Modern Windships; Phase 2

9. Simulations

Introduction
Assumptions and Restrictions
Results from the VPP
Simulations of Combined Propeller and Wind Forces
Polar Diagrams and Input to Weather Route Optimisation
Summary and Conclusion, using the VPP.
Notes on the Propulsion System, VPP and Weather Routing
Weather Routing
Assumptions and Restrictions
Trade Patterns
Calculations
Results
Weather Routing, Conclusion

It was determined at an early stage in the project to use computer simulation to determine the ship’s speed depending of wind direction, wind speed and the propeller and rudder settings, due to the complex dependencies of the aero- and hydrodynamics forces. This kind of simulation is common in professional yacht racing and is known as a VPP (Velocity Prediction Program).

The large benefit using VPP-programs is the possibility to quickly evaluate a large number of different scenarios and routes. For reliable weather routing and fuel consumption prediction a VPP simulation is essential. The VPP was developed by DMI/SL, see Ref. 7.

DMI/meteo has subsequently used the resulting VPP-databases for weather routing, optimisation and fuel consumption calculations, see Ref. 8.

Introduction

The Wind Ship Simulator, the actual VPP code, was developed by DMI/SL. It utilised a commercial program, MatLab, see Ref. 49, for quick development. An optimisation module available separately for MatLab was also used. This however meant that a MatLab-license is a prerequisite to run the actual VPP.

A "common" VPP in Professional Yacht Racing performs the task of predicting the max. sailing speed given the wind angle, wind strength, wave height and direction, and sometimes other related data for sailing. This way a sailing crew can be given a theoretical "target speed" when competing. Data is normally presented in what is called a "polar diagram", that is the max. attainable sailing speed is presented as a function of wind angle and speed. This way it is easy for a helmsman to determine whether the crew is sailing the yacht optimally, or if further trimming of the sails is necessary.

The Wind Ship Simulator was more complicated as the installed engine power was adding another independent parameter. The ship speed was now not only dependent on winds and waves, but also on thruster settings (thruster load and direction).

The resulting polar diagrams could be presented in two ways, either as attainable ship speed given a specific wind and thruster setting, or thruster load in order to attain a target ship speed in given wind conditions.

In the modern WindShip project the VPP was subdivided into two separate modules:

1. Generation of the result databases.

2. Interpolating the results in the database based on the users input.

In the first module the input to the VPP were the aerodynamic and hydrodynamic coefficients measured in the wind tunnel and towing tank respectively (described in chapter "8. Hull Design" above). From these, and a given set of wind speeds, wind directions and thruster settings the ship’s speed, thruster load, drift angle etc. were calculated and stored in large matrices. This was typically a process that required several days of processor time on an average PC.

In the second module a simple user interface to interpolate within the result matrices was displayed. By asking the user to provide wind data and target speed or thruster loading, the VPP interpolated the result database and presented the results of the calculation. The interpolation in the second module was a very fast process compared to recalculation of the database.

By dividing the VPP in two modules a user does not need to wait for a complete calculation of the VPP every time a specific result is asked for. A screenshot of the VPP user interface is shown below, see Figure 64.

Figure 64. Look here please

Figure 64. Screenshot of the VPP user interface.

Assumptions and Restrictions

In order to make a manageable code, which could produce result databases in a realistic time-scale, the following assumptions were made:

	Added resistance due to waves. The added resistance from waves was calculated according to strip theory. The added resistance consists of a term Raw(0) which is the added resistance in head seas. Raw(0) is then multiplied with a function f(a ) depending on the relative wave direction. For further information, please refer to Ref. 10 and Appendix 2.
	Load and ballast. All measurements were performed using fully loaded condition. The VPP therefore predicts the speed for a fully loaded WindShip. In the subsequent economical calculations performed by Mærsk Broker a reduction of the fuel consumption due to ballast sailing was incorporated.
	Heeling. A balancing of the heeling angle was not performed in the VPP. Measurements showed that the influence of heeling was small at small heeling angles. Calculations further showed that the heeling angle would be less than 8° during all normal sailing conditions. All calculations were therefore performed under the assumption of an upright ship.

