James Grimshaw
Acknowledgement
I would like to thank my advisor D T Craft for all his help during this project, without his assistance
none of the work done in this project would have been possible. I would also like to thank my family
and friends for all their support during this process.
i|Page
James Grimshaw
Declaration
I James Grimshaw declare that the work presented in this report is all my own work. All work that
has been taken from other papers or projects has been correctly referenced. If any work is found
that has not been referenced contact me immediately and it will be amended immediately.
ii | P a g e
James Grimshaw
Table of Contents
Acknowledgement ................................................................................................................................... i
Declaration .............................................................................................................................................. ii
Table of Figures ....................................................................................................................................... v
Table of tables ........................................................................................................................................ vi
1.
Summary ......................................................................................................................................... 1
3.
Introduction .................................................................................................................................... 2
4.
3.1.
3.2.
4.2.
Previous Literature.................................................................................................................. 5
5.
6.
6.1.1.
6.2.
6.2.1.
6.2.2.
6.2.3.
6.2.4.
6.3.
7.
6.3.1.
6.3.2.
7.1.1.
7.1.2.
7.1.3.
7.1.4.
7.1.5.
7.2.
7.2.1.
7.2.2.
James Grimshaw
7.2.3.
8.
7625805
Conclusion ..................................................................................................................................... 40
8.1.
Conclusions of Results........................................................................................................... 40
8.1.1.
8.1.2.
8.2.
9.
Recommendations ........................................................................................................................ 42
9.1.
9.2.
Modelling in 3D ..................................................................................................................... 42
10.
Bibliography .............................................................................................................................. 43
11.
Appendices ................................................................................................................................ 45
11.1.
Appendix A ........................................................................................................................ 45
11.2.
Appendix B ........................................................................................................................ 47
11.2.1.
11.2.2.
11.3.
Appendix C ........................................................................................................................ 50
11.4.
Appendix D ........................................................................................................................ 52
iv | P a g e
James Grimshaw
Table of Figures
Figure 1 - Photo of the Baden Baden (NNoN, 2012) ............................................................................................... 4
Figure 2 - Photo of the Barbara (SDTB, 2010) ......................................................................................................... 5
Figure 3 - Photo of the E-Ship 1 (Floch, 2011) ........................................................................................................ 5
Figure 4 - The Magnus effect (Celli, 1997) .............................................................................................................. 9
Figure 5 - Original mesh used for single cylinder case .......................................................................................... 11
Figure 6 - Final refined mesh used for single cylinder case .................................................................................. 11
Figure 7 - Plot of wall Yplus across the cylinder wall in the original mesh bias factor of 460 .............................. 13
Figure 8 - Plot of wall Yplus across the cylinder in the original mesh bias factor of 160 ...................................... 13
Figure 9 - Plot of wall Yplus across the cylinder wall in the final refined mesh .................................................... 13
Figure 10 - Contour of turbulent kinetic energy with original settings ................................................................ 15
Figure 11 - Contour of turbulent kinetic energy with revised time step .............................................................. 15
Figure 12 - Plot of coefficient of lift against the flow time for the with the revised time step ............................ 16
Figure 13 - Mesh after 2nd refinement for the single cylinder case .................................................................... 17
Figure 14 - Comparisons of the coefficient of lift for various meshes for the single cylinder case ...................... 17
Figure 15 - Geometry used for preliminary simulations the for two cylinder case .............................................. 19
Figure 16 - Mesh used for preliminary simulations for the two cylinder case ..................................................... 19
Figure 17 - Geometry used for final simulations for the two cylinder case.......................................................... 20
Figure 18 - Mesh used for final simulations for the two cylinder case ................................................................. 20
Figure 19 - Contour of turbulent kinetic energy at =1.5 for single the cylinder case ......................................... 22
Figure 20 - Contour of turbulent kinetic energy at =2 for the single cylinder case ............................................ 22
Figure 21 - Contour of turbulent kinetic energy at =4.5 for the single cylinder case ......................................... 22
Figure 22 -- Contour of turbulent kinetic energy at =5 for the single cylinder case ........................................... 22
Figure 23 - A graph showing the coefficient of lift for various spin ratios for the single cylinder case ................ 23
Figure 24 - Contour of turbulent kinetic energy at T/D=2 with flow at 90 for the two cylinder case ................ 25
Figure 25 - Contour of turbulent kinetic energy at T/D=4 with flow at 90 for the two cylinder case ................. 25
Figure 26 - Contour of turbulent kinetic energy at T/D=2 with flow at 45 for the two cylinder case ................. 25
Figure 27 - Contour of turbulent kinetic energy at T/D=4 with flow at 45 for the two cylinder case ................. 25
Figure 28 - Plot showing the coefficient of lift around two cylinders at T/D=2 with flow at 90 ......................... 26
Figure 29 - Plot showing the coefficient of drag around two cylinders at T/D=2 with flow at 90 ...................... 26
Figure 30 - Plot showing the coefficient of lift around two cylinders at T/D=4 with flow at 90 ......................... 28
Figure 31 - Plot showing the coefficient of drag around two cylinders at T/D=4 with flow at 90 ...................... 28
Figure 32 - Plot showing the coefficient of lift around two cylinders at T/D=2 with flow at 45 ......................... 29
Figure 33 - Plot showing the coefficient of drag around two cylinders at T/D=2 with flow at 45 ...................... 30
Figure 34 - Plot showing the coefficient of lift around two cylinders at T/D=4 with flow at 45 ......................... 31
Figure 35 - Plot showing the coefficient of drag around two cylinders at T/D=4 with flow at 45 ...................... 31
Figure 36 - Contour of turbulent kinetic energy for the final simulations at T/D=3 with the inlet flow at 90 .... 33
Figure 37 - Contour of turbulent kinetic energy for the final simulations at T/D=4 with the inlet flow at 90 .... 33
Figure 38 - Contour of turbulent kinetic energy for the final simulations at T/D=3 with the inlet flow at 45 .... 33
Figure 39 - Contour of turbulent kinetic energy for the final simulations at T/D=4 with the inlet flow at 45 .... 33
Figure 40 - A plot showing the coefficient of lift around two cylinders for the final simulations at T/D=3 with
flow at 90............................................................................................................................................................. 34
Figure 41 - A plot showing the coefficient of drag around two cylinders for the final simulations at T/D=3 with
flow at 90............................................................................................................................................................. 34
Figure 42 - A plot showing the coefficient of lift around two cylinders for the final simulations at T/D=4 with
flow at 90............................................................................................................................................................. 35
Figure 43 - A plot showing the coefficient of drag around two cylinders for the final simulations at T/D=4 with
flow at 90............................................................................................................................................................. 35
v|Page
James Grimshaw
Figure 44 - A plot showing the coefficient of lift around two cylinders for the final simulations at T/D=3 with
flow at 45............................................................................................................................................................. 36
Figure 45 - A plot showing the coefficient of drag around two cylinders for the final simulations at T/D=3 with
flow at 45............................................................................................................................................................. 36
Figure 46 - A plot showing the coefficient of lift around two cylinder for the final simulations at T/D=4 with
flow at 45............................................................................................................................................................. 37
Figure 47 - A plot showing the coefficient of drag around two cylinder for the final simulations at T/D=4 with
flow at 45............................................................................................................................................................. 37
Figure 48 - Mesh used in turbulent pipe flow tutorial .......................................................................................... 45
Figure 49 - Velocity vector for turbulent pipe flow tutorial ................................................................................. 45
Figure 50 - Pressure contour for turbulent pipe flow tutorial .............................................................................. 45
Figure 51 - Plot of the wall Yplus for turbulent pipe flow tutorial ........................................................................ 45
Figure 52 - Velocity vector for laminar flow around a cylinder tutorial ............................................................... 46
Figure 53 - Contour of turbulent kinetic energy for laminar flow around a cylinder tutorial .............................. 46
Figure 54 - Conotur of turbulent kinetic energy for turbulent flow around a cylinder tutorial ........................... 46
Figure 55 plot of the wall Yplus for turbulent flow around a cylinder tutorial .................................................. 46
Figure 56 - Contour of the cell courant number for the single cylinder case ....................................................... 47
Figure 57 - Contour of the cell courant number for the two cylinder case, preliminary mesh T/D=2 ................. 47
Figure 58 - Contour of the cell courant number for the two cylinder case, preliminary mesh T/D=4 ................. 48
Figure 59 - Contour of the cell courant number for the two cylinder case, final mesh T/D=3 ............................. 48
Figure 60 - Contour of the cell courant number for the two cylinder case, final mesh T/D=4 ............................. 49
Table of tables
Table1 Table showing aspect ratio and orthogonal quality of the chosen mesh
11
Tabl3 2 Table showing the effect of refining the mesh on the results
16
Table 3 Table demonstrating the quality of the meshes used for the various spacing ratio
19
Table 4 Table demonstrating the quality of meshes for the final simulations
21
Table 5 Table showing the coefficient of lift and drag for the various spacing ratios and flow anglespreliminary mesh
32
Table 6 Table showing the coefficient of lift and drag for the various spacing ratios and flow angles
final mesh
38
Table 7 Gantt chart demonstrating how time was planned and managed during the project in
semester 1
50
Table 8 Gantt chart demonstrating how time was planned and managed during the project in
semester 2
51
vi | P a g e
James Grimshaw
1. Summary
In recent years ship designers have been looking for possible methods to increase the efficiency of
their ships. Many different means have been looked into; the use of LNG as a fuel, more
aerodynamic hulls and more efficient propellers. However, it is a method first designed in the 1920s
that is proving one of the most promising. Flettner rotors use the Magnus effect to produce a means
of secondary power with very little required power input. This report will use Ansys-Fluent to
examine the propulsive qualities of multiple Flettner rotors, the aim of this report is to find what the
effects of interactions between Flettner rotors are and to hopefully find a value for the minimum
spacing ratio that can be used for future ship design. For the single cylinder case it was found that
for a flow with a Reynolds number of 200,000 the vortex shedding was suppressed at a spin ratio of
1.5. As the spin ratio of the cylinder was increased the coefficient of lift produced also increased, at a
spin ratio of 5 a maximum coefficient of lift of -8.60 was recorded. For a two cylinder case
preliminary simulations found that a spacing ratio of 2 gave a greater average coefficient of lift, with
a value of 22.59, than a spacing ratio of 4 which produced a coefficient of lift of 21.98. But the
average coefficient of drag recorded for a spacing ratio of 2 gave was 14.21 compared to 13.64 for a
spacing ratio of 4. As the difference between the coefficients of lift was larger than that for the
coefficients of drag, plus the weight saved by using a spacing ratio of 2 on a vessel, it was found that
using ta spacing ratio of 2 would be more efficient. However, when using a more refined mesh
running with spacing ratios of 3 and 4. It was found that for a spacing ratio of 4 coefficients of lift
and drag of 21.14 and 13.36 where recorded. These values make for a more efficient Flettner rotor
than the values of 21.09 and 13.73 for the coefficient of lift and drag for a spacing ratio of 3.
