Third International Symposium on Marine Propulsors

smp’13, Launceston, Australia, May 2013

A Holistic Design Approach for Propulsion Packages

Lars Greitsch1 , Markus Druckenbrod2 , Sven Bednarek2 , Hans-Jürgen Heinke3
1
Mecklenburger Metallguss GmbH (MMG), Waren (Müritz), Germany
2
Hamburg University of Technology, Hamburg, Germany 3 Potsdam Model Basin, Potsdam, Germany

ABSTRACT dages is highly dependent on the operational condition, like

This paper discusses an extended approach to ship design, for example thrust loading, draught or drift angles, the de-
which is focused on the propulsion package consisting of sign of these components has to be carried out more in-
propeller and rudder and special hub following geometries tegrated. Especially the interaction between propeller and
in order to ensure high overall efficiencies. Therefore the rudder has a significant influence on the overall performance
approach includes a newly developed optimisation setup of propulsion systems. A design of propeller and rudder as
for so called junction caps between propeller hub and pro- one unit is to be recommended. Therefore for example the
pul-sion bulb. Calculations and model tests give first im- distance between rudder and propeller and the twisted rud-
pression of the influence of the way of interaction between der design has to be based on simulations considering the
the cap shape and the costa bulb. operational profile or at least a ”multi-point”-specification.
More detailed propulsion improving devices like propul-
Keywords sion bulbs came more into the focus. In addition special
propeller caps help to reduce the losses due to the hub
Design Methods, RANSE, Propulsive Efficiency, Rudder, vortex. The research project ”BossCEff” concentrates on
Optimisation the possibilities to increase the propulsive efficiency by re-
1 INTRODUCTION
ducing the losses in the hub near flow. The project was
funded by the German Ministry of Economics. The devel-
New requirements regarding propulsive efficiency demand oped methods are part of the holistic approach presented in
better and more holistic approaches within the field of ship this paper.
design. The conventional design procedure of single and
separated design steps carried out by each supplier on its 2 THE DESIGN APPROACH
own is not capable enough to ensure the design quality we The new holistic approach includes the assessment of main
need to reduce the fuel oil consumption to the required de- dimensions based on the operational profiles and the nu-
gree. Especially if the operational bandwidth of the planned merical design of hydrodynamic details as for example the
vessel should be considered within the design phase. ”Multi- leading edge of the rudder as well as the hub near region
Point”-Design or design based on statistical profiles will in- including an optimized combination of junction cap and
crease the operational efficiency of future vessel compared propulsion bulb (see: figure 1).
to the conventional designs based on one single specifica-
tion point. The propulsion system consisting of propeller
and rudder and additional propulsion improving devices
has also to be designed considering such more realistic op-
erational requirements.
In addition propulsion efficiency itself became significantly
more important after the last economic crises and due to
the steady rise of fuel prices. But the design procedures
for merchant vessel still do not consider these new require-
ments. Often the main dimensions used to be fixed in a
project phase where neither the propeller nor the rudder
designer is yet involved. This may lead to suboptimal so-
lutions and a potential for improvement based on a holistic
approach. Furthermore the conventional design procedures
consider only one specification point. Depending on the
real operating profile the differences between this artificial
optimum and a more adapted design might be significant.
Especially in case of more detailed hydrodynamic designs
the use of operation adapted design procedures becomes
more important. Figure 1: Schematic design approach
As the interaction between propeller and hull with appen-

263
In order to cover all different level of information within butions a Monte-Carlo algorithm can be used in order to
the design phases the approach is consecutive. Thus the recalculate discrete operation situations with the same fre-
optimisation of the main dimensions can be carried out in quency behaviour as the input data. For each diced random
a very early phase whereas later on the detailed design can number between 0 and 1 the Monte-Carlo algorithm results
be based on a more final status. Starting based on a conven- in a specific value of the distribution of interest (see: figure
tional set of input data, given as a specification point, the 3).
design considerations can be refined by deciding some rep-
resentative operational key points within the assumed op-
erational profile (”multi-point”-design). The last and most
detailed way to introduce more realistic requirements to the
design is the fully operational based design. This requires
simulation methods covering a statistical description of the
transportation task as well as the operational behaviour of
the vessel. This can be done by use of force-based ma-
noeuvring algorithm considering the vessel operation and
Monte-Carlo algorithm covering the statistical input data
(Greitsch et. al. 2009).

