Académique Documents
Professionnel Documents
Culture Documents
Weather-related ship loss statistics are assessed as a proportion of total losses to ascertain the
requirement for route planning. A routeing assessment for suitability of on-board application
is undertaken and a moveable grid is then proposed incorporating a time optimization pro-
cedure for the on-board system.
4 t 7 is-2
41 8 R. M O T T E AND S. CALVERT VOL. 41
Loss
Stress of weather
founderings
Totals and abandonment
The ratio of weather-related losses (i.e. stress of weather, foundering or abandonment) to all
losses varies from 02 to 047 but is generally above one third. It has been and continues to be
a growing category, relatively, as a reason for ship loss. (Source: The Underwriters'
Association.)
Effective avoidance must lead to safer and smoother voyages and on occasions,
as a bonus, to minimizing passage time. Better predictability of arrival time, for
vessels with massive unit loads, indicates a further requirement. The general
trend in the 1980s is for short holding time in port with faster, more specialist
carrier units. Expensive terminal facilities are provided with a timetabled usage
by a consortium. The programming of such an operation has to be done with a
minimum tolerance of voyage time fluctuation; thus each vessel's schedule is
important to the whole scheme. There is little requirement to reduce passage
time; the demand is not to increase time on passage by, for instance, heaving to
in a storm for a day. This tendency towards specialization and through transport
systems suggests a demand for strategic routeing. The long-term savings accruing
by reducing heavy-weather damage both to ship and cargo on strategic routes are
the main reason, however, for adopting routeing procedures. The penalties of
' off-hire' clauses relating to overdue vessels on time charters, the carriage of
sensitive cargoes and susceptibility of damage to vessels in ballast, are additional
arguments for a storm avoidance approach.
Thus on-board routeing should be conducted with the fundamental philosophy
of storm or gale avoidance. A rigorous analysis of an individual vessel's
behavioural characteristics and, therefore, performance related to relative wind
force and direction should be undertaken to learn and catalogue the limitations
of environmental conditions to which that vessel should be exposed. It then
follows that a good estimate should be made of:
(i) the short-term prevailing conditions that the vessel may immediately
experience on departure;
(ii) the medium-term or mid-ocean development of depressions;
(iii) the long-term or far ocean features likely to occur before the vessel
completes her passage.
It is of limited value to attempt to route vessels on the evidence of surface
information alone. It is always necessary to have an indication of storm movement
and future development, particularly for the second half of the passage. Such
indication is only attainable with the assistance of upper air information to
evaluate the growth and steering likelihood of depressions.
Optimization routines are generally based on a matching of ship performance
data with sea surface data, relating the seakeeping characteristics to an analysis and
the predicted sea wave fields. Unfortunately for the routeing analyst these
predicted fields are only available for up to two or a maximum of three days
ahead. Retrospective analysis using optimization techniques is, on the other hand,
a highly accurate tool. As accurate or prevailing sea data are not normally
available and the ' optimization' is undertaken using raw and often inaccurate
forecasted sea fields, after day one the term ' optimization' in such an application
may be perceived as a misnomer. It can lead to misinformation in that short-term
indicators of measured or calculated sea states will not accurately relate to the
position and/or intensity of middle latitude transient storms during the later
periods of a passage. Initial decisions on advised courses should relate to the
passage as a whole. Such subjective decision making, on board, can have
considerable advantages when backed up with an optimization programme, and a
420 R. MOTTE AND S. CALVERT VOL. 41
trans-ocean grid can be biased to the locality of the proposed route. For example,
should the 500 mb flow indicate major storm activity to the south of a
conventional route so that the prognosis advises a northerly passage, a grid can
be positioned so as to avoid unnecessary southerly coverage. This reduces the
need for high capacity machines for on-board operation and reduces the storage
of unwanted environmental data.
In devising a working algorithm for an ocean wide grid which, of necessity,
will use a set of sea states (Petri et al.5) for several time and space stages, it
is necessary to be economical with data. For example, a basic ship-based
program with, say, 10 space states and 10 time slots at each stage will give 100
variations at each stage and so there will be 10000 ways to travel. All routes must
be examined in the dynamic programming; it follows that any pre-planning
which reduces input data by causing a ship to proceed within a reduced grid size
and strategic decision making which reduces comparisons in the optimization
procedures, will assist the operation.
3. MODEL F O R M U L A T I O N . In designing an on-board system for optimal
weather routeing several aspects were dependant upon the capabilities of the
computer and its memory size. There is a need for a simple recursive algorithm
and model using accurate predicted sea state data and ship data. Dynamical
programming (DP) was therefore accepted as the basis for the initial trials for an
on board system, since:
(i) The model is a simple powerful tool for calculating the optimum route
across an ocean.