However, more work is yet required to simulate the spacing ratio of 2 at using this final mesh to
produce a definitive solution.
1|Page
James Grimshaw
3. Introduction
3.1.
At the moment the shipping industry produces 2.7% of the worlds total Co2 emissions and by 2020
is aiming to cut its emissions by 20%. This has now provided a market for ships using Flettner rotors.
The International Maritime Organisation (IMO) has recently made new amendments to the MARPOL
regulations in order to reduce the emissions from shipping. The allowable Sulphur Oxide (SOx)
emissions from the exhaust gases of ships have been reduced from 4.5% to 0.5% by 2012 and in
Sulphur Emission Control Areas (SECAs) it is to be limited to 0.1%. Nitrogen Oxide (NOx) emissions
are to be lowered to 3.4g/kw by 2016 in designated emission areas and 14.4g/kw outside these
areas. All new ships above 400 tonnes delivered after 2015 will have to have an improved efficiency
of 10%, for the same category of ships delivered after 2024 their efficiency will have to be improved
by 30%. Looking at these regulation changes it is no wonder Flettner rotors are being looked into
with a renewed interest, it is also evident why many shipping companies are trialling such propulsion
methods given the dramatic increase in oil prices over the past years and the further increase
forecast.
3.2.
The research presented in this report will be specifically looking into the effects of the interaction
between multiple Flettner rotors, simulations will look into the effects that the wake of one Flettner
rotors may have on a second rotors lift, as well as looking into the effects of interacting pressure
fields between two Flettner rotors. From finding these points the report hopes to try and find a
minimum spacing ratio between two Flettner rotors at which a high level of efficiency can be gained
from the rotors to make them a practical source of propulsion for ships.
As Flettner rotors can only be used with certain wind directions, at the point where the incoming
flow will be on the limit of a useable wind there will also be the maximum influence from the wake
of the first rotor onto the second rotor. Hence a range of simulations with flow at this inlet angle will
be produced. The flow will also be examined so the inlet angle is perpendicular to the rotors,
assuming the Flettners are to be aligned length ways down the vessel. This will give the inlet flow
condition at which the maximum interaction between pressure fields will be found. As the two
extremes of the flow are used as boundaries it can be assumed that any negative effects that occur
at these boundaries will be less at all other inlet angles. From the results produced from these wind
speeds average values for the coefficients of lift and drag at specific spacing ratios can be calculated.
Allowing for a minimum spacing ratio (T/D) to be produced, where T is the distance from the centre
of one rotor to another and D being the rotor diameter. This spacing ratio will assist in allowing
designers to maximise the lift produced. The simulation will be run at a Reynolds number of 200,000;
this is not quite at the same Reynolds number that would be expected from wind that would be
around the 500,000 mark. However, there is little difference in the effects of increasing the Reynolds
number when they are in this region. Therefore, in order to save on computational costs the
Reynolds number is kept at a lower value.
Three cases will be run in total to achieve these objectives; the first case is that of a single cylinder
which will give a value for at which spin ratios the vortex shedding is suppressed and what
coefficients of lift can be expected from a single cylinder. This may not seem like a necessary case,
2|Page
James Grimshaw
however, data for flow around a cylinder at such a high Reynolds number as used in this project is
limited, which is why it hoped to provide some from the work in this project. As simulating the flow
around a single cylinder will increase user experience on the Ansys-fluent software, as well as
demonstrate what modelling approach will be necessary for simulations of this nature it is seen as
beneficial to the final aims. The second case will be with two cylinders at two spacing ratios of 2 and
4; this case will use a slightly coarser mesh in order to gain an idea of what the trends can be
expected without changing the setup of the simulation from the first case and keeping the
computational costs down. The final case is again the simulation of two cylinders, however, a more
refined mesh will be used to try and gain more accurate results. There may be some changes in the
spacing ratios used in this final case depending upon the results of the previous case.
The relevance of this research can be seen in the design for the autonomous cloud seeding ships
which have been proposed as a method to combat the rising temperature of the earth. It has been
suggested that a remotely driven fleet of ships powered using Flettner rotors could be used in the
Pacific Ocean to spray salt water into the air in an attempt to seed clouds and make them more
reflective in order to reflect more of the sun rays. It has been suggested that multiple Flettner rotors
will produce enough power and allow the correct amount of control for these vessels. However,
these vessels need to use their power as efficiently as possible which requires a balance between
getting the most from the rotors but also keeping down the weight and size of the vessel. This
cannot be done if the Flettner rotors are separated at a large spacing ratio through fear of
interacting pressure fields. In a report on the proposed design of the cloud seeding ship S H Salter
states The case for two rotors is quite strong because of the extra agility and very tight turns.
However the high pressure field round one rotor may interfere with the low pressure field of the
other and so they must be as far apart as possible. (Salter, 2007). This is why this report hopes to
find a minimum distance at which the interaction between the pressures will no longer be an issue.
Further information on the cloud seeding ships can be found in (Latham, et al., 2012).
3|Page
James Grimshaw
4. Previous Work
4.1.
In 1925 Anton Flettner designed and built the first ship driven using Flettner rotors. The
breakthrough came from the work of Flettner and the German scientist Ludwig Prandtl, who found
that they could exploit the Magnus effect and use it to create enough lift to drive a ship. The first
ship ever fitted with these Flettner rotors was the Buckau. In 1926 Flettner changed the name of the
Buckau to the Baden Baden, as seen in Figure 1, it then successfully crossed the Atlantic in 1926
arriving before its sister ship which was powered by conventional sails, proving that this theory could
be used for shipping across the Atlantic. Flettner rotors proved to be very efficient compared to
traditionally powered diesel ships and could also be used in a wider range of wind speeds and
directions than conventional sail powered ships. In all Flettner ships the purpose of the rotors is to
provide extra propulsion as a second drive, the advantage of this was that only an extremely small
amount of fuel used, however, a disadvantage is that Flettners can only be used with certain wind
directions similar to that of sail ships. Compared to sails Flettner rotor ships were significantly lighter
and required significantly less work to drive the ship, giving the rotor ships increased performance.
However, due to low diesel prices and the economic downturn of the 1920s research into Flettner
rotors were scrapped. (Thiiink, 2012)
A second ship the Barbara, as seen in figure 2, was built in 1926 by the German navy and was used in
the Mediterranean as a cargo ship. The ship itself produced extremely promising results and proved
the benefits of Flettner rotors to the world. The ships captain reported that the Flettner rotors
generated an extra 600hp of power and could increase the speed of the ship by 2-3 knots whilst only
having a fuel consumption of 1kg/hour. The captain also reported that this extra power was
available from the rotors for around 30-40% of the time. Unfortunately even with all these bonuses
the Barbara was still reverted back to a normal cargo ship at the start of the Second World War.
4|Page
James Grimshaw
In 2010 the wind turbine manufacturer Enercon began using the E-ship 1 to transport its parts
around the world. Enercon found that their parts did not fit into the standardised cargo units,
causing the company increased costs. To cut out these costs it was deemed cheaper to build their
own specially designed ship, due to the companies green outlook it was decided to create a more
environmentally sound ship. The E-ship 1 uses a wide range of methods to increase its efficiency; a
new aerodynamic hull, new efficient propellers for use with the conventional diesel electric motor,
Flettner rotors and an oversized rudder to efficiently change the trajectory of the force from the
rotors. The power for the Flettner rotors comes from the exhaust gases, which are sent through a
Siemens gas turbine. The same exhaust gases are also used to heat the ships interior. All of these
changes have reduced the ships fuel consumption by 30-40% compared to that of a standard ship.
One of the environmental bonuses of the E-ship 1 is that it has no nitrogen oxide or sulphur oxide
emissions in ports. (Thomas, 2010)
4.2.
Previous Literature
As previously discussed in the introduction the objectives of the project are to produce data
regarding the spacing ratio (T/D) of Flettner rotors and the effect this has on the performance. This is
done with the autonomous cloud-seeding ship in mind, (Latham, et al., 2012) gives detailed
information regarding the application and uses of the cloud-seeding ships. It is suggested the ships
will be used to robotically patrol the Pacific Ocean and spray sea water up into the atmosphere, it is
hoped that this will increase the amount of heat reflected back into space by the clouds in a bid to
5|Page
James Grimshaw
decrease the earths temperature. The proposed plans include the propulsion for the catamaran
vessel to be provided by multiple Flettner rotors. The power for these Flettner rotors will come from
a turbine underneath the vessel which as dragged through the sea to produce enough power to turn
the Flettner rotors. Flettner rotors have been chosen for many reasons, it is easier to change the
rotational direction and speed of a cylinder electronically than to raise and lower a sail, the power
needed to rotate the cylinders is 5-10% of the power required to power the vessel using a
conventional rotor at the same speed. Another benefit is that Flettner rotors are self-limiting which
will be beneficial in high winds. All of this information is provided in the (Salter, 2007) report which
documents many of the possible arrangements of Flettner and vessel design.
