes do not satisfy the momentum equation, PISO method overcomes this with a neighbour correction and a skewness correction The non-iterative time-advancement (NITA) scheme is used as this speeds up the transient simulation by making the splitting error of the same order as the truncation error. 43 Spatial Discretization Scheme Fluent stores discrete scalar values of the cells by default, however, the face scalar values of the cells are required to calculate convection terms in the governing equations of the project. These are important in determining the accuracy of the solution, For the discretization of the momentum equation, the bounded central difference scheme is automatically selected for LES in its view of low number diffusion. This selection is performed by default by FLUENT and is therefore seen as the best selection for the project. For the discretization of the Pressure equations, the PRESTO! option is used PRESTO! (PREssure Staggering Option) uses discrete continuity to stagger the control volume and therefore calculates a staggered pressure. All model settings were chosen with computational cost and accuracy in consideration to get the best balance between an accurate yet relatively quick solution. 25 4.4 Boundary Conditions In setting the boundary conditions it is imperative that they best resemble the actual scenario that will occur in the loch where the seaweed will be grown. If boundary conditions differ from actual conditions that are in the loch then the project research would not be valid. Boundary condition options vary dependent ‘on which boundary has been selected. For this project the velocity flow inlet is set at one boundary closest to the cylinder on the X-Y plane. From Tefler’s hydrographical report the velocities used in the project can range between 0 and 0.4 m/s for near surface fluid in Loch Etive, for the current project a range of velocities will be used of 0.2mis, 0.3m/s and 0.4m/s to see the velocity profiles at the higher velocities in the Loch. The opposite facing boundary is set as the flow outflow. No-slip wall boundary conditions are set for the solid surfaces. This applies to the cylinder surface and also the base of the model. Applying a wall boundary condition prevents fluid passing through it and therefore velocities of fluid would approach zero towards the wall. This is the reason that the lateral and top boundaries are set to symmetry conditions. Symmetry setting means that the boundary acts like a mirror where fluid velocity does not reduce but does not cross the boundary either. The symmetry condition is chosen for the lateral boundaries as the fluid in the loch would not have wall boundaries close by unless the seaweed was grown near a solid boundary (e.g. a rocky face). 4.5 Solution Initialization Solution initialization is vital in ensuring that initial conditions are similar to the values produced in the converged results. Solution initialization allows a definition of flow variables and therefore the software initializes the flow field to these values. Initial values are the values that are set for the beginning of the flow from that boundary. ‘Compute from’ allows the user to compute values from a particular zone, in terms of the current project all variables will be calculated from a velocity set at the velocity inlet. 26 Anon dimensional time-step (At U, / D) of 0.005 is used to simulate the full flow cycle in the model. Once these settings have been customised, the simulation can begin. After each iteration FLUENT reports residuals, which are the sum of conserved quantities, back to the user on a window which can also be used to view the drag convergence history of the model. Both graphs can be used to predict how the simulation is converging (or diverging) and can be used to observe initial values of drag of the models. 