(ii) The computer capabilities could be easily matched to the size of grid
mesh required and the speed of computation. The Acorn Cambridge
Workstation which was used is quoted to be as fast as a lightly loaded
mainframe system but is a stand alone machine.
(iii) The methodology can incorporate more than one objective function and
more importantly, for micro-based systems, the calculation constraints
can be levelled both linearly and directly.
(iv) Grid points can be labelled by latitude, longitude, objective policy
decisions and environmental parameters.
The DP grid can be represented by a series of legs or zones, across an ocean,
where each zone consists of several latitude points. It was found convenient for
the grid to be geared to the shortest distance across the ocean, that is the great
circle route. Motte et al.,6 studied the grid mesh size for a rectangular grid
system, and found that a four degree of longitude by a half degree of latitude mesh
was adequate. In effect this grid stretches out the east—west dimension and
therefore, provides for a closer great circle fit.
Taking the great circle as the central line between departure and destination,
the DP grid can be constructed by taking perpendiculars at, for instance, ^8
intermediate points. Such a grid consists of 60 stages in the east—west direction
and can be constructed in the north—south direction to a number of points
depending upon the distance taken along the perpendicular from the great circle.
Mesh size is thus easily controllable and the discontinuous routes on a large grid
can be smoothed by reducing the north—south point spread.
NO. 3 WEATHER ROUTEING PROCEDURES 42 I
limited states). Beyond that value, voluntary speed loss is not included in the
algorithms although the vessel is required to 'heave-to'. If the weather is
persistent and time is removed for heave-to then the time available for the
remainder of the voyage may become zero. If such a situation occurs, typically
during the winter months, then an omission flag can be triggered withdrawing
calculations from that point. This restriction has to be waived near the start point
of a voyage as it may prevent the route reaching its goal during further
iterations.
In transiting a zone from an ' end' point, several ' next' points are looked to.
(See Fig. i.)
Course deviation is restricted in the program to a number of points either side
of the central or 'end' point, depending upon the grid mesh size. Power
limitations are governed by the set service speed and correspond with the speed
reduction limitations. The full program flow is shown in the Appendix.
4. DYNAMICAL PROGRAMMING WITHIN THE COMPUTER MODEL.
Dynamical programming was favoured over other possible models for its
simplicity and power. The methodology was first proposed in 1957 and the form
of the deterministic ship optimization equations are as follows:
terminal cost function J(Xn) = A(* destlnatlon ) (1)
Latitude steps
Fig.
The environmental inputs. The data used for the time optimization programs
were initially entered by manual /computer overlay block means to produce an
array data file for each day of weather. For ease therefore, a one degree grid
system was adopted and DP grid point weather data found by close matching of
the two grids. The data were taken from wave charts and consist of significant
wave heights and directions. The data are produced for each one degree grid
point in the ocean and since the optimal grid can be of the order of 2000 points,
each holding several parameters, the process becomes tedious. For analysis
purposes output from the UK Meteorological Office's wave model is used and
is more accurate and efficient. The number of data points is not increased but
rather the number of data parameters (see Fig. 2). There is a need to investigate
the computer calculation time for such a large data box, either to reduce the data
volume or expand the grid spacing. The data arrays can be represented by a three-
dimensional grid, as shown. Each layer of the grid box represents one data
parameter, where that parameter is defined by latitude and longitude. At the
present stage the data box is only two layers deep, representing wave heights and
directions. However, a simple empirical formula for wind speed and an assumed
wind direction are used to provide the inputs to ship response formulae. Wave
and wind data from the Meteorological Office model provides many more
accurate inputs, taking the grid to nine layers deep (see Fig. 2).
As stated by Chen et ah, the data field can be represented by a data plane,
where: <P = latitude; A = longitude; subscript 1,2,3 = points one, two and
three; and z = data parameter in use.
R. MOTTE AND S. CALVERT VOL. 41
i = . . . 11
r
i = 6
i = 5
i = 4-
/ = 3
i = 2
60
Grid points
i = 1
30
80 W 0 W
-«— Longitude
Fig. 2. Grid data box. i, significant wave height; 2, swell wave direction; 3,
significant wave period; 4, swell wave period; j , wind speed; 6, wind direction; 7,
swell wave height; 8, wind wave height; 9, wind wave period; 10, current speed; 11,
current direction
is:
(A-A,)
A2-At
A, —A,
A 2 -A,
(5)
The formulation of intermediary data values requires three points and these are
taken to be the two points involved in the transit plus one other next to the
destination point. Three intermediary points are used between transit points and
their latitudes and longitudes 0, A are found by thirding the distance between
transit points and inversing Mercator's formula. This then occurs for each data
parameter to give a closer representative value than a simple mean for the transit.