In the (Mittal & Kumar, 2003) paper on flow past a single rotating cylinder the calculations where ran
at a Re = 200 for spin ratio 0 5, where is defined as
with
and
the circumferential velocity. They found that for < 1.91 vortex shedding occurred, but that
for values higher the vortex shedding disappeared up until spin ratios of 4.34 4.70. At this
range it was found that only one sided vortex shedding appeared. Mittal & Kumar suggest that the
vortex shedding is stopped as the large area of slow moving fluid that builds up at the front of the
cylinder when = 0 reduces as the spin ratio increases and then reappears at 4.34 4.70. In the
(Prandtl, 1926) paper it was suggested that the lift produced by a Flettner rotor could never exceed
4, however Mittal & Kumar found that at their maximum rotations their lift coefficient surpassed
Prandtls limit in a 2D simulation.
(Kumar, et al., 2011) experimented with the effects of two cylinders side by side using the Hydrogen
bubble technique (PIV). The experiment was ran with Re ranging from 100 up to 500, spin ratios
from 0 up to 5 and spacing ratios (T/D) 1.8 to 7.5. In these experiments it was again found that the
flow stabilizes as the spin ratio is increased. In this experiment the two cylinders are spun both
inward and then outward, which even though different to the rotation used in this project will still
provide some insight into the interaction of pressure fields. For a Re=100 it was found that as T/D
was increased from 1.8 up to 4 the spin ratio at which the vortices were suppressed increased from
1.2 up to 1.7 and that any change in T/D does not change the suppression spin ratio any further. For
higher Reynolds numbers it was found that the vortices were only suppressed for a T/D of 4 and 7.5.
Unfortunately the paper only looks into the effects of T/D on the spin ratio at which the vortices are
suppressed and have no mention of the effect on the coefficient of lift which would have given
useful reference values to check the validity of the simulation.
The (Akoury, et al., 2008) paper also found that for a spin ratio > 2 vortex shedding was
suppressed. They also found that at high spin rates the flow entered a second mode of instability
where a series of counterclockwise vortices are shed in the upper layer of the flow. The findings
from this report are very similar to the findings of Mittal & Kumar (2003) for the 2D case. Akoury
goes onto show how in the 3D case the Reynolds number at which this second mode appears (Re crit)
increases. Further work by (Kumar, et al., 2012) produced more results which strongly agree with
Mittal & Kumars original work. The experimental procedure used a water tunnel along with the
hydrogen bubble technique and PIV. Re = 200, 300 & 400 were used and the spin ratios varied from
0 to 5. It was recorded that vortex shedding was suppressed when > 1.95 and the flow remained
stable until the spin ratio was in the region 4.34 < < 4.70.
6|Page
James Grimshaw
Research presented in (Craft, et al., 2012) on Flettner rotors with the addition of Thom discs is
particularly focused on research into Flettner rotors with a look towards the application of rotors,
compared to other reports which are more focused on just the fluid mechanics of flow around a
cylinder. The work done by Craft is therefore very relevant to that being done in this project, a
URANS solver has been selected as the solver whilst the Reynolds number are 140,000, 800,000 and
1,000,000 which are much closer to those used in this project than any others papers reviewed. For
this reason it will be very important to compare the results gained for a single cylinder from my
project with that shown in this paper as a method to validate the accuracy of the simulation. Even
though there are many similarities Craft uses a 3D model rather than the a 2D model, there will be
some differences between the results as in the 3D model there will be oscillations in the flow in that
third dimension which will not be taken into account when only using a 2D model. However, Craft
fortunately compares the differences between the two model types and find there is actually very
little difference. For a 2D model with a =2 the CL=5.68 and the CD=0.150, for a 3D model with the
same flow conditions CL=4.80 CD=0.186. Craft also does some comparison between the URNAS
model and the LES model by Karabelas and finds that the difference between the two is in fact much
closer than it used to be. It should be noted that Craft uses more non-standard model for turbulence
and wall functions which will give greater accuracy.
(Seifert, 2012) paper looks into the possibilities of the use of the Magnus effect for aerospace
applications, however, there is a section looking into the previous uses of Flettner rotors on ships. It
details the Baden Baden and Barbara which were both designed in the 1920s and demonstrated the
possibilities of Flettner rotors. Interestingly it shows the weight saved due to the removal of sails and
addition of Flettner rotors which was considerable. There is also a look into the possible ways in
which Flettner rotors could be incorporated into an aerofoil to provide extra lift, giving more
impetus to Flettner rotor research. Although the only experimental work done in this paper is with
regards to aerofoils, a separate paper by (Thouault, et al., 2009) is used to provide some general
data about flow around a cylinder. In this data it is found that for a Reynolds number of 72,000 the
vortex shedding is suppressed at =2. It should also be noted that before the vortices were
suppressed the Strouhal number for the flow was approximately 0.2.
Recent work by (Karabelas, 2010) uses a Re = 140,000 which is closest to the Reynolds number used
during this report. Karabelas used a spin ratio from 0 up to 2 and aimed to again find the critical spin
ratio at which the flow becomes stable for this higher Reynolds number compared to that of Mittal
& Kumar (2003). The flow was found to stabilize at = 1.3, however, it was also found that the flow
is transitional for a few dimensionless time units, demonstrating a von Karman vortex street before
the flow fully stabilises. Unfortunately Karabelas did not go to any higher spin ratios so it is not
possible to see at what spin ratios the second phase of instability appears for this higher Reynolds
number. It is also worth noting which methods Karabelas used to gain his results given the
similarities to this report. The chosen program was fluent 6.3 which ran the Large Eddy Simulation
solution which is an extremely accurate method. The discretization selected was a low diffusive
central difference scheme of 2nd order accuracy. PISO was used as the pressure based solver. The
point implicit Guass Seidel method was chosen as the iterative method aided by an algebraic multigrid method to accelerate the method. Karabelas chose not to apply a wall function but instead
refine the mesh sufficiently at the wall and used a y+ < 5 for all spin ratios.
7|Page
James Grimshaw
Another valuable source concerning the computational side of flow around a cylinder is the work
done by (Catalano, et al., 2033). In this report Catalano compares the use of a large eddy simulation
(LES) solution to a ReynoldsAveraged Navier-Stokes (RANS) solution. When the flow becomes
unstable a URANS solver will be used in place of a RANS. Catalano compared the computational
results with experimental results such as Achenbach (1968) and those done by Falchsbart in
Zdravkovich (1997). Catalano ran the simulations at Reynolds numbers 0.5 x 106, 1 x 106 and 2 x 106.
It was found that the LES solution using a wall layer model gained the best results. A wall layer
model allows good approximation of the near wall eddies without the need for heavy mesh
refinement at the near wall saving on the computational costs. Catalano used a DES eddy simulator
in which the entire attached boundary layer is modelled. A very important feature of this report is
that at the Re = 0.5 x 106 it is found that the model predicts the flow reasonably well.
8|Page
James Grimshaw
5. Background Knowledge
Flettners rotor theory is based on the Magnus effect which was first described in 1852 by German
physicist Heinrich Magnus. The Magnus effect states that when a cylinder is rotating in the flow path
of a viscous fluid, as seen in figure 4 where the flow is moving from left to right and the cylinder in
rotating in a clockwise direction. The velocity of the fluid moving over the top of the cylinder
increases, whilst the velocity of the flow across the bottom of the cylinder is reduced. The change in
velocity creates a local change in pressure at the top and bottom of the cylinder. This creates a
region of high pressure at the bottom of the cylinder and a low pressure at the top resulting in a lift
force acting in the direction shown on figure 4. This change in velocity is caused by the interaction
between the fluid and the cylinder wall where there is no slip between the two, thus is can be
assumed the angular rotation of the cylinder directly effects the pressure difference and therefore
the amount of lift produced. This effect is often taken advantage of in sports such as tennis and
football to make the ball curve. It has been proved that the Magnus effect can stop vortex shedding
in the wake of the cylinder, this is believed to be caused by the large area of slow moving fluid at the
front of the cylinder when there is no angular rotation reducing as the angular rotation increases.
A comparison can be made between the lift created by a Flettner rotor and that of an aerofoil; the
lift in an aerofoil is created due to the displacement of the fluid downward due to the shape of the
aerofoil. This relies on Newtons second law that the net force on an object is equal to its rate of
momentum change and the third law every reaction has an equal and opposite reaction. Therefore
as the aerofoil deflects the momentum of the fluid downwards, a net force is produced equal to the
momentum of the flow displaced, this then creates an equal and opposite reaction giving the
aerofoil a lift force. The same occurs around a Flettner rotor, due to the rotation of the cylinder and
the pressure fields created the momentum of the flow is deflected and as a reaction to this a lift
force is produced which will provide the propulsion necessary to drive a ship. In a similar way to how
the lift force of a Flettner rotor can be compared to that of an aerofoil, so can the drag force. The
drag force in an aerofoil is due to the low pressure area to the rear of the aerofoil creating a suction
force which pulls the aerofoil back. The same occurs around a Flettner rotor with an area of low
pressure created behind the cylinder due to the separation point of the flow and the flows inability
to reform to the rear of the object.
9|Page
James Grimshaw
Choice of Simulator
Ansys-Fluent is an industry standard package and is commonly used for academic high-level research
giving it the correct credentials for use in this project. Ansys Fluent is an improved version of the
Fluent CFD package which was the original fluids package. The package itself comes with a wide
range of intuitive tools to aid with the modeling, meshing and post processing. Whilst the package
also boasts an impressive range of fluid dynamic options to enable a wide range of simulations to be
accurately run. At the moment there is a lot of research work going into the improvement of the
Ansys Fluent package to try and close the gap between the RANS and LES solvers making the use of
this package particularly favorable. The amount of support available from all Ansys packages makes
them an extremely useful tool. The addition of the Ansys workbench allows the management and
development of various simulations effortless.