4.6 Model Validation Before results can be analysed the single cylinder model must be validated against results from existing literature this is to ensure that the model settings and under relaxation factors are correct. For the single cylinder model, a flow Reynolds numbers of 1.4 x 10° is used as is used in the experimental data of Cantwell & Coles [19]. The single model is also verified against experimental data concluded in the experiments mentioned above. In their project they experimented with a circular cylinder attached to the bottom surface of the tank, in this current project the cylinder is hanging from the water, this difference must be excused in verifying the model as there is a distinct lack of research investigating a hanging cylinder in a flow. This exception has been allowed due to the Plew [12] showing that fluid deflects around the patch and minimal flow goes under the cylinder. In the validation study the single cylinder should compare well against Cantwell & Coles well documented experiments. At this Reynolds number the drag coefficient of the single cylinder body is converged to 1.231 which coincides with the experimental data (Cs = 1.237) as shown below: 27 Key: é - Single Cylinder Co lo 1 to 1 1 tots? Re Figure 9 Cylinder drag coofficients at various Reynolds numbers Pressure Coefficient around Cylinder Surface edge Cantwell & Coles (1983) me current Project Degrees Figure 10 - Comparison of simulated and measured time-averaged pressure coefficients at cylinders surface The minimum C, occurred at 90° and 270° on the circle when positi oned with O° onto the oncoming flow as shown above. There is significant difference in the pressure coefficient at the lee-side of the cylinder which may cause slight errors 28 in the velocities of the measured data. This has been accounted for as the drag coefficient of each model will still be able to be gathered and compared. Figure 11 describes how the angle around the cylinder surface is obtained. Flow of water (oy Figure 11 - Angle on cylinder surface The calculated Strouhal number S; = 0.192 was slightly higher than the measured Strouhal number of S; =0.179, although it compares favourably with Plews [12] calculated value of S,= 0.19. Model Validation against Cantwell & Coles (1983) 12 ra os . wm current Projet 206 3 Cantwell & Coles (1983) oa 02 o 0 2 «6 8 4% Figure 12 - Comparison of simulated and measured time-averaged stream-wise velocities at the symmetry line x/D=0 29 Model Validation against Cantwell & Coles (1983) ‘© Cantwell & Coles (1983) Velocity U/U, 4 Current project, 02 _ 2 a ° 1 2 x/D Figure 13 - Comparison of simulated & measured time-averaged stream-wise velocities along 2/D=3 Model Validation against Cantwell & Coles (1983) oa Cantwell & Coles (1983) mCurrent Project, Figure 14 - Comparison of simulated & measured time-averaged stream-wise velocities along 2/D=1 From the previous Figures it can be seen that the simulated results and the measured results show similar patterns but differ in magnitude. Figure 12 is a plot of compared values of simulated and measure time averaged stream-wise 30 velocities along the symmetry line (WD = 0). These results show a similar tendency to rise fast and approach a maximum U/U,. For the measured results the time-averaged velocities in the symmetry line only reach 80% of the maximum stream-wise velocity but the simulated results approaches much closer to 100%. Figure 13 shows the simulated and measured time-averaged stream-wise velocities along z/D = 3. In this graph it is clear to see the behaviour of the velocity on the lee-side of the cylinder. In the simulated results the velocity achieves a maximum velocity of around 80% where in the measured tests, the lee-side velocities dropped to around 63%. Both results show a symmetric range of velocities about the centre line symmetry x/D=0. Figure 14 presents the simulated and measure time-averaged normal velocities along z/D = 1 symmetry plane. Large normal velocities can be seen at the lee- side of the cylinder, this conveys the strength of the vortices that should be created, Viewing Figures 12-14 it can be said that the LES settings chosen in FLUENT are not entirely accurate as the values do not match perfectly, however, since the graphs are of the similar trends they can be used as an estimate of the velocity flow field and drag coefficient of the selected models. 31 5 Results 5.1 Single Cylinder model Figure 15 — Single cylinder velocity vectors at 0.2m/s Figure 16 - Single cylinder velocity contours at 0.2m/s 32 Figure 17 - Single cylinder velocity vectors at 0.3 m/s Figure 18 - Single cylinder velocity contours at 0.3m/s 33, Figure 19 - Single cylinder velocity vectors at 0.4 m/s Figure 20 - Single cylinder velocity contours at 0.4 mis 5.2 Cylinders in a row Figure 21 ~ Row of cylinders velocity vectors at 0.2 mis. Figure 22 - Row of cylinders velocity vectors at 0.3 mis 35 Figure 23 - Row of cylinders velocity vectors at 0.4 mis 5.3 Cylinders in a Column Figure 24 - Column of cylinders velocity profile at 0.2 m/s Figure 25 - Column of cylindors volocity vactors at 0.3 