J.G. Vermeer
Feasibility study on the possiblities of implementing a computer based
scheduling application at Scaldis Reefer Chartering.
Masters thesis,
Report 96.3.LT.4793, Transport Technology, Logistic Engineering.
Scheduling of vessels is the assignment of cargoes (demand for capacity) to
vessels (supply of capacity), in such a way that all technical and time
constraints are met, and the pre-set goal achieved.
The demand for transportation capacity, the cargo owners wanting to ship
their cargo to buyers, is a very dynamic factor in the scheduling process.
The available cargo quantities are never certain, changing from year to
year and from day to day. Each season though is characterised by a low and
a high season, the last being roughly from February through April/May.
The supply side of the market is largely characterised by the small amount
of shipping companies. Seatrade, of which Scaldis Reefer Chartering is a
subsidiary, is the world's largest supplier of transportation capacity. The
supply side has its own major uncertainty, being the position of the vessel
at any time in the future. Because of both the uncertainties from the
supply and demand side, scheduling far ahead in time is useless.
The chartering practice is the interaction between the demand and supply
side. Scaldis has its own people