6.2.
The purpose of spending time modeling a single cylinder case is twofold; firstly there is limited
available data for the spin ratio at which vortex shedding is suppressed for flows using URANS solver
at a Reynolds number as high as 200,000. Secondly there is no data available for a two cylinder case
allowing for some form of validity of the simulation. As the level of mesh refinement and simulation
setup can be carried across to the two cylinder case it also provides a method of validating the
results for the two cylinder case.
James Grimshaw
flow. It was then altered after the Von Karman Street was suppressed until the mesh became grid
independent. This was done in order to increase the accuracy of the calculation to get results that
could be validated by previous work. The mesh used for this part of the simulation can be seen
below in figure 6. From comparing the two it is possible to see that the final mesh is far more refined
than the original, this final mesh was found to be grid independent. Compared to the original mesh
the results produce from the final mesh will be far more accurate, however, the computational costs
required from this final mesh are greater.
6.2.1.2.
Checking the quality of the mesh:
In order to ensure that the mesh is of sufficient quality to produce accurate results that will
converge there are two important values to check. The aspect ratio is the measure of stretching in
the cell, the ideal value for this 1. This is only achievable when the cell is a perfect square; however,
it is highly unlikely to have a mesh near this value due to the impracticality of using square cells. It is
therefore considered a high quality mesh when the aspect ratio is around 4 or 5; the maximum
aspect ratio a mesh can have to be considered of a sufficient quality to successfully run is around 10,
although in some cases the aspect ratio can go as high as 35. The orthogonal quality of the mesh is a
measure of how deformed a cell is with regards to its adjacent cells. This quality is extremely
important when trying to enhance the stability of the simulation. The orthogonal quality of the mesh
is measured between 0 and 1 with 1 being a perfect mesh. A mesh with a minimum orthogonal
quality over 0.05 will be stable and produce accurate results. The values for the aspect ratio and
orthogonal quality for the final mesh seen in figure 6 can be seen in table 1 below. The mesh used in
this simulation is of extremely high quality as demonstrated below meaning the results will have a
high degree of accuracy.
Quality
Aspect Ratio
Orthogonal quality
Minimum
1.0057
0.9546
Maximum
2.6057
0.9999
Average
1.4682
0.9958
Table 1 - Table showing aspect ratio and orthogonal quality of the chosen mesh
11 | P a g e
James Grimshaw
6.2.1.3.
Choice of solver:
An Unsteady Reynolds Averaged Solver (URANS) was chosen due to the available computational
facilities. A Large Eddy Simulator (LES) would be the idea solver to use as this method directly
calculates the turbulence of the flow rather than a URANS solver which needs a turbulence model to
be included. This model does not directly calculate the turbulence instead it simply approximates the
turbulence, giving the simulation a further degree of error. Unfortunately the computational costs
needed for a LES solver are far too large.
6.2.1.4.
Choice of turbulence model:
As previously mentioned the turbulence model is used to obtain the Reynolds stresses of the flow.
There are various models available for use; the most common turbulence model type is the two
equation turbulence model. These models have become industry standard in recent years and are
the most commonly used for engineering problems. These two equation models include two
transport equations which represent the turbulent properties of the flow, it also allows for effects
like convection and diffusion to be accounted for. All two equation models use the Boussinesq eddy
viscosity assumption; this assumes that the Reynolds stress tensor, ij, is proportional to the mean
strain rate tensor, Sij. This allows the effect of turbulence on mean flow to be thought of similar to
that of molecular viscosity affects to laminar flow, in a turbulent case the Viscosity will be much
larger and will vary in space. One benefit of this approximation is that it allows for more intuitive
variables such as turbulence intensity and turbulence length scale to be used in a simulation.
However, the relationship is only true for simple flows such as boundary layers and wakes. But for
flows with strong curvature or strongly decelerated flows the relationship does not hold true,
incurring errors in the simulation (CFD, 2007). As the flow used in this simulation is fairly simply it is
acceptable to use a two equation turbulence model.
The K- turbulence model was chosen for this simulation, this is the most common two equation
turbulence model. In this model the K represents the turbulence kinetic energy which determines
the energy of the turbulence. The represents the turbulent dissipation rate which determines the
scale of the turbulence. One benefit of this particular model is the relative simplicity and low
computational costs that come with it, in this simulation the turbulent intensity at the inlet was set
to 2% and the turbulent viscosity ratio set to 5.
6.2.1.5.
Near-wall treatments:
There are two types of wall treatment; Low Reynolds Number Treatment (LRN) and High Reynolds
Number Treatment (HRN). A LRN integrates every equation up to the viscous sub layer of the near
wall flow. This means that the first computational cell must have its centroid in y+ 1. This results in
an extremely fine mesh; this will increase the computational costs of the simulation greatly. A HRN is
a wall function which relies on the log-law velocity profile of the flow. This assumes that the flow at
the near wall region takes the shape of a log-law. In order to do this the first computational cell must
have centroid with a y+ 20-80, this gives a coarser mesh than that of a LRN decreasing the
computational costs of the simulation. HRN can also enhance the convergence rate and numerical
stability of the simulation. For this reason it was chosen to use a standard HRN wall function to
approximate the near wall flow (CFD, 2011). In order to obtain the correct y+ value for the near wall
mesh it is necessary to run a short simulation and plot a graph of the y+ around the wall. From
looking at the plot it is possible to see whether the majority of the points plotted are in between the
20-80 range that is necessary to be able to use a wall function. A standard wall function was deemed
12 | P a g e
James Grimshaw
appropriate for this simulation. Figures 7 to 9 below show the y+ of the near wall mesh of the original
mesh and the altered mesh. Figure 5 shows the original mesh taken from tutorial which has a bias
factor of 460. For this reason it can be seen that the average wall y+ is relatively low as the mesh is
extremely refined at the near wall region. In order to try and improve the effects of the wall function
the bias type was changed to 160 to produce the plot seen in figure 8. However, it was found that
this mesh was too unrefined to give accurate results; the mesh was therefore refined using the
region adaption function. This allowed the mesh to be refined whilst leaving the near wall region
coarse enough to effectively use a wall function. Figure 9 represents the wall y+ for this mesh, it can
be seen from the plot that the majority of points lie well within the acceptable range allowing the
use of a wall function.
Figure 9 - Plot of wall Yplus across the cylinder wall in the final refined mesh
6.2.1.6.
Choice of Pressure Velocity Coupling:
For this simulation a Pressure Implicit with Splitting of Operators (PISO) coupling was selected. The
benefit of using PISO is the reduction in iterations per time step; this is done due to the addition of
two additional corrections; neighbor correction and skewness correction. Given the slight skewness
in the mesh used it is a further benefit to have the skewness selection to increase the accuracy of
the simulation (Ansys, 2009).
13 | P a g e
James Grimshaw
6.2.1.7.
Choices of discretization schemes:
In order to increase the accuracy of the simulation the discretization schemes for momentum,
turbulent kinetic energy and turbulent dissipation rate were all set to second order upwind. This
reduces the error of the simulation but at the same time increases the computational costs. Due to
the complexity of the flow it is seen as a necessary cost. The pressure gradient was set to the
PRESTO! scheme, PRESTO! was chosen as this scheme actually calculates the pressure on the cell
faces rather than simply interpolating on the faces as other schemes do. Whilst the gradient was set
to least squares cell based, these were chosen in order to give the simulation the necessary accuracy
whilst attempting to keep computational cost to a minimum.
The Strouhal number is a dimensionless number which describes oscillating flows where f is the
frequency of vortex shedding, D is the diameter of the cylinder and U is the inlet velocity of the flow.
In order to find an approximate value of the Strouhal number it is assumed that the frequency is 0.2,
this is similar to the frequency that would be expected with the given flow conditions as shown in
(Thouault, et al., 2009) paper. When an approximate Strouhal number is found by ensuring that one
vortex will appear for every 100 time steps gives a starting value for the time step.
The CourantFriedrichsLewy condition (CFL condition) is used to ensure convergence during the
solving of partial differential equations with the method of finite differences. The condition can be
seen above, where t is the time step, U is the inlet velocity and x is the cell length. It is important to
remember that the condition must be true for those cells at the near wall region which are smaller
than that at the start of the flow path. As stated above the CFL number must be approximately equal
to 1, however the lower the number the more accurate the simulation will be. In the fluent package
there is a means of plotting a contour of the cell courant number, this gives the various values of the
courant number. This is a useful tool when changing the mesh to ensure that the courant number is
always in its appropriate region. The plots of courant number for the various meshes used can be
seen in appendix B.
14 | P a g e
James Grimshaw
The above equation is a method to gain an idea of the dimensionless time, it is used in simulations to
find the time that the simulation must model the flow for to ensure all features of the flow have
been fully developed. In the equation as shown above ts is the time that the flow must be simulated
for, U is the velocity of the inlet flow and D is the diameter of the cylinder. When ts is such that the
answer is between 50 and 100 the flow will be appropriately developed so that all the major
features are present. This dimensionless number can then be rearranged to give a physical time that
the simulation must run for. Once the time step has been found using the two conditions above the
number of time steps necessary can also be found using ts from this equation.
The figures below demonstrate the flow around the cylinder before the changes to the time step
and after the changes. In figure 10 before the changes to the time step the flow was to diffusive for
turbulence to occur. As the time step was too large the smaller eddies in the flow were not picked
up leading to inaccuracies in the simulation. Once the time step was reduced so that the flow could
be calculated in more detail the Von Karman Street became present in the flow as demonstrated in
figure 11.