mis Figure 26 - Column of cylinders velocity vectors at 0.4 mis 37 5.4 Cylinders at an offset formation Figure 28 - Offset formation of cylinders volocity vectors at 0.3 mis Figure 29 - Offset formation of cylinders velocity vectors at 0.4 mis 5.5 Cylinders in a triangular formation Figure 30 - Triangular formation of cylinders velocity vectors at 0.2 mis 39 Figure 32 - Triangular formation of cylinders velocity vectors at 0.4 mis 40 5.6 Cylinders in a square formation Figure 34 - Square formation of cylinders velocity vectors at 0.3 m/s, a Figure 35 - Square formation of cylinders velocity vectors at 0.4 mis 5.7 Drag Coefficients 3 25 g Pp? single Row Bis oe" = Column 5 1 Se oriset se Triangle os me Square ° ° 02 04 06 Free-Stream Velocity (m/s) Figure 36 - Comparison of drag coefficients for all models 42 6 Discussion 6.1 Introduction For the each cylinder a flow approached, a general pattern emerged. Flow would reduce in velocity as it advanced towards a cylinder and then divert around the body. The flow would increase in velocity around the cylinder achieving a maximum velocity at 90° and 270° coinci ding with the maximum pressure coefficients. The flow then decreases in velocity on the lee-side of the cylinder with a minimum velocity at 180° just behind the cylinder surface. For each of the different models tested, significantly different velocity patterns were apparent. It should be noted that all velocity vectors have been scaled up to be able to improve clarity of the velocity magnitudes. 62 Single Cylinder Model From Figure 15, a free-stream velocity of 0.2mis is used with the single cylinder model; the velocity profile shows a typical pattern without the vortices at the lee- side of the model. As the flow approaches the body the velocity is reduced to zero due to the affect of the boundary layer. The flow is diverted around the body and acceleration takes place. This acceleration is due to having to travel a further distance than the rest of the free-stream velocity. On the outside of the cylinder surface where maximum pressure occurs, the velocity increases to almost double the free-stream velocity. At maximum pressure the velocities are not symmetrical either side of the cylinder, this is probably due to the model being validated inaccurately. In Figure 16 the contours of the velocity field are shown. Minimum velocities are ‘shown towards the surface of the cylinder with bulbous contours emerging from the lee-side of the cylinder surface. It takes almost 2 full diameter lengths to return to free-stream velocity. Figures 17 and 19 show the vectors of free-stream velocity 0.3m/s and 0.4m/s respectively. The velocity patterns are similar to that of Figure 15. The velocity 43 reduces significantly at the frontal area of the cylinder and accelerates to a velocity twice the magnitude of U,. The velocity then reduces significantly at the lee-side of the cylinder where the wake should appear. In both figure 15 and 17 the velocity takes around 1.5 to 2 diameter length to return to U,, Figure 18 represents the contours of the velocities at U, = 0.3m/s on the single cylinder. The patter is similar to that of Figure 16. The flow is slow in front of the cylinder and at the lee-side, maximum velocities occur at the points closest to the tank walls. The velocity then accelerates once again to U,, like Figure 16 this occurs in bulbous-shaped contours. Figure 19 illustrates the contours of the velocity profile at a U, = 0.4m/s. This Figure shows a different pattern to that of Figures 18 & 19. Although the contours close to cylinder surface are similar, the outermost contour after the cylinder extends to the boundaries of the plane. 6.3 Rowof Cylinders Model In Figures 21, 22 & 23 the velocity profiles of the flow around the modelled row of cylinders are shown. Each model shows similar pattems where the flow slows at the front of the cylinders and increases in velocity at the maximum pressure locations on the cylinder. In the gap between the cylinders accelerating flow is up to 4 times the magnitude of U,. This is due to flow being diverted into this gap from flow approaching cylinders on the outside of the gap. Interestingly Figure 23 shows a different flow pattern in the wake of the cylinders. In Figure 21 & 22 the flow returns to U, almost immediately but in Figure 23 this is not the case. A gentle parabola shows the return to U,, which occurs approximately 1.5 diameter lengths behind the middle cylinder. This can be explained by an increase in turbulence, therefore it takes longer for the energy dissipate and resume normal flow. In Figure 21, 22 & 23 on the outside of the outer cylinders the velocity increases to twice the size of U,. This conveys a pattern where for each cylinder the flow has to divert around the velocity will increase by twice, meaning in a spacing 44 where there is only 1 cylinder to be diverted there will be an increase of double U,. For a gap with 2 cylinders there will be an increase of quadruple U,. 