NO. 3 WEATHER ROUTEING PROCEDURES
The data plane formula (equation ($)) is used 31' times, where i is the number of
data parameters listed in Fig. 2. The inclusion of the added data will go hand in
hand with improved ship response algorithms. The ship response formula then
provides an optimal parameter for the intermediate zone transits.
The ship response. The response of the vessel in a seaway is of prime importance
to the weather optimal routeing programs. The development of the programs has
involved the speed loss curves proposed by James,9 however, the further
improvement of these formulae by Babbedge'' has also been included and it is
hoped that the next stage of development will be a simplified ship model to
describe all motions applied to the sea spectra by superposition techniques. There
will then be a balancing of power and resistance from which fuel and/or cost
objective functions can be found.
The James speed-loss curves describe the speed loss of a vessel under the
influence of wave height and direction. Typically there are three curves,
representing three sea regimes — head, beam and following — however, no tests
were made with this model as further advances were made to use the
Babbedge1' formula, which is described later. The advantages and disadvantages
of the James speed loss curves are :
(i) They can easily be described by third degree polynomials from
experimentally derived data. The information from curves supplied in the
Bales13 paper was tried out using the curve fitting routines.
(ii) The equations are simple and therefore are ideal for repetitive use in the
routeing programs since many calculations are made through the grid. It is
questionable if the size of the DP grid can still be used when the response section
is expanded to incorporate a ship model.
(iii) The curves can only be found by extensive experiments at sea, thus they
are individual and require much time and effort.
(iv) The curves only describe a wave encounter angle of 1 20 degrees for each
sea regime. Thus differing combinations of wave heights and directions will effect
the same speed reduction whereas in reality this is not so.
(v) The curves only utilize a small amount of data and, in fact, only use wave
direction as a pointer to which equation to use.
The form of the equations is as follows:
speed = ai + biHx + q Hi + </j W| (6)
Where Hi_ = significant wave height; i = (1) head sea, (2) beam sea, and (3)
following sea; a, b, c, d are coefficients of the polynomial.
The equations have been modified by Babbedge'' but still are individual to the
ship. Observations carried out aboard four ships were analysed statistically in
order to establish the main causes for ship speed loss and to incorporate them in
a speed reduction equation. The equation thus does away with the sea regime and
actively uses the wave direction as a parameter in the equations. The equations
also incorporate power, maximum speed in calm conditions, wind speed, wind
direction, wave height, wave direction, sea temperature and ship displacement,
but still fall short of fully describing the ship response since the seaway is only
described by significant wave height and primary direction. It is well known that
426 R. MOTTE AND S. CALVERT VOL. 41
sea spectra for many differing seas can have the same significant wave heights and
directions, thus the response of the vessel will be predicted to be the same in
cases where it can be grossly dissimilar. The Babbedge formulae still have the
same advantages of the previous approach, however, they improve the routeing
algorithms at this stage.
The form of the Babbedge equations are:
-10-2
r i
+ 0-04 (0— 12) — 0000 1 2 (A — 37ooo) + o-27 (7)
Ww = °°7S VI (8)
where Hw = significant wave height; and Kw = wind speed.
The formula can be used for a specific speed in calm conditions with due
respect to the maximum power available for the ship. Under the present program
for time optimization, power is constant and speed is maximized with respect to
the particular meteorological conditions.
Equation (7) formed the initial basis for trials as mentioned but it was felt that
the amount of calculation required might be unduly increased by variations for
sea temperature and/or ship displacement. A series of trials were then simulated
using varying significant wave heights, displacements and temperatures to
establish the effect these parameters have on the speed calculated. The wind
speed was formulated on the Scott formula, equation (8), as used by Babbedge.
However, this formula is not strictly accurate and the formula adopted by the UK
Meteorological Office incorporated a simple constant for wave height in the
presence of no wind. The standard curve for comparison was that without this
parameter (see Figs 3—6). The Babbedge equations can be rearranged to optimize
for power given a speed to be maintained across the ocean. This demonstrates the
versatility of such equations and their importance in micro based systems.