15 | P a g e
James Grimshaw
Figure 12 - Plot of coefficient of lift against the flow time for the with the revised time step
Figure 12 above shows the variation in the coefficient of lift against flow time, the graph shows how
the instability in the flow grows until the periodic oscillations caused by the vortex shedding reaches
it maximum. This complies with research gathered in other experimental papers and proves the
validity of the simulation set up. The figure above was taken from the final mesh chosen after it was
found to be grid independent.
Figure
Figure 5
Figure 6
Figure 13
No. Cells
18432
73524
293874
% Change
39%
0%
Table 2 - Table showing the effect of refining the mesh on the results
16 | P a g e
James Grimshaw
Figure 13 - Mesh after 2nd refinement for the single cylinder case
0.10
Coefficient of Lift
0.05
Coefficient of lift
original mesh
0.00
0
200
400
600
800
1000
1200
1400
Coefficient of lift
1st refinement
Coefficient of lift
2nd refinement
-0.05
-0.10
-0.15
Figure 14 - Comparisons of the coefficient of lift for various meshes for the single cylinder case
Figure 14 shows a plot comparing the coefficient of lift for the three meshes. The time across the
bottom has been altered to give a non-dimensional time that allows all three plots to be plotted
together and highlight the variation in results. Given the clear differences between the results for
the original mesh and after the first refinement it is clear that the mesh is not grid independent. It
should be noted that there has been a large increase in the refinement of the mesh, demonstrated
17 | P a g e
James Grimshaw
by the large increase in the number of cells. Even though this large increase in refinement means
that the computational costs of the simulation will increase, this is seen as an acceptable cost due to
the increase in the accuracy of the simulation. Figure 13 demonstrates the mesh used for the final
refinement of the mesh, as the figure shows the mesh has been heavily refined to check for the gird
independence of the mesh seen in figure 6. As demonstrated in table 2 and figure 14 there is no
difference between the two meshes up to three decimal places, therefore the mesh created from
the first refinement can be classed as grid independent. Given the scale of the refinement, the
computational costs required to run this final mesh are extremely large making the mesh seen in
figure 6 the mesh used in for the final simulations.
usual.
6.3.
18 | P a g e
James Grimshaw
6.3.1.1.
In order to achieve a high quality mesh it is vital to carefully plan the geometry before the meshing.
Figure 15 demonstrates the geometry produced by splitting an oval using various projection lines,
the two semi-circles at either end allowed for mapped meshing which has a high orthogonal quality
and low aspect ratio whilst also allowing the refinement to be specified to those areas where it is
required most, such as the near wall area. Ideally the two outer regions of the centre could also be
mapped to give a high quality mesh, however, doing this reduces the quality of the mesh in the
central section which is a more important area due to the increased activity. The central section is
separated as above in order to control the mesh, using the projections shown above the mesh can
be refined towards the near wall region by imposing a bias factor. Figure 16 shows the final mesh
used, the mesh is more refined at the near wall area and coarser closer to the inlet and outlet to
save on computational costs. The values for the orthogonal quality and aspect ratio can be seen
below in table 3; the mesh for the preliminary results for a spacing ratio of 2 and 4 can be seen. Due
to the increase in complexity in the geometry the orthogonal quality and aspect ratio values have
become somewhat worse. However, they are still within an acceptable range to produce accurate
results.
Spacing ratio
T/D=2
T/D=4
Quality
Aspect ratio
Orthogonal quality
Aspect ratio
Orthogonal quality
Minimum
1.0036
0.5192
1.0001
0.3719
Maximum
15.0681
0.9999
14.5756
0.9999
Average
1.8547
1.8457
1.9041
0.9743
Table 3 - Table demonstrating the quality of the meshes used for the various spacing ratio
James Grimshaw
coefficients of lift and drag produced by the two cylinder case will be as accurate as the results for
the single cylinder case. These simulations will again be ran at a spin ratio of 5 with the inlet flow
angle at 45 and 90 whilst the Reynolds number will be kept at 200,000. The spacing ratios used in
these final simulations will be 3 and 4 due to the findings of the preliminary case.
6.3.2.1.
In order to have a mesh which can be refined to the necessary level, whilst still having a high level of
quality it is important to start with a good quality geometry. Figure 17 demonstrates the geometry
used and highlights how the mesh was divided with the use of projections. As with the preliminary
simulations the two end sections have been left to be mapped meshed, whilst the central section
has been divided up into more sections to give increased control over the mesh. The two semi circles
around each cylinder allow for mapped meshing around all faces of the cylinder meaning that the
near wall region is sufficiently refined with a high quality of mesh. The central area has been
separated up into many more sections than done previously, this allows the regions closest to the
centre, where the majority of the activity will occur, to be highly refined but will also be of the
highest possible quality to ensure accurate results. Again various bias factors were used on the
projection lines to have a high level of control over the mesh, this allowed the mesh to be refined
where necessary and coarser in areas of low activity.
As grid independence has already been found once it would be an inefficient use of time to run
three simulations to find the gird independence of this final mesh. Instead the meshes can be
compared to the mesh used in the single cylinder case. Comparisons can be made between the
reduction factor and the smallest cell size of the two. The reduction factor is the rate at which the
cell size reduces and is found by rearranging the expansion of the cell size. The reduction factor is a
function of the bias factor and number of nodes. The smallest cell can be found by using the number
of nodes and reduction factor. By ensuring that the smallest cell values match and that the reduction
factor of each section is either equal to or greater than that of the single cylinder case ensures that
the mesh is as refined if not more so. The reduction factor is defined as;
James Grimshaw
Where L is the length of the section, r is the reduction factor and n is the number of nodes. The full
derivation of the reduction rate and smallest cell size can be seen in appendix D.
The orthogonal quality and aspect ratio of the new mesh can be seen below; as the mesh used for
this case is more refined than that used for the preliminary simulations the quality of the mesh has
been degraded. However, due to the time spent creating a quality geometry the values are still
within the acceptable region. The values of which can be seen in table 4.
Spacing ratio
T/D=3
T/D=4
Quality
Aspect ratio
Orthogonal quality
Aspect ratio
Orthogonal quality
Minimum
1.001
0.4862
1.001
0.6274
Maximum
7.3891
1
9.1856
1
Average
1.5039
0.9773
1.5148
0.9887
Table 4 - Table demonstrating the quality of meshes for the final simulations
Due to the increased complexity to the geometry, which leads to a poorer quality mesh, it is
necessary to change various parts of the setup in order to correct problems with the turbulent
viscosity. The under relaxation values control what percentage of the new answer and what
percentage of the old answer are used for each iteration, therefore a large under relaxation factor
gives a lower simulation time. The various under relaxation factors relevant to the turbulence were
reduced to try and stop the divergence in the turbulence values. Further changes were made after
plotting the cell courant number, using the function in the fluent software, which demonstrated that
for this simulation the number was far too large reaching a maximum of 14 in a number of places. To
ensure that the accuracy of these final simulations was as high as possible the time step was reduced
further to ensure that all the smallest eddies in the flow could be recorded. In order to encourage
convergence the pressure-velocity coupling was changed from PISO to SIMPLEC for the early stages
of the simulation and then changed back to reduce the calculation time when the simulation was
converging constantly. During the simulations further adaptions to the setup were required, at the
start of the 45 case it was found that the simulation was not completely converging, the residuals
were getting very close but not to the exact value. Instead of reducing the time step further or
increasing the number of iteration per time which could have enabled the residuals to converge the
time step was increased. This enabled the flow to move on from its original state which was causing
the issue. After a given time has elapsed the time step was then lowered and it was found that the
solution now converged.
21 | P a g e
James Grimshaw
Figure 19 - Contour of turbulent kinetic energy at =1.5 for single the cylinder case
Figure 21 - Contour of turbulent kinetic energy at =4.5 for the single cylinder case
Figure 20 - Contour of turbulent kinetic energy at =2 for the single cylinder case
Figure 22 -- Contour of turbulent kinetic energy at =5 for the single cylinder case
22 | P a g e
James Grimshaw
Coefficint of Lift
0.0
-2.0
-4.0
-6.0
Spin Ratio of 2
Spin Ratio of 4.5
-8.0
Spin Ratio of 5
-10.0
-12.0
0
10
15
20
25
Run time
Figure 23 - A graph showing the coefficient of lift for various spin ratios for the single cylinder case
James Grimshaw
for this spin ratio was -9.69 which occurred within the first 5 seconds of the flow. As with the
coefficient of lift for a spin ratio of 2 the lift decreases after its reached the maximum until it levels
out, the coefficient of lift at this point is -7.40. The coefficient of lift at this point is around 60%
bigger than for a spin ratio of 2, however, the spin ratio has increased by 125% meaning that in
order to gain a 60% in lift a much greater proportion of work must be done to the Flettner rotor to
provide the higher rotational speed.
James Grimshaw
7.2.
Figure 24 - Contour of turbulent kinetic energy at T/D=2 with flow at 90 for the two
cylinder case
Figure 26 - Contour of turbulent kinetic energy at T/D=2 with flow at 45 for the two
cylinder case
Figure 25 - Contour of turbulent kinetic energy at T/D=4 with flow at 90 for the two
cylinder case
Figure 27 - Contour of turbulent kinetic energy at T/D=4 with flow at 45 for the two
cylinder case
25 | P a g e
James Grimshaw
7.2.1.1.