The size of the result databases were kept limited by choosing the following calculation domains:

	Thruster load: 0-100%
	Wind Speed: 0-20m/s
	Wind direction: 0-180° (symmetrical)
	Ship speed: 0-15 knots

Some limited extrapolation could also be performed, extending the calculation domain further.

Results from the VPP

The most important use of the VPP output was generating input to the weather routing. It is also interesting to see how the WindShip performs in different conditions, some polar diagrams are therefore presented below, see Figure 65 and Figure 66 below. All possible combinations could obviously not be shown. Further results are available in Appendix 11, which in turns are reprints from Ref. 7.

Figure 65. Look here please

Figure 65. WindShip speed polar in 3 m/s true wind, varying the engine power between 0-100%.

Figure 66. Look here please

Figure 66. WindShip speed polar in 16 m/s true wind, varying the engine power between 0-100%.

From the polar diagrams we concluded that:

The modern WindShip could sail upwind up to true wind angles of about 40° . It should be noted that the sails were not "flattened" during the measurements, since the angle between the flap and mast was fixed in the measurements. With flatter sails the upwind capabilities might have been even better.

There were regions in the polar diagrams in which WindShip could not sail. These "no-go zones" were typically in higher winds combined with low diesel engine thrust and sailing upwind. The difference between the aerodynamic and hydrodynamic pressure centres was too large to be balanced other than by high engine thrust, see section "Balancing the WindShip" in chapter 8. In a "real world" situation the captain would ease on the aft-most sails, in order to balance the ship better. This is common practise, and not unique to the WindShip. The easing of the sails would result in other aerodynamic coefficients if measured, and the VPP would give a calculated speed. Due to lack of resources these variations of sail trim were not measured, thus the VPP used coefficients where all sails were giving max CL. This resulted in larger "no-go zones" than necessary.

"Tacking downwind" will be necessary in order to achieve max. speed when sailing in the wind direction. The sails were set so that at true wind angles of 155° -160° the WindShip stopped using the high lift profiles for generating lift, and was merely "pushed" down the wind. Since the profiles are of stiff shape it might be beneficial to allow for lift generation "deeper down the wind" (down to maybe 170° ). This should be investigated by use of wind tunnel measurements and the VPP in a future phase of the modern WindShip.

Max. ship speed in no wind condition, using only the diesel engine was 13.5 knots. (Due to the chosen engine arrangement, with primary focus on manoeuvrability of the ship, the fuel consumption at this speed was higher than for a conventional product carrier at the same speed. See "Fuel consumption of the Modern WindShip.", chapter 10)

The best sailing performance was achieved at a true wind angle of approximately 100° . Without using the engine the following performance could be anticipated, see Table 19 below:

Table 19. The modern WindShip speed at 100° true wind, 0% engine power.

It should be noted that the ship’s speed exceeds 15 knots already at 12 m/s 100º true wind. As stated in section "Static Force Measurements of the Hull in Towing Tank" in chapter 8 above, the hydrodynamic coefficients were only measured up to 15 knots of ship speed. Some limited extrapolation can be performed, but from 15 m/s of wind and upwards, where the ship’s speed exceeds 20 knots, the results should be evaluated with great care.

Simulations of Combined Propeller and Wind Forces

As indicated above the modern WindShip Simulator differs from other common VPP’s as the thrust level and thruster direction was also included in the equations. This is necessary, since the Wind Ship will often sail using a combination of sail and engine power.

All velocity prediction programs perform iterative loops in order to balance the aerodynamic forces with the hydrodynamic in relation to the input data given by the user. For the modern WindShip simulator an extra iterative loop was thus necessary, to account for the thruster settings. This of course added to the computer time used.