Equation (7) can be represented by:
giving,
^ = 6-44 l n P — (10)
K }
'" w a v e - K c a l m wind] ++ P
NO. 3 WEATHER ROUTEING PROCEDURES 427
•x—«
2000 1500
17-14- 12-86
_ 14-29 j 70-77
o
x 11-43 £ fl-57
2 8-57 g. *•«
5 71 4-29
206 2-74
0 0
«<J 72-65 76 87 2708 4-22 <J<J 72-65 76S7 27-06
Sig wave ht (M) Sig wave ht (M)
Fi F
g- 3 'g- 4
Fig. 3. Wave height versus speed for constant power with variations. Dart Atlantic:
maximum speed, 20 kt; d.w.t., 41 000: A, following sea; • , beam sea; + , head sea.
Wave factor: x , following sea; D, beam sea; • , head sea
Fig. 4. Wave height versus speed for constant power with variations. Dan Atlantic:
maximum speed, i j k t ; d.w.t., 41000. Temperature factor: x , following sea; n,
beam sea; • , head sea. For key to other symbols see Fig. 3
15 00 1500
12-86 12-86
10-71 70-77
613 6-43
4 29 4-29
2 14- 2 14
0 0
0 4-22 8-43 12-65 16-87 2108 4-22 8-4-3 12-65 16-87 21-08
Sig wave htlM) Sig wave ht (M)
F
'g- S Fig. 6
Fig. j . Wave height versus speed for constant power with variations. Dart Atlantic:
maximum speed, i j k t ; d.w.t., 41000. Displacement factor: x , following sea; n,
beam sea; • , head sea. For key to other symbols see Fig. 3
Fig. 6. Wave height versus speed for constant power with variations. Dart Atlantic:
maximum speed, 1 c kt; d.w.t., 41 000. Wind factor: x , following sea; n , beam sea;
• , head sea. For key to other symbols see Fig. 3
428 R. MOTTE AND S. CALVERT VOL. 41
19-66 r
1667
H-06
O
6-43
5-52
A T
0
I J
1660-8 3721-6 55624. 74-43-2
Max power
9304-0
Power (HP)
Fig. 7. Wave height versus power for constant speeds. Dart Atlantic: d.w.t., 41 000.
A, j kt; T, 10 kt; +, i j k t , encounter angle, 00
19 68 r
1687
14-06
11-24
0)
1 8-43
5-62
2-81
Max power
j
" 0 18608 3721-6 5582-4 74-432 9304 0
Power (HP)
Fig. 8. Wave height versus power for constant speeds. Dart Atlantic: d . w . t . , 41 000.
A, j k t ; T, 10 kt, + , i j k t ; encounter angle 1800
NO. 3 WEATHER ROUTEING PROCEDURES 429
Certain graphs have been produced to show the variation of power setting against
wave heights, to maintain a required speed (see Figs 7 and 8).
c. R O U T E I N G C O M P A R I S O N . Several runs were made for the North
Atlantic based on the Dart Atlantic formulae as described by Babbedge.'' These
runs were then compared with the routes advised by Oceanroutes (UK) Ltd, for
the same period of weather which was chosen to be during the winter months
(Nov. 86—Dec. 86). Obviously the ships routed are not identical and so large
discrepancies can occur. However, with this in mind the Dart Atlantic was set at
different service speeds and different loading conditions in order to give a
different actual vessel speed during transit. The deadweight tonnage (d.w.t.)
only provides one input to the speed estimate so different combinations of d.w.t.
and set service speed may provide the same speed estimate.
The information provided by Oceanroutes was the actual advised route sailed
(only open ocean), the type of vessel involved, the d.w.t. of the vessel and the
departure and destination times. It was hoped that the vessel matching could be
much closer, but the only possible matching was that of d.w.t. and not
service speed. The initial routes were run on a DP grid derived by taking points
at increments of 60 nautical miles from the great circle route. The routes advised
show some degree of discontinuity but changing the grid mesh size or
incorporating a smoothing algorithm through the rough route removes the
jaggedness.
A routeing case is shown in Fig. 9, west-bound from Cork (S. Ireland), to New
York (N. America). Several runs were performed with differing loading
characteristics and set service speed. The program runs therefore illustrate
theoretical vessels for the period in question. Fig. 9(6) illustrates a run for
42000 d.w.t. and a set service speed of i$kt, Fig. 9(c), however, illustrates
the same run for a 48 000 d.w.t. vessel at a service speed of 18 kt. The latter run
is intended to represent the Oceanroutes advised course as given to a 48 000
d.w.t. bulker. The routes are performed on the larger DP grid system and
therefore show a degree of discontinuity. However, Figs 10(a) and (b) show
the same runs on a finer grid, at 14 and 15 kt and it may be noted that smoothing
of the optimum route removes extreme values of course change.