Coefficient of lift
14.0
12.0
10.0
8.0
Cylinder 1
6.0
Cylinder 2
4.0
2.0
0.0
-2.0 0
10
15
20
25
Run time
Figure 28 - Plot showing the coefficient of lift around two cylinders at T/D=2 with flow at 90
Coefficient of drag
8.0
6.0
Cylinder 1
4.0
Cylinder 2
2.0
0.0
0
-2.0
10
15
20
25
Run time
Figure 29 - Plot showing the coefficient of drag around two cylinders at T/D=2 with flow at 90
The peak in coefficient of lift and drag seen in both plots above can be attributed to the production
of a single vortex. During the emission of this vortex the coefficient of lift reaches a maximum, the
lift then reduces and comes to a constant value at around 20 seconds. Comparing the coefficient of
lift seen in figure 28 above for the two cylinder case with that seen in figure 23 for the same spin
ratio but for a single cylinder the average lift value is far higher. The maximum value reached for
cylinder 1 was 6.13 and the final value was 5.72, the maximum value for cylinder 2 was 16.8 and the
final lift value was 14.5. Compare this with -10.80 and -8.60 for a single cylinder and the average
value is far greater. This would mean that a system of two cylinders with a spacing ratio of 2 would
26 | P a g e
James Grimshaw
produce a total output of 20.22 which is more than two times greater than that of a single cylinder.
This occurs as some flow which has already been turned by the first cylinder will then be turned
again by the second cylinder resulting in a greater total displacement and hence a greater total lift
force is produced, which will increase the coefficient of lift as seen above.
Figure 24 shows the contour of turbulent kinetic energy of the which corresponds to the lift and drag
plots seen in figure 28 and 29, the figure demonstrates how the rotation of the cylinder deflects the
flow to produce the resultant lift force. The equation to find the coefficient of lift can be seen below,
this demonstrates how a higher lift force will results in a greater coefficient of lift:
Where L is the lift force, is the fluid density, v is the flow velocity and S is the surface area, all
variables in the equations are constant for all simulations bar the lift force therefore it must be this
force which alters the coefficient of lift. Looking at the contour it is very easy to see where the high
pressure areas to the left hand side of the cylinder and the low pressure areas to the right hand side
of the cylinders form. Interestingly there seems to be no interaction between the low pressure field
of cylinder 1 and the high pressure field of cylinder 2 which was one of the predicted reasons for a
decrease in lift when using two cylinders. It can be seen from figure 24 that the wake of cylinder 2
actually moves around the wake of cylinder 2.
The variation in the coefficient of drag around the cylinder is just as important as the variation in the
coefficient of lift when investigating the variation in cylinder set up of Flettner rotors for the
propulsion of ships. As figure 29 demonstrates, for a flow inlet angle of 90 the coefficient of drag is
much lower than the lift produced. The final values of lift are 1.53 for cylinder 1 and 0.632 for
cylinder 2 giving a total coefficient of drag for the system of 2.162 which is close to ten times less
than that of the coefficient of the lift. The coefficient of drag is calculated using the following
equation:
Where Fd is the drag force, is the fluid density, v is the fluid velocity and A is the body area. As all
variables bar the drag force stay constant across the simulations the differences in the coefficient of
drag are due to the drag force. The drag force is created by areas of low pressure forming at the rear
of an object in a flow path; one possible reason for an increase in drag force could be down to the
separation points of the flow moving further upstream of the object. This decreases the pressure
behind the object and can occasionally be seen on turbulence contours as an increase in wake area.
27 | P a g e
James Grimshaw
7.2.1.2.
Coefficient of lift
12.0
10.0
8.0
6.0
Cylinder 1
4.0
Cylinder 2
2.0
0.0
-2.0
10
15
20
25
Run time
Figure 30 - Plot showing the coefficient of lift around two cylinders at T/D=4 with flow at 90
Coefficient of drag
4.0
3.0
2.0
Cylinder 1
1.0
Cylinder 2
0.0
-1.0
-2.0
10
15
20
25
Run time
Figure 31 - Plot showing the coefficient of drag around two cylinders at T/D=4 with flow at 90
Plots of the coefficient of lift and drag for the flow around a rotating cylinder with a spin ratio of 5 at
a spacing ratio of 4 can be seen in figures 30 and 31 above. These figures demonstrate how the flow
develops as time passes and also what lift and drag forces can be expected from the two Flettner
rotors in this arrangement. Looking at figure 30 the maximum coefficient of lift experienced by
cylinder 1 is 9.04 whilst the maximum coefficient of lift experienced by cylinder 2 is 13. This
maximum is created when a single vortex is produced; this vortex is a product of the starting setup
of the simulation as the vortex is produced before the rotation of the cylinder can have any effect.
28 | P a g e
James Grimshaw
The coefficient of lift for cylinder 1 settled at 8.28 whilst it was 11.4 for cylinder 2. This means the
total coefficient of lift across the two cylinders was 19.68, which is only slightly less than the total lift
produced when the cylinders are at a spacing ratio of 2 which is 20.22. This decrease could be
caused by the increase in the distance between the two cylinders, which causes less flow that has
already been turned to be deflected further by the second cylinder reducing the total deflection of
flow, hence the reduction in the overall lift force produced.
The contour of turbulent kinetic energy around the cylinders can be seen in figure 25; this figure
shows that at this spacing ratio there is no chance of a loss in lift due to pressure interaction
between the two cylinders. However, through comparisons with figure 24 it can be noted that the
flow visibly deflects more with a spacing ratio of 2 which confirms the above explanation as to why
the coefficient of lift is less.
Figure 31 shows the plot of the coefficient of drag against the run time, the final drag values that are
produced from this simulation are 0.834 for cylinder 1 and 0.812 for cylinder 2. This gives a total
coefficient of drag for the system of 1.646, this value is far less than that for a spacing ratio of 2.
Comparing the contours of turbulent kinetic energy in figures 24 and 25 it appears as though the
wakes behind the cylinder in figure 25 are much smaller than those in 24 indicating to differences in
the pressure areas to the rear of the cylinders. As the cylinders in figure 24 are so close the flow is
affected by both objects creating a larger wake.
7.2.1.3.
Coefficient of lift
0
-2 0
10
15
20
25
-4
-6
Cylinder 1
-8
Cylinder 2
-10
-12
-14
-16
Run time
Figure 32 - Plot showing the coefficient of lift around two cylinders at T/D=2 with flow at 45
29 | P a g e
James Grimshaw
30.0
25.0
20.0
15.0
Cylinder 1
10.0
Cylinder 2
5.0
0.0
0
10
15
20
25
Run time
Figure 33 - Plot showing the coefficient of drag around two cylinders at T/D=2 with flow at 45
Figures 32 and 33 show the coefficients of lift and drag for the flow around two cylinders. The angle
of the inlet flow for these cases is at 45 to the x axis which means the effect of the wake of one
cylinder on the lift of a second cylinder can be seen in the figures. The maximum lift generated for
cylinder 1 and 2 is -11.5 and -14.3, whilst the final lift values are at -10.06 and -12.59. This produces
a total lift for the system of -22.65, in a practical use not all of this lift will be used to propel a vessel
as some lift will be lost through the use of a rudder.
Figure 26 shows the contour of turbulent kinetic energy created by a flow with a 45 inlet angle and
a spacing ratio of 2. Comparing the wake in figure 26 with the wake seen in figure 24 which has the
same spacing ratio, the obvious difference is the direction of the wake which is why in figure 32 the
coefficient of lift is negative whilst in figure 28 the coefficient of lift is positive. A further difference is
that wakes appear to be closer to interacting; this is due to the flow inlet angle and could cause a
loss in lift. However, the lift produced by these cylinders is greater than that for an inlet flow of 90.
This increase could be caused by a greater deflection angle in the wake; however, the deflection of
the flow appears to be greatest in the 90 simulations. It is important to remember that even though
the deflection may appear to be greater, the momentum of the flow may be less which would lead
to a lower lift force being produced.
The plot of the coefficient of drag seen in figure 33 shows values far greater than those seen in any
of the previous cases, for cylinder 1 the drag coefficient is 7.55 whilst for cylinder 2 the coefficient is
18.70 giving the system a total coefficient of drag of 26.25. This is greater than the lift produced
from the rotor and would therefore produce a force in the opposite direction than that desired if
used on a ship. However, this is to be expected as this is the known limit of the use of a Flettner
rotor. Looking at figure 26 a combination of the flow inlet value and the low spacing ratio has
produced a very large wake with extremely low pressure to the rear of the cylinders resulting in the
very large coefficient of drag seen in figure 33.
30 | P a g e
James Grimshaw
7.2.1.4.
0.0
-2.0 0
10
15
20
25
-4.0
Cylinder 1
-6.0
Cylinder 2
-8.0
-10.0
-12.0
-14.0
Run time
Figure 34 - Plot showing the coefficient of lift around two cylinders at T/D=4 with flow at 45
25.0
20.0
15.0
Cylinder 1
10.0
Cylinder 2
5.0
0.0
0
10
15
20
25
Run time
Figure 35 - Plot showing the coefficient of drag around two cylinders at T/D=4 with flow at 45
The coefficient of lift and drag for the final simulation using the preliminary mesh can be seen in
figures 34 and 35 above. The cylinders in this simulation where at a spacing ratio of 4 and had a flow
inlet angle of 45 to the x axis. The maximum coefficient of lift recorded were -13.16 for cylinder 1
and -13.18 for cylinder 2. The final values at which the lift settled where -12.04 for cylinder 1 and 12.23 for cylinder 2, these values produces a total lift for the system of -24.27 which is greater than
any other system simulated up to this point. Due to the increase in space between the cylinders this
allows more flow to be turned producing a greater lift than that seen in figure 32 for the same inlet
31 | P a g e
James Grimshaw
flow angle but for a greater spacing ratio. As previously discussed the greater the mass flow rate of
fluid that can be turned the greater the lift output will be.
The coefficient of drag as plotted in figure 35 for cylinder 1 has a final value of 9.78 whilst cylinder 2
has a final value of 16.33, giving the total system a final coefficient of drag of 26.11. As already
discussed previously this is much greater than those seen for flow with an inlet angle of 90 for
various reasons. But more importantly the value is slightly less than that with the same inlet angle
but with a spacing ratio of 2. Given that the coefficient of lift is also greater with this increased
spacing ratio this means that the overall output of the system will be greater than that using a
smaller spacing ratio.