3 SIMULATION OF THE PROPULSIVE PROFILE Figure 3: Recalculation of discrete operation conditions

The necessary input data has to be representative for the
assumed operational profile of the projected vessel. Thus These so determined operational situations can be accessed
it has to be a very clear description of the desired trans- by a force-based manouevring algorithm which is capa-
portation task. Based on later investigations focused on ble to calculate the equilibrium condition (see: figure 4).
rudder cavitation within the operational bandwidth a sta- Dices which lead to manouevring forces that can not be
tistical specification of the transportation task by frequency covered by the vessel design (e.g. maximum engine power
distributions of vessel speed and floating and weather con- exceeded) are rejected from the algorithm. Thus only diced
ditions arised as sufficient to cover these influences. Figure and simulated operating points which can be covered by the
2 shows typical frequency distributions of vessel speeds for ship design are saved within propulsive profile. From pre-
a container vessel (orange) and a RoRo vessel (blue). vious investigations can be derived that 2000 valid samples
are sufficient to reproduce the operational input data (El-
jardt et. al. 2009).

Figure 2: Typical vessel speed distributions

In order to ensure that the results of the simulation can pic-

ture the realistic vessel operation the data has to be checked
for possible statistical dependencies. Only independent in-
put data can be processed as a single condition. In case
of higher correlations between the data (for example wind
and sea state) the data has to be accessed by considering the
dependencies. A possible mathematical tool for these data
check is Kendall’s correlation factor, which indicates the
level of correlation in a range between -1 and 1, whereas
values near 0 stand for a weak correlation respectively less
dependencies in a statistical way.
These statistical distributions have to be preprocessed and Figure 4: Operation simulation algorithm
described as cumulative distributions to be accessible for
a numerical processing. Based on these cumulative distri- In the following the propulsive profile consisting of the

264
relevant system answers can be used as the data base for
subsequent calculations and evaluations. In case of rud-
der design aspects all parameters with influence on the rud-
der flow have to be considered (see table 1). But this data
has to be processed within the simulated data sets repre-
senting discrete operating points. According to the above
mentioned convergence criterion the propulsive profile also
consist of 2000 data sets in order to keep the relevance to
the desired operational profile.

Table 1: Propulsive profile

value unit
vessel speed vs,i [m/s]
wake fraction wef f,i [−] Figure 5: Propulsion package
propeller pitch P/Di [−]
propeller speed ni [1/min]
rudder angle δi [◦ ] far determined according to the rules and are used as a start
cross flow vc,i [m/s]
geometry, whereas the design of the propulsion bulb will
draught at a.p. dap,i [m]
be introduced within the procedure later on. Therefore the
focus is on the design of the twisted leading edge and the
An evaluation of the design variants based on input data so called junction cap for a reduced gap between propeller
with this fine resolution can only be carried out by use of hub and propulsion bulb. The design of the twisted leading
potential flow methods. In case of investigations with a edge stand as an example of a statistical design evaluation
higher numerical effort, either because of the necessity of combined with a direct optimisation based on the hydrody-
the consideration of viscous effects or because of too many namic average. Conversely the design of the junction cap is
design parameters, calculations based on an averaged op- based on a generic optimisation based on a simplified, but
erating condition have to be carried out. For each design viscous numerical model.
aspect has to be checked whether the numerical model can
be used directly on the statistical data set or the design step 4.1 The design of the twisted leading edge
has to be carried out based on this kind of representative of In analogy to the idea of cavitation reduction due to twisted
the statistical data. But in order to keep the influence of the leading edges of the rudder profiles the aim of coming closer
operational profile on the design step this averaging has to to the shock free entry of the flow is the basis of high ef-
be carried out for the technical consequence but not for the ficiency rudder blades. As mentioned above this design
input data. aspect is to be directly based on the operational profile.
4 DESIGN CATEGORIES Therefore a start geometry has to be calculated within the
whole propulsive profile. Here the local pressure on the
Considering the propulsive profile is the basis of a design rudder blade ploc,i is a direct function of the propulsive pro-
which is more adjusted to the needs. But in case of a file consisting of i discrete operating situations.
propulsion package consisting of a solid rudder and a fixed
pitch propeller (see figure 5) there is need for a transfer
from the discretisised datasets back to averaged but best
ploc,i = f (vs,i , wef f,i , P/Di , ni , δi , vq,i , dap,i ) (1)
representing ”specification” point.
Depending on the design demand each design aspect has to For a transfer from these i single pressure distributions on
be based on the most suitable design procedure. Accord- the rudder blade towards a representive flow situation an
ing to the possibilities of increasing the depth of detail can arithmetic average is be calculated for panel, repectively
adapted. As a first distinction table 2 shows the identified each local pressure.
design aspects and possible data bases.
i
Table 2: Design aspects X
ploc,k
k=1
design aspect design procedure pav,loc = (2)
n
rudder main dimensions conventional, acc. class
propulsion bulb acc. best pratice guideline Displaying all these averaged local pressure values leads
twisted leading edge averaged, equivalent flow to an averaged pressure distribution which is keeping the
junction cap generic optimisation characteristic of the flow but with considering all occuring
flow situations and the related frequencies. Figure 6 shows
the averaged pressure distribution (A) in comparison to the
The first two design aspects are so far left out in the oper-
pressure distribution for the original specification point (B).
ational based design. The rudder main dimensions are so

265
The example stands for the same Ro-Ro-ferry as the above ing in direction of the rudder chord because of the pro-
shown speed profile was derived from (Greitsch 2011). The peller’s slipstream.
direct comparison show the difference especially in the low Resulting from this 3D-2D-transfer a distribution of equiv-
pressure region at the upper leading edge due to the slip- alent 2D inflow conditions can be derived. These inflow
stream of the propeller. The pressure distribution based on conditions are suitable as input data for geometrical opti-
the original specification point did not cover the average misations of the rudder profiles.
extend of this pressure region.
Looking for a leading edge flow which comes closest to the
ideal shock free entry the shape of the profile is optimised
by use of the same tangent method and based on a simple
2D-calculation following the profile theory.

Figure 6: Averaged pressure distribution

Based on the averaged pressure distribution a determina-

tion of the inflow conditions for the rudder sections ac-
cording to the method of equivalent flow (Greitsch 2011)
is carried out. Here the inflow condition is equivalent if the
pressure distribution on the rudder surface shows the most Figure 8: Optimisation of rudder profiles
similar characteristic as a 2D-pressure distribution coming
from calculations considering the same geometry as the In comparison between the symmetric profile to the twisted
rudder section but different inflow conditions depending on profile it can be seen that the pressure distribution is re-
angle of attack and inflow speed. For this equivalence con- duced due to the more smooth inflow (see figure 8). There-
ditions height and location of the pressure minimum is to fore an adaption to the flow field given by the slipstream
be observed. The comparing process is carried out by a tan- is achieved. The rudder resistance within the slipstream is
gent method. The result is an angle of attack and an inflow reduced. These considerations lead to a distribution of dif-
speed where the 2D-pressure distribution is most similar to ferent rudder profiles alongside the rudder height. A fully
the founded pressure distribution on the rudder blade. S-shaped rudder blade follows the discrete rudder sections
(see figure 9)

Figure 9: S-shaped rudder blade

Figure 7: Equivalent pressure distribution 4.2 Cap shape optimisations

Propeller hub caps of course have been already investi-
Of course these pressure distribution cannot be fully con- gated. Either in order to evaluate the capabilities of the
gruent, because in case of the 2D-calculations a homoge- numerical methods (Junglewitz 1996) or in order to investi-
neous inflow is assumed, whereas in case of the pressure gate the influence of different cap shapes on the dead water
distributions the velocity of the surrounding flow is chang- area behind the cap (Maksoud 2004).

266
The aim within the project ”BossCEff” was to find an opti-
misation algorithm in order to find the most suitable junc-
tion cap for a given rudder and propulsion bulb geometry
and considering the propeller slipstream. Within the op-
timisation domain the propeller is represented by a body
force model in order to decrease the numerical effort (see
figure 10).

Figure 12: Geometry parameterisation

Figure 10: RANS domain

The optimisation has to include the parametric definition

and generation of the geometry, the mesh generation and Figure 13: Propeller influence
the numerical calculation of the propulsor performance. In
order to achieve an automatic processing the algorithm has
to be able to prepare all numerical setups for each variant in the volume mesh for the calculation method for viscous
by itself (see figure 11). flow, resulting in a remarkable reduction in grid size and in
calculation time.