Polar Diagrams and Input to Weather Route Optimisation

The Polar Diagrams shown Appendix 11 were calculated by DMI/SL. A general problem when presenting Polar diagrams for the modern WindShip is the necessity to plot 4 quantities, namely:

	The ships speed v
	The wind speed w
	The true wind direction d
	The engine power p

In a "normal" polar diagram for sailing ships the last item, engine power p, is not relevant. For a WindShip it becomes of vital interest, as the ship will need to run the engines in order to reach its destination port in time.

The polar diagrams were presented in Ref. 8 as matrices instead of the conventional polar plots. See Figure 67 below.

Figure 67. Look here please

Figure 67. WindShip performance as 3D matrix plot instead of polar plot.

For the calculations by DMI/meteo, databases of the modern WindShip performance were necessary. These calculations were performed at Pelmatic Knud E Hansen. The aerodynamic coefficients were slightly modified, to account for the effect of using five and four masts, as well as the original six masts. The modification simulated the folding together of one or two of the aft-most masts. The results showed that the ship’s balance was better, the "no-go zones" were smaller, although the ship’s speed was lower due to the reduced lift. The data bases calculated by the VPP were transferred to DMI/meteo by email for subsequent calculations. Each database took approximately two days on a normal Intel PC to calculate.

Summary and Conclusion, using the VPP.

The WindShip VPP proved an indispensable tool for further evaluating the economic feasibility of the Wind Ship. It was also very useful in understanding the behaviour of the WindShip in different conditions. It is highly recommended to continue developing and using the VPP in future modern WindShip projects. Before one actually builds a WindShip the VPP, weather routing tools and economical models developed within phase 2 of the project should be brought to extensive use. Combined with measurements of different propulsion alternatives there is a real chance of choosing an optimal engine and propeller.

The sailing performance of the modern WindShip was judged adequate. Improvements in sailing characteristics would have an immediate effect on the total economy of a modern WindShip, and are therefore a high priority. Several parameters mentioned above remain to be investigated in depth, including sail trim and balance optimisation.

Notes on the Propulsion System, VPP and Weather Routing

The WindShip machinery is not only used to propel the ship forward as on conventional ships, but also to provide necessary turning moments to balance sail forces. Moreover, even if the service speed is a constant, the ships actual speed may vary in an interval of 5 to 20 knots depending on the weather. The original engine package on the WindShip was chosen mainly to act as steering and emergency machinery, with the optional capacity to drive the ship in absence of wind.

During the tests at DMI/SL a fixed pitch propeller was used. There are some torque and rpm limitations on a fixed pitch propeller. These can be difficult to adhere to under the varying conditions facing a WindShip. To use a variable pitch propeller would partially solve these problems.

Using a variable pitch propeller fairly well represents the output from the VPP, where only power output and steering angle are presented, regardless of ships speed. It would be relatively simple also to list the rpm and torque necessary for the needed propulsion output. One would then find quite a number of "unfeasible" solutions using a given fixed pitch propeller, where either the torque or rpm are out of bounds.

Thus one would need to test a large number of propeller, rudder and diesel engine combinations in order to find the most economical combination. It should be noted that these combinations should be run through the weather routing as well, not only the VPP, to find the most economic alternative. Due to budget limitations it has not, in the current project, been possible both to develop these tools, and completely run a specific ship with many different propulsion options, in order to determine the optimum propulsion package.

Weather Routing

In collaboration with DMI/meteo an exhaustive weather routing was performed, see Ref. 8. Weather routing is a tool used to predict and make use of weather prognosis. By using the polar diagrams calculated above using the VPP and meteorological data, one can find the optimal route between two harbours. The optimisation is normally performed in order to minimise the sailing time. In the case of the modern WindShip the code was modified in order to minimise the fuel consumption, under the restriction to reach each destination in a given time.