Figure 27 shows the contour of turbulent kinetic energy that forms due to the rotation of the
cylinder. Comparing the area of the wakes in this figure to those in figure 26 it is clear why the
coefficient of drag is less for a spacing ratio of 4. As previously discussed during the comparison of
the 90 simulations as the bodies are closer together this means the flow has a larger area to travel
around creating the large wake seen in figure 26 which in turn leads to the regions of lower
pressure.
Spacing
ratio
Flow
angle
T/D=2
45
90
Average
T/D=4
Average
45
90
Cylinder 1
Cylinder 2
Total
Total
Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient
of lift
of drag
of lift
of drag
of lift
of drag
-10.06
7.55
-12.59
18.70
-22.65
26.25
5.72
1.53
16.8
0.63
22.52
2.16
22.59
14.21
-12.04
9.28
-12.23
16.33
-24.27
25.61
8.28
0.83
11.4
0.81
19.68
1.64
21.98
13.64
Table 5 - Table showing the coefficient of lift and drag for the various spacing ratios and flow angles-preliminary mesh
Table 5 demonstrates that through use of a greater spacing ratio the efficiency of Flettner rotors is
not improved due to the reduction in the drag coefficient. However, this difference is so slight that
by using a spacing ratio of 2 the vessel size needed would be reduced which would increase the
efficiency of the rotors. This suggests that by making the autonomous cloud seeding ships smaller
the efficiency will be far greater. The table also indicates that at a wind speed of 45 Flettner rotors
become redundant which could be a potential problem for the autonomous cloud seeding ships as it
is proposed that Flettner rotors would be their only source of propulsion.
Validity of results
In order to make efficient use of the simulation time available it was decided to run preliminary
simulation, the results of which can be seen above. As the mesh used for these simulations was not
as refined as the mesh found to be grid independent in the single cylinder case the results produced
by these simulations will not be as accurate as could be achieved. This is due to the mesh being
slightly coarser and of slightly less quality leading to an increase in errors in the results. However, the
exact values of these simulations are not being compared to any other definite values, instead only
the trends and patterns that these simulations produce are used. In the single cylinder case both the
original and refined meshes produced results with the same oscillating trends; the only differences
were the definite values. The validity of the results will be able to be looked into more detail when
the results for the final simulations are analysed in the next section.
32 | P a g e
James Grimshaw
Figure 36 - Contour of turbulent kinetic energy for the final simulations at T/D=3 with
the inlet flow at 90
Figure 37 - Contour of turbulent kinetic energy for the final simulations at T/D=4 with
the inlet flow at 90
Figure 38 - Contour of turbulent kinetic energy for the final simulations at T/D=3 with
the inlet flow at 45
Figure 39 - Contour of turbulent kinetic energy for the final simulations at T/D=4 with
the inlet flow at 45
33 | P a g e
James Grimshaw
7.2.2.1.
Comparisons with preliminary results
Comparing the plots of lift and drag for with a spacing ratio of 4 for the preliminary and final
simulations the results are extremely similar. Even comparisons between the definite values show
very little difference between the preliminary and final simulations. This implies that the conclusion
arrived at in the previous section still holds true for these more accurate results. As running the
simulation at a spacing ratio of 2 in these simulations would simply recreate the same plots a spacing
ratio of 3 can now be used to find whether this provides a better rotor efficiency than that at 2.
7.2.2.2.
12.0
10.0
8.0
Cylinder 1
6.0
Cylidner 2
4.0
2.0
0.0
-2.0 0
10
15
Run time
20
25
Figure 40 - A plot showing the coefficient of lift around two cylinders for the final simulations at T/D=3 with flow at 90
Coefficient of drag
5.0
4.0
3.0
2.0
Cylinder 1
1.0
Cylinder 2
0.0
-1.0
-2.0
10
15
20
25
Run time
Figure 41 - A plot showing the coefficient of drag around two cylinders for the final simulations at T/D=3 with flow at 90
34 | P a g e
James Grimshaw
7.2.2.3.
Coefficient of lift
12.0
10.0
8.0
6.0
Cylinder 1
4.0
Cylinder 2
2.0
0.0
0
-2.0
10
15
Run time
20
25
Figure 42 - A plot showing the coefficient of lift around two cylinders for the final simulations at T/D=4 with flow at 90
Coefficient of drag
Cylinder 1
Cylinder 2
10
15
20
25
Run time
Figure 43 - A plot showing the coefficient of drag around two cylinders for the final simulations at T/D=4 with flow at 90
Through comparison of the plots of the coefficients of lift and drag for a spacing ratio of 3 in figures
40 and 41 with that for the same coefficients with a spacing ratio of 4 in figures 42 and 43 the ratio
which gives the greatest rotor efficiency at this flow angle can be found. The total coefficient of lift
for two cylinders with a spacing ratio of 3 is 19.33 and the coefficient of drag is 1.64. Whilst for a
spacing ratio of 4 the coefficient of lift is 18.97 and the coefficient of drag is 1.51. Therefore at a flow
inlet angle of 90 a greater coefficient of lift and drag is produced with a spacing ratio of 3. This
makes choosing between the two spacing ratios difficult as the difference between the efficiency of
the two is so small. The coefficient of lift is greater the smaller spacing ratio as more of the fluid is
deflected resulting is a greater change in momentum. Whilst this spacing ratio is beneficial for the
35 | P a g e
James Grimshaw
lift in this case it is detrimental for the drag. As the cylinders are relatively close the path the flow
takes around the outside makes the pressure difference across the cylinder greater resulting in an
increase in the coefficient of drag.
7.2.2.4.
Coefficient of lift
2.0
0.0
-2.0 0
10
15
20
25
-4.0
Cylinder 1
-6.0
Cylinder 2
-8.0
-10.0
-12.0
-14.0
Run time
Figure 44 - A plot showing the coefficient of lift around two cylinders for the final simulations at T/D=3 with flow at 45
25.0
20.0
15.0
Cylinder 1
10.0
Cylinder 2
5.0
0.0
0
10
15
20
25
Run time
Figure 45 - A plot showing the coefficient of drag around two cylinders for the final simulations at T/D=3 with flow at 45
36 | P a g e
James Grimshaw
7.2.2.5.
Coefficient of lift
0.0
-2.0 0
10
15
20
25
-4.0
-6.0
Cylinder 1
-8.0
Cylinder 2
-10.0
-12.0
-14.0
Run time
Figure 46 - A plot showing the coefficient of lift around two cylinder for the final simulations at T/D=4 with flow at 45
25.0
20.0
15.0
Cylinder 1
10.0
Cylinder 2
5.0
0.0
0
10
Run time
15
20
25
Figure 47 - A plot showing the coefficient of drag around two cylinder for the final simulations at T/D=4 with flow at 45
Comparing the coefficients of lift and drag for a spacing ratio of 3 in figures 44 and 45 with those for
a spacing ratio of 4 in figures 46 and 47 the total coefficient of lift produced by each system is -22.84
and -23.31. Therefore in order to create the maximum coefficient of lift from a system of two
Flettner rotors with an inlet flow angle of 45 it is best to have a spacing ratio of 4. It also appears
that in order to have the minimum coefficient of drag it is also more beneficial to have a spacing
ratio of 4 rather than 3. Therefore a spacing ratio of 4 gives a vessel a greater advantage, in terms of
rotor efficiency, than a spacing ratio of 3 at this flow inlet angle.
The reasons for this are much the same as for the preliminary case, for a spacing ratio of 4 more
fluid is deflected by the rotor increasing the momentum of the flow deflected hence increasing the
37 | P a g e
James Grimshaw
lift force produced by the rotor. For a spacing ratio of 3 the cylinders are close enough that the
separation points of the fluid from the body create a larger pressure difference across the cylinder
producing a larger drag force.
Spacing
ratio
T/D=3
Flow
inlet
angle
45
90
Cylinder 1
Coefficient Coefficient
of lift
of drag
-11.46
9.24
7.23
0.84
45
90
-11.9
7.77
Average
T/D=4
Average
9.92
0.74
Cylinder 2
Total
Total
Coefficient Coefficient coefficient coefficient
of lift
of drag
of lift
of drag
-11.38
16.57
-22.84
25.81
12.10
0.80
19.33
1.64
21.09
13.73
-11.41
15.3
-23.31
25.22
11.2
0.77
18.97
1.51
21.14
13.36
Table 6 - Table showing the coefficient of lift and drag for the various spacing ratios and flow angles final mesh
Table 6 demonstrates the final coefficient of lift and drag values produced for flow around two
cylinders for the final simulations. The flow angles of 45 and 90 are the boundaries of Flettner use
selected for these simulations, however, due to the effect of the flow angle has on the coefficient of
lift the values produced have been averaged to try and give a clearer image of which spacing ratio is
the preferred choice. Table 6 demonstrates that across a range of wind angles a spacing ratio of 4
will produce an increased coefficient of lift and a reduced coefficient of drag. This will give the
Flettner rotors an increased efficiency compared to rotors with a spacing ratio of 3. In order to
calculate the average values the modulus of the values were taken.
38 | P a g e
James Grimshaw
latter stages of the simulation there were still the occasional instances where the simulation did not
fully converge.
Due to the time constraints of this project it was decided to stop all simulations for the two cylinder
case after twenty five seconds; this was decided upon through the same non-dimensional process as
used in the single cylinder case, it was found that by this point all the flow characteristics would have
been present. Looking at the plots of lift and drag by this time it appears that the flow has plateaued,
however, the values were still changing. The values stay constant up to three decimal places but
after that the values are not constant giving the final results seen in table 6 further degree of
inaccuracy.
The average values found for the coefficient of lift and drag only use two flow angles, these values
are the outer bounds and were selected especially for that reason. However, to gain a more accurate
solution a number of flow angles between these bounds should be simulated. Unfortunately due to
the time constraints of this project this was not possible.