6.4 Column of Cylinders Model Figures 23, 24 & 25 show the velocity vectors of the model where the cylinders are in a column aligned parallel to the oncoming flow. A clear pattern emerges for all 3 velocities. As mentioned previously the flow slows to zero as it approaches the first cylinder then increases as it diverts round the body. The flow for these models continue with a 50% increased velocity then slows down when the spacing between cylinders occurs. The flow then increases again when diverting around the next cylinder in the column. This pattern repeats until the final cylinder when the flow diverts round the last cylinder, at the lee-side the flow slows, before increasing to U, within 1 to 1.5 diameter lengths, In the spacing between cylinders, the flow is minimal. Interestingly the flow further out from the cylinders is also increased by 25-30% even when up to 2 diameters length away on the x-axis. This shows that the current formation has an effect across the full flow and is not just specific to the relevant area in close proximity of the cylinders, 6.5 Offset Cylinders Model Figures 26, 27 & 28 show the velocity vectors of the model with cylinders in an offset formation. These figures showed similar patterns across all 3 free-stream velocities. The flow experienced an increase of 250%, this was noted in the spacing between the first row of cylinders. The flow velocity then occurs a reduction of 10-15% then increases once again to accelerate around the second row of cylinders. Behind the second set of cylinders was the greatest distance to return to free-stream, taking a length of approximately 1.5 diameter lengths. A bell- shape profile is formed beyond the final row in which the flow returns to U,. 45 6.6 Triangular Formation of Cylinders Model Figures 29, 30 & 31 display the velocity vectors of cylinders in a triangular formation. During this simulation, there was little acceleration as the flow diverted round the first cylinder; this did not match the pattern that emerged from the single cylinder simulations. This was because the flow did not have to divert dramatically to the next cylinder causing a smaller acceleration, Beyond the first cylinder, there was acceleration around the next row of 50% on the outer curves of the cylinders. A 100% increase in velocity occurred in the spacing between the 2 cylinders in the 2" row of the model. The flow stays relatively constant as it diverts round the 3" row of the model. Minimum velocities are seen directly behind the 3° row of cylinders at 180° on the cylinder surface. Flow returns back to free-stream velocity within half a diameter, showing minimal interference on the flow by the structure. 6.6 Square Formation of Cylinders Model Figures 32, 33 & 34 present the velocity vectors of cylinders arranged in a 3 x3 formation. During the simulation, the flow velocity increased 200% between cylinders, and then proceeded to decrease to U,, then increase 200% in each spacing of the cylinders as the flow accelerates. Minimal velocities are once again seen immediately in the lee-side wake of cylinders. A bell-shape curve shows the velocity returning to U,, the maximum distance taken to return to U, is 2 diameter lengths away from the middle cylinder in the final row. 6.7 Drag Coefficients Idealised drag coefficients were given by the drag convergence history on FLUENT. These may not be entirely accurate due to the mesh limitations but give a good indication of the trends and allow comparisons of drag coefficients between models with the various velocities. 