6. C O N C L U S I O N . The intention of this work at Plymouth Polytechnic is to
produce an on-board advisory weather routeing package. In designing such a
system it is necessary to take into account the capabilities of the machine and so
this paper has tried to outline both the limitations and methods for utilizing that
capability.
It is suggested that advantages can be gained by using a simple optimal
algorithm and ship response algorithm for the calculation of optimum time
routes. However, there is a trade-off in this approach when further optimal
parameters are required. The moveable grid system is shown to reduce both
calculation time and required memory space and is considered an ideal method
for including constraints in the optimal calculations. There is a need for
observing the 500 mb flow details and biasing the grid depending upon the results
since accurate forecasts do not exist for a complete North Atlantic transit.
Optimal routeing for the whole voyage would require complete and accurate data
I I I I
30
80 75 70 65 60 55 50 45 40 35 30 25 20 15 10 5 0
(a)
60
55
80 75 70 65 60 55 50 45 40 35 30 25 20 15 10 5 0
(b)
SO 75 70 65 60 55 50 45 40 35 30 25 20 15 10 5 0
(c)
Fig. 9. North Atlantic crossing, (a) d.w.t., 39000; speed, I J kt: (b) d.w.t., 42000;
speed, I J kt: (c) d.w.t., 48400; speed, 18 kt. A, advised route; O, optimum route;
G, great circle
NO. 3 WEATHER ROUTEING PROCEDURES 431
40 Departure time
20-2
Departure Arrival
35
4900 700 40-40 62-40
30
SO 70 60 50 40 30 20 10 W-E
Current ship status
(a)
(b)
Fig. 10. (a) Smoothed d.w.t., 48000; speed, 14IU: (fc) smoothed d.w.t., 48000,
speed 1j kt
432 R. MOTTE AND S. CALVERT VOL. 41
for the transit, however this is not available and it is believed that observing a
steering level and and long range forecasts will provide better results than relying
upon climatological 'fill-in' data.
The authors believe that it is essential to concentrate on the physics of the
operation rather than to have the mathematics 'take control'. For example, the
problem of relating vessel response to the sea state demands in the optimization
routine a good estimate of sea conditions for the complete passage. How is this
to be best achieved? Assumptions have to be made that the prediction for the
forecasted period remains good when the next set of data takes over. Is this a
realistic proposition, or can the two sets of data be examined and a linear time-
varying sea state obtained for the period in question?
REFERENCES
1
Heijboer, D. (1974). Weather routeing — a modern aid to navigation. Fairplay International.
8 March.
2
Motte, R. and Laurence, C. A. (1985). Fuel consumption on container ships on the North
Atlantic. This Journal, 38, 2j8.
3
Motte, R. (1983). Ship based weather routeing using dynamical meteorology. This Journal,
36, 480.
4
Motte, R. (1973). Weather Routeing of Ships. London: Stanford Maritime.
5
Petrie, G. L., Bongort, K. J. and Maclean, W. M. (1982). A New Approach to Vessel Weather
Routeing and Performance Analysis. New York Metropolitan Section of SONAME.
6
Motte, R., Manhire, B. J. and Higham, J. (1985). Diagnostic and dynamical modelling for
weather routeing of ships using on-board micro-computers. Proc. Int. Conf. On Computer
Applications in the Operation and Management of Ships and Cargoes. Royal Institute of Naval
Architects.
7
Chen, H. T., Frankel, E. G., Fiore, A. E. and Carleton, H. (1976). Optimisation of ship
weather routeing. In Ship Operation and Automation Symposium, pp. 27—3 j (eds. Pitkin, Roche and
Williams).
8
Zoppoli, R. (1972). Minimum time ship routeing as an N-stage decision process. J. Appl.
Meteorol., 1 1 , 4 2 9 .
9
James, R. W . , (1957). Application of Wave Forecasts in Marine Navigation. US Naval
Oceanographic Office, Sp-i.
10
Townsin, R. L. and Kwon, Y. J. (1983). Approximate formulae for the speed loss due to
added resistance in wind and waves. J. Royal Institute of Naval Architects, 125, 199.
" Babbedge, N. H. (197J). Ship Speed Analysis. M. Phil. Thesis, Plymouth Polytechnic.
12
Bellman, R. (19J7). Dynamic Programming. New York: Princetown University Press.
Bales, S. L. (1976). The Use of Seakeeping in Ship Operations: A Status Report. D. W . Taylor
Naval Ship Research and Development Centre.
NO. 3 WEATHER ROUTEING PROCEDURES 433
V
Calculate data parameters
for intermediate points based
on data plane
T Within limits
END