It must also always be remembered that all URANS simulations no matter how good the geometry,
mesh or setup always carry some measure of error down to the mathematical principles that they
function on.
39 | P a g e
James Grimshaw
8. Conclusion
8.1.
Conclusions of Results
8.2.
Review of project
The tutorials at the start of the project were extremely useful; they introduced many important
concepts with creating geometries one of the most vital lessons learnt. During the setup of the single
cylinder case the basics that were learnt in earlier university modules were forgotten and,
unfortunately, it was assumed that more complicated settings needed to be altered to try and create
an accurate simulation. However, shortly into the second semester there was a decision stop and
reevaluate the setup. After this the single cylinder was soon working, now that the first results were
produced it was then possible to begin to create the poster. At the poster presentation Dr Craft and
Dr Iacovides asked thought provoking questions concerning at what flow angles it would be
necessary to run the simulations at and how to improve the accuracy of the results by finding a grid
independent mesh. After these questions it was very simple to produce accurate simulations of the
flow for the single cylinder case. The next area of difficulty came when trying to create a mesh as
refined as the single cylinder case for the two cylinder case. After some initial difficulties it was
decided to run preliminary simulations to aid in the understanding of how to correctly mesh complex
geometry and of what could be expected from the flow. Whilst the preliminary simulations were
40 | P a g e
James Grimshaw
running it was then possible to start to create a suitable geometry and mesh for the final
simulations. After many days of trial and error style design work, a mesh was created that had
similar quality and refinement as the mesh in the single cylinder case. The final simulations could
then be run.
During this project a familiarity with the software was developed. Unfortunately a lot of time was
wasted in the first semester especially stuck on the single cylinder case, however, it was an
important learning curve which was beneficial in the latter stages of the project. If the time at the
beginning hadnt of been wasted whilst stuck on the single cylinder case there could potentially have
been enough time to simulate a spacing ratio of 2 with the final mesh which have given a more
conclusive answer.
41 | P a g e
James Grimshaw
9. Recommendations
9.1.
The results presented in this project are far from giving a complete picture of the flow around
multiple Flettner rotors. This means that more work is necessary to get an accurate solution for at
what spacing ratio Flettner rotors are at their most efficient. The first step would be to increase the
accuracy of the results. By increasing the quality of the mesh used for these simulations the accuracy
of the simulation could be increased whilst at the same time being able to use time saving measures.
This would give a clearer picture of what is going on in the fluid mechanics of the problem and a
better estimate of what sort of efficiencies can be achieved for a 2D case. One method of doing this
would be to invest more time in creating a well-structured geometry especially in the area of
interaction between the two cylinders. Due to the complex nature of the geometry at this point the
mesh was not of a quality as high as that seen in the single cylinder mesh. Given more time the areas
of low orthogonal quality and high aspect ratio could be smoothed out to increase the overall quality
of the mesh and increase the accuracy of the results.
As already previously seen in the comparisons between the coefficient of lift produced in (Craft, et
al., 2012)for a single cylinder and that produced for a single cylinder from the work in this project
the values are slightly different. This could be due to the Reynolds number of the flow or due to the
difference in the simulation setup. In Crafts paper a more specialized turbulence model and near
wall treatment were employed to increase the accuracy. In comparison to a LES simulation the
results produced were found to be very close to those produced from the LES simulation, meaning
the results have a high level of accuracy. The employment of these improvements would increase
the accuracy of the results and give further understanding of the fluid mechanics that are at work. As
the average values of coefficient of lift and drag upon which the various spacing ratios were
compared only took the outer to limits of the possible flow angles a more accurate average values
could be produced by simulating with more flow angles. From this point simulating various spacing
ratios can recommence, a good starting point would be to create geometry with a spacing ratio of 2
to check what happens to the coefficient of lift. The next step would be to try a spacing ratio of 3.5
to try and find the minimum spacing ratio for Flettner rotors.
9.2.
Modelling in 3D
After the work to increase the accuracy of the simulations in a 2D case a 3D case then needs to be
run, this would give a better idea of what would actually occur. A 2D case cannot take into account
the fluctuations in the flow that are known to occur in the third dimension and are known to affect
the coefficient of lift and drag of a Flettner rotor. Looking to crafts paper again the coefficient of lift
and drag for the 3D case were 4.80 and 0.188, whilst for the 2D vase the values were 5.68 and 0.150.
When it is understood what occurs during the interaction between the rotors changes and
improvements can be made in their design to increase the efficiency. One such design change which
should be tested is the addition of Thoms discs. The addition of these discs, spaced along the
cylinder has been found to stop the growth of fluctuations in the third dimension and increase the
coefficient of lift from the rotors. It has been proposed that the autonomous cloud seeding ships will
feature Thoms discs to increase the efficiency of the rotors. Therefore the whole system needs to
be simulated with the inclusion of Thoms discs to find the most efficient spacing ratio.
42 | P a g e
James Grimshaw
10. Bibliography
Akoury, R. E. et al., 2008. The Three-Dimensional Transition in the Flow Around a Rotating Cylinder.
Journal of Fluid Mechanics, 607(10.1017/S0022112008001390), pp. 1-11.
Ansys, 2009. Ansys Share net. [Online]
Available at: https://www.sharcnet.ca/Software/Fluent12/html/ug/node785.htm#sec-uns-solve-pvel
[Accessed 2013].
Catalano, P., Wang, M., Accarino, G. & Moin, P., 2033. Numerical Simulation of the Flow Around a
Circular Cylinder at High Reynolds Number. International Journal of Heat Fluid Flow, Volume 24, pp.
463-469.
Celli, V., 1997. Virginia University Physics Department. [Online]
Available at: http://galileo.phys.virginia.edu/classes/311/notes/aero/node2.html
[Accessed April 2013].
CFD, O., 2007. CFD online. [Online]
Available at: http://www.cfd-online.com/Wiki/Standard_k-epsilon_model
[Accessed 2013].
CFD, O., 2011. CFD Online. [Online]
Available at: http://www.cfd-online.com/Wiki/Two_equation_models
[Accessed 2013].
Craft, T. J., Iacovides, H., Johnson, N. & Launder, B. E., 2012. Back to the future: Flettner-Thom rotors
for maritime propulsion?. Turbulence, Heat and Mass Transfer , Volume 7.
Flettner, A., 1925. AnwendungderErkenntnissederAerodynamikzumWind-vortrieb vonSchiffen..
Zeitschriftf ur FlugtechnikundMotor-Luftschiffahrt, 16(3), pp. 52-66.
Floch, M., 2011. Ship Spotting. [Online]
Available at: http://www.shipspotting.com/gallery/photo.php?lid=1458180
[Accessed April 2013].
Karabelas, S. J., 2010. Large Eddy Simulation of High-Reynolds Number Flow Past a Rotating Cylinder.
International Journal of Heat and Fluid Flow, Volume 31, pp. 518-527.
Kumar, S., Cantu, C. & Gonzalez, B., 2012. Flow Past a Rotating Cylinder at Low and High Roatation
Rates. Journal of Fluids Engineering, 133(10.1115/1.4003984).
Kumar, S., Gonzalez, B. & Probst, O., 2011. Flow past Two Rotating Cylinders. Phys.Fluids, Volume 23.
Latham, J. et al., 2012. Marine Cloud Brightening. Phil. Trans. R. Soc. A, Volume 370, pp. 4217-4262.
Mittal, S. & Kumar, B., 2003. Flow Past a Rotating Cylinder. Journal of Fluid Mechanics,
476(10.1017/S0022112002002938), pp. 303-334.
43 | P a g e
James Grimshaw
44 | P a g e
James Grimshaw
11. Appendices
11.1.
Appendix A
Figure 51 - Plot of the wall Yplus for turbulent pipe flow tutorial
45 | P a g e
James Grimshaw
46 | P a g e
James Grimshaw
11.2.
Appendix B
Figure 56 - Contour of the cell courant number for the single cylinder case
Figure 57 - Contour of the cell courant number for the two cylinder case, preliminary mesh T/D=2
47 | P a g e
James Grimshaw
11.2.2.2.
Figure 58 - Contour of the cell courant number for the two cylinder case, preliminary mesh T/D=4
11.2.2.3.
Figure 59 - Contour of the cell courant number for the two cylinder case, final mesh T/D=3
48 | P a g e
James Grimshaw
11.2.2.4.
Figure 60 - Contour of the cell courant number for the two cylinder case, final mesh T/D=4
49 | P a g e
James Grimshaw
Week
12
Week
11
Week
10
Week
9
Week
8
Week
7
Week
6
Week
5
Week
4
Week
3
Task/Week
Week
2
Appendix C
Week
1
11.3.
Initial research
into project
Laminar pipe
flow tutorial
Turbulent pipe
flow tutorial
Laminar cylinder
flow tutorial
Turbulent
cylinder flow
tutorial
Literature
review
Simulation of a
single cylinder
Interim report
Interim report
deadline
Table 7 - Gantt chart demonstrating how time was planned and managed during the project in semester 1
Key
Completed on time
Not Completed on time/
actual time taken
50 | P a g e
Week
10
Easter Holidays
Week
9
Week
8
Week
7
Week
6
Week
5
Week
4
Week
3
Week
2
Task/Week
Week
1
James Grimshaw
Simulation of
single cylinder
Simulation of
rotating cylinder
Presentation
Poster
Poster deadline
Simulating various
spin ratios
Poster
presentation
Refined mesh to
increase accuracy
Creation of mesh
and geometry for
preliminary
simulations
Preliminary
simulation for two
cylinder case
Final report writing
Creation of final
geometry and
mesh
Final simulations
for two cylinder
case
Report submission
Table 8 - Gantt chart demonstrating how time was planned and managed during the project in semester 2
51 | P a g e
James Grimshaw
11.4.
Appendix D
)
)
52 | P a g e