46 The offset formation of seaweed shows the greatest change in drag coefficient with an increase of 0.4 when velocity is changed from 0.2m/s to 0.3m/s this is maybe due to the formation not being in a streamline formation as oppose to the triangular formation. Here velocities do not increase rapidly when there is little change to the direction of flow as it travels from one cylinder to another. Also in Figure 36, the trend for all models tends to be that as the velocity increases the drag coefficient increases slightly. This can be explained by the increase of turbulence of the flow as the Reynolds number increases. An increased turbulence in the wake of the flow means there will be a reduction in pressure and a resultant increase in pressure drag [9] The single cylinder model has the lowest drag coefficient, as the model experiences the lowest drag force. The square of cylinders model experiences the greatest drag coefficient. This is assumed to be because of the large wake that will be caused by the formation of cylinders. In terms of real life seaweed these drag coefficients would not represent the above data because of the reconfiguration of seaweed. As seaweed is put under larger velocities the seaweed, due to lack of stiffness, bends over with the flow, therefore reducing the frontal area of the seaweed [4]. This reduction in frontal area causes the seaweed drag coefficient to reduce with increasing velocities therefore would show a trend with an inverse slope of the one show in Figure 36. Limitations of the FLUENT software dictate that the stiffness of the seaweed cannot be modelled therefore the idealised drag coefficient is shown. 6.8 Other data Unfortunately due to the unstructured mesh, computer limitations and user inexperience the vortices known to be seen in the wake of a circular cylinder in a turbulent flow were not seen in the results. These vortices would provide the project with vorticity graphs and would enable the information on the effects of flow around the various models to be gathered. 47 7 Recommendation for future work Firsly, a project of this magnitude must be carried out on a computer with a large enough processor to run a finer mesh. This would enable the vortices to been seen in the wake of the project. Having a finer mesh would also ensure that the drag coefficients are of a much more accurate nature. For the current project countless hours were spent trying to correct this error on classroom computers with little success. CFD software is improving all the time and hopefully, in the near future, the stiffness and the change of frontal area of the seaweed will be able to be modelled, This development would allow the drag coefficient to be compared with measured data In terms of expansion of the current project, the model must be extended to include the structure that supports the seaweed. This change will greatly affect the forces around the seaweed due to the structure and also provide a force of its own including a gravitational force, drag through the flow and a pull on the seaweed. The structure of the seaweed carrier will also have to be able to resist change in wave height and also an inertia force. The investigation of forces on the structure would involve enough research to be a project in its own entity. Expanding on the size of the seaweed farm will cause the flow in the loch to react differently in terms of the flow behaviour towards the farm. Increasing the width of the farm would present an obstacle for the flow and a project could be created on the effects of full seaweed farm on the overall pattern of flow in the loch It has been suggested that the vast majority of the flow diverts around the farm as opposed to going under [12]. An investigation can be carried out to see whether shape and density affects this theory. This would also lead to being able to see how density affects the velocities in the centre of the patch. This is important as a regular flow of water is required to ensure the seaweed gets the nutrients it requires. This flow also ensures that the seaweed is moving and therefore has a greater chance of absorbing more photons from the sun. The 48 density is also an important factor in the aspect that if it is too high, seaweed stipes will overlap one another and therefore restrict its availability to exposing itself to the sun and to a passing flow. Another development that can be considered is the effect of roughness of the cylinders. For seaweed, the roughness of the skin would have an effect on the flow around the patch. Patches with different roughnesses could be tested in order to see how the velocity fields differ as oppose to this project where all the cylinders are of a smooth nature. One aspect that can also be considered would be to carry out experimental data on the velocity fields around the seaweed. This would provide relevant data in terms of drag coefficients and flow pattems which could be used to compare against numerical simulations done in CFD programmes such as FLUENT. 