Assumptions and Restrictions

A number of assumptions were made in order to increase the "realism" of the weather routing. These assumptions were mainly concerned with hard weather behaviour. At high wind speeds the VPP can often predict high sailing speeds downwind, in excess of 20 knots, see Table 19 above. This was not considered realistic, in a "real" situation the captain would reduce speed due to structural considerations, slamming etc.

To describe this behaviour in a simple manner the following conditions were imposed on the weather routing:

	The ship speed was limited to 20 knots. A max. extra-polation of 5 knots from the data measured by DMI/SL thus resulted.
	In winds above 30 m/s the speed was set to 5 knots, and 100% engine power.

Especially the last point is punishing. The weather routing optimisation will thereby automatically avoid regions with winds above 30 m/s, since it is very costly in fuel consumption.

A fairing of the polar data supplied by the VPP program was performed prior to incorporating the data in the weather routing algorithm. The fairing was performed to fill "gaps" in the polar matrixes, where the VPP clearly predicted unphysical solutions. The effect of balancing by sail trimming was thereby also somewhat accounted for.

Trade Patterns

The collection of routes in which a ship sails is called a trade pattern. The used trade patterns were supplied by Mærsk Broker as being typical for product carrier of the WindShip size. One trade pattern was located in the Atlantic/Mediterranean region, and the other in the Indian-Pacific region. The routes and the sailing distances are shown in Table 20 and Table 21 below:

Table 20. Trade routes in the Atlantic trade pattern

Table 21. Trade routes in the Indian-Pacific trade pattern

A map of the individual trade routes can be seen below in Figure 68:

Figure 68. Look here please

Calculations

The calculations were performed using four optimisation methods, allowing for maximum comparison and enabling a detailed post-processing.

A general approach was used, where three "representative" years (1993 -94 -95) were chosen by DMI/meteo, and the trading routes were sailed using one ship every day in both directions. This approach enables following of an individual ship from port to port during the –93 to -95 period, taking any number of port days into account.

Speed

A very important factor is the ship’s speed. Increasing the ship’s speed will have negative effects on the fuel consumption. A higher speed will on the other hand result in higher productivity and better earning capacity. The weather routing was therefore performed using four individual ship speeds: 10 – 11 – 12 – 13 knots. All weather routing calculations were thus repeated four times for the different speeds.

Since the WindShip is a sailing ship, speed will vary underway depending on the weather. This means that the speed requirement in reality is a time requirement, or an average speed requirement. For a given trade route there is a given time in which the WindShip should reach it’s destination. This was also how the optimisations were performed. The task became to minimise the energy consumption, under the condition to reach destination within a fixed time, depending on the target speed in use.

Optimisation methods

The four optimisation methods used were:

1. Sailing using great circle navigation. No weather prediction was involved, the ship simply sailed along the great circle course, using whatever wind was available. If the WindShip due to weather was lagging behind or ahead of schedule, speed was adjusted so that the next waypoint was reached within the original time schedule.
   This method of calculation was meant to simulate a captain sailing without taking any weather information into account, and should result in the highest fuel consumption.
2. Sailing an optimal route based on historically known weather. Since we were using years –93 – 94 – 95, we knew the weather from historical data. Therefore we could calculate the best possible route, something we can not do in practice, since we are always depending on forecasts in "real life".
   This should yield the lowest possible fuel consumption attainable by the modern WindShip.
3. Sailing along a route only taking into account weather prognosis available at the start of the trip. This corresponded to the captain getting the latest weather report when the ship left port, but not receiving any further weather prognosis underway. The captain sails according to the prognosis, even if the actual weather differs from what was forecasted.
   This rather "stubborn" approach proved to be one of the costliest in terms of fuel consumption.
4. Sailing along an optimised route using weather prognosis updated once every 24 hrs. This corresponded to a "normal" weather routing, where the captain is constantly updated with new weather prognosis, and adjusts their actions accordingly. It should be noted that the routing calculations were based on the prognosis available at the simulated time-points. WindShip performance was then calculated based on the real weather.
   In terms of fuel consumption a value in-between methods 1 and 2 should be the result. The closer to the values of method 2, the better the weather routing. If we are close to the fuel consumption values using method 1 the weather routing was ineffective.