Figure 11: Optimisation process

This requires a fully parametric description of the cap shapes

to be optimised. In order to cover even more complicated
cap shapes (see figure 12) the description of the geome-
try was based on two independent functions f (x) and g(x)
(Druckenbrod et. al. 2012). Figure 14: Hub near flow

For the propeller representation via the body force model

the TUHH inhouse code panM ARE was used. For the 5 MODEL TEST RESULTS
hub near flow the influence of the propeller blades has to
Within the research project ”BossCEff” first model tests
be introduced into the viscous calculation within the opti-
were carried out based on the free research geometry KCS
misation domain. An impression of these propeller influ-
3600TEU from KRISO. The model tests were targeting the
ence shows figure 13.
single influences of relevant design details like for example
For the coupling purpose the pressure values on each panel the distance between rudder shaft and the propeller plane
of the propeller calculation are transformed into body forces or presence of a propulsion bulb.
and imported into ANSYS CFX by adding them to the
For identifying the potential of the design of the twisted
source term of the time-averaged Navier-Stokes equations.
leading egde to rudders have been built. The first rudder
Hence, the propeller geometry does not need to be reflected

267
was based on symmetric rudder profiles, whereas the sec- design aspects can be captured. Depending on the numeri-
ond differs only in the twist of the leading edge. Both rud- cal modeling which has to be chosen for a proper capturing
der were tested with and without propulsion bulb. In ad- of the influences this statistical description has to be aver-
diton various different cap shapes were tested. Figure 15 aged. This averaging is carried out for the effect and not
gives an overview about the model test matrix which has for the input data. This ensures, that the flow characteristic
been investigated. will not be lost during averaging.
This statistical design can be carried out for the leading
edge design of the rudder without lack of operational in-
fluence. The evidence can be provided by realculation of
the modified geometry again within the whole operational
profile.
The design of the junction cap is carried out within a sim-
plified domain, by replacing the propeller by its forces.
This saves numerical effort, which can be put on an ex-
tension towards a ”multi-point”-optimisation in order to
achieve the link between cap optimisation and operational
profile. The domain already is capable for combined opti-
misation of junction cap and propulsion bulb, whereas the
existing calculations only consider a cap optimisation for a
fixed bulb geometry. This has to be introduced as the next
step. A design including all mentioned improvements will
lead to an increase of propulsive efficiency of 2,5% up to
3% compared to a state of the art rudder-propeller configu-
ration. Scheduled model test will bring verificaton of these
values.

REFERENCES
Figure 15: Model test matrix
Greitsch, L., Eljardt, G. & Krüger, S. (2009)
First International Symposium on Marine Propulsors
The model test results show a noticeable influence of the Trondheim, Norway.
distance between rudder and propeller. Here the higher ef-
Eljardt, G., Greitsch, L. & Mazza, G. (2009) ‘Operation-
ficiency (abt. 1%) was achieved with a smaller distance
Based Ship Design and Evaluation’
between propeller and rudder. In addition the rudder vari-
International Marine Design Conference
ant with twisted profiles shows a higher efficiency than the
Trondheim, Norway.
symmetric geometry. In this case the difference was about
1.2% on the propulsive efficiency. The propulsion bulb in Greitsch, L. (2011) ‘Entwicklung eines Entwurfsverfahrens
any case has shown positive effect on the efficiency. The für Schiffsruder auf der Basis statistischer Betriebspro-
single effect was measured as a decrease of 1.3% of propul- file’.
sive power. Ph.D. thesis, Hamburg University of Technology, Ham-
burg, Germany
6 CONCLUSIONS
Junglewitz, A. (1996) ‘Der Nabeneinfluss beim Schrauben-
As the design of propulsion devices is more and focused propeller’.
on even small design details which will bring small but fur- Ph.D. thesis, The University of Rostock Rostock, Ger-
ther percentage of efficiency the effort to put on a design many
procedure is increasing. In order to scope with all main in-
Abdel-Maksoud, M., et al. (2004) ‘Numerical and Experi-
fluences this paper gives the first impression of an advanced
mental Investigation of the Hub Vortex Flow of a Ma-
design approach, which will be capable to cover the realis-
rine Propeller’,
tic influences on the propulsor design. This is considered
25th Symposium on Naval Hydrodynamics St. John’s,
by a statistical description of the operational profile which
Newfoundland and Labrador, Canada
leads into a propulsive profile after simulations based on
the numerical model of the vessel. Druckenbrod, M., et al. (2012) ‘Geometric Modelling for
Optimisation of Propeller Hub Caps’,
The so derived propulsive profile is able to be processed in
15th Numerical Towing Tank Symposium Cortona, Italy
further partial design procedures. Therewith all necessary

268