49 8 Conclusion A large eddy simulation was executed to study the interaction between turbulent flow and different arrangements of circular cylinders. The main velocity paths around the cylinders were simulated and results were discussed. Velocities dramatically increased when diverting around cylinders as acceleration occurs, this is because a longer path is followed than the free-stream velocity. This acceleration causes an increase in drag force and subsequently an increase in force coefficient when there are more cylinders to bypass around. From this project it can be concluded that the best suited velocity fields would be of a triangular formation or a column. This is due to their streamline attributes, as this causes little interference to the surrounding flow. Furthermore, extensive research needs to be carried out on the flow fields to ensure that all conditions are accounted for. As this project only accounts for the velocity flow fields around the seaweed bodies and not the influence of support structure. The project has not been completed to the full extents that the specification required. This was due to computers available having not large enough processors, a structured mesh package not being available on classroom computers and each model taking a high computational run time. Results obtained are achieved through the best means and resources available in the University Engineering department. 50 References [1] Farming Seaweed Could Be a Cheap Way to Produce Biofuel Photograph. New Microbe Turns Sugary Seaweed into Fuel. Bio Architecture Lab, 19 Jan. 2012 Web. 14 Feb 2012 . [2] SandJensen, Kaj. "Drag and Reconfiguration of Freshwater Macrophytes." Freshwater Biology 48.2 (2003): 271-83. [3] Denny, M., and B. Gaylor. "The Mechanics of Wave-swept Algae." The Journal of Experimental Biology 205 (2002): 1355-362. [4] Boller, M. L., and E. Carrington. "The Hydrodynamic Effects of Shape and Size Change during Reconfiguration of a Flexible Macroalga.” Journal of Experimental Biology 209.10 (2006): 1894-903. [5] Buck, B., and C. Buchholz. "Response of Offshore Cultivated to Hydrodynamic Forcing in the North Sea." Aquaculture 250.3-4 (2005): 674-91. [6] Stevens, C., D. Plew, N. Hartstein, and D. Fredriksson. "The Physics of Open-water Shellfish Aquaculture." Aquacultural Engineering 38.3 (2008): 145- 60. (7) Buck, Bela Hieronymus, and Corneila Maria Buchholz. "The Offshore- ring: A New System Design for the Open Ocean Aquaculture of Macroalgae." Journal of Applied Phycology 16.5 (2004): 355-68. [8] Demes, K. W., E, Carrington, J. Gosiine, and P. T. Martone. “Variation in Anatomical and Material Properties Explains Difference in Hydrodynamic Performances of Foliose Red Macroalgae (rhodophyta)."Phycological Society of America 47 (2011): 1360-367. 51 [9] Massey, B. S., and John Ward-Smith. "Chapter 8: Boundary Layers and Wakes." Mechanics of Fluids. Cheltenham, England: S. Thornes, 1998. 312-75 [10] | Zdravkovich, M. M. (1990). "Conceptual overview of laminar and turbulent flows past smooth and rough circular cylinders.” Journal of Wind Engineering and Industrial Aerodynamics 33: 53-62. [11] Sumer, B. M., N. Christiansen, and J. Fredsoe. "The Horseshoe Vortex and Vortex Shedding around a Vertical Wall-mounted Cylinder Exposed to Waves." The Journal of Fluid Mechanics 332 (1997): 41-70. [12] Plew, David R. "The Hydrodynamics of Long-line Mussel Farms." Thesis. University of Canterbury, 2005, [13] Telfer, T. (2008). Hydrographic report for a potential fish farm site in upper Loch Etive. Stirling: University of Stirling. [14] Guo, Y., J. Zhang, and L. Zhang. "Numerical Simulation of 3D Flow around an Overlapping Cylinder." Proceedings of the ICE - Maritime Engineering 163.2 (2010): 49-56 [15] _ Rogallo, R. S., and P. Moin. "Numerical Simulation of Turbulent Flows." Annual Review of Fluid Mechanics 16.1 (1984): 99-137 [16] ANSYS. ANSYS FLUENT 12.0 Meshing Help. ANSYS Inc., release 12.1 edition, November 2009c. [17] ANSYS. ANSYS FLUENT Theory Guide. ANSYS Inc., 13 edition, November 2010 [18] ANSYS. ANSYS FLUENT 12.0 User's guide. ANSYS Inc., 13 edition, April 2009a [19] Cantwell, Brian, and Donald Coles. "An Experimental Study of Entrainment and Transport in the Turbulent near Wake of a Circular Cylinder." Journal of Fluid Mechanics 136 (1983): 321-74. 52