The weather routing optimisation was performed by calculating a "route-tree", evaluating the different route-alternatives in terms of fuel consumption. To lower the number of alternatives a selection was constantly performed during the progress of the calculation, where solutions showing high fuel consumption were eliminated.

A typical example of a "route-tree" can be seen below, in Figure 69.

Figure 69. Look here please

Figure 69. Typical route-tree on leg 5. Optimal route marked with bold line.

As can be seen from Figure 69 above there were a few waypoints that the WindShip had to sail through on each route. They were placed by DMI/meteo in such a way to "steer" the route calculation around hindrances, such as southern India in the route shown above. As they effectively were a form of restriction on route choice, it is recommended that another "softer" steering algorithm is used in an eventual future re-analysis.

Results

The power consumption results from the weather routing were recalculated as fuel consumption numbers. They can be found in Figure 70 and Figure 71 in chapter 10.6 below. The results presented here will instead focus on the overall tendencies that could be observed.

Wind statistics

A large amount of wind data was compiled. Data on wind strength and direction was collected for each route at points 200 Nm apart.

Seasonal weather variations could be observed, clearly affecting the fuel consumption of the WindShip. Variations along each route could also be observed, typically a higher average wind speed can be found far out at sea.

For these preliminary calculations the average wind is the most important parameter, as it indicates how much energy is available for the sails. For each route the wind speed was averaged, resulting in Table 22 below.

Table 22. Look here please

Table 22. Three year average wind speed along each route.

Compared to phase 1 we see that it is only on route 11, New York – Rotterdam, where the average wind speed equalled 8 m/s. This was the assumed average wind speed from phase 1, used to determine the necessary sail area, engine installation etc. All other routes have significantly lower average wind speeds, ranging from 3.2 to 7.1 m/s. This fact significantly reduced the power available from the rig.

Fuel consumption

Trends in the fuel consumption were clearly visible, depending on route and season. Effects of using the four different weather routing methods described above were also significant. Generally the difference between method nr.1 and method nr.4 were not big enough to motivate the extra cost of weather routing. On the other hand method nr.2, the "optimal way" showed significant savings. With better predictions there is thus a significant fuel saving potential.

The fuel saving at higher ship speeds were somewhat disappointing. At 13 knots ship’s speed the modern WindShip even consumed more fuel on some routes than the standard product carrier. There was however a good reason for this. There is no possibility to sail effectively at 13 knots average speed on many of the routes, there was simply not enough wind. At the same time the propulsion unit was chosen mainly as an emergency and steering unit, not to propel the ship for extended periods of time. The small diameter (and thus cheaper) propellers are relatively ineffective at higher speeds, resulting in a higher fuel consumption.

Weather Routing, Conclusion

The weather routing performed by DMI/meteo served multiple purposes. It resulted in a realistic and large database for further calculations and comparisons. In this sense it was also used to provide information for economical comparisons with conventional ships, described further down in the report. More information can be post-processed from the large database in a future development of the WindShip.

The use of weather routing proved indispensable when realistically calculating the economy of deploying a WindShip. The tool developed and used at DMI/meteo is now available for further calculations on other WindShips. Further development of this tool, both on the optimisation algorithms and the weather prediction reliability would immediately show beneficial effects on the costs of using WindShips instead of conventional tonnage.

Fuel consumption numbers were somewhat disappointing, mainly due to the lack of favourable wind on the chosen routes. Further discussion on the subject can be found in section "10. Feasibility Study" in chapter 10. See also Appendix 10.

The amount of data made available through the calculations performed by DMI/meteo was so large that a complete post-processing has not yet been performed. In a future phase of the WindShip project the available data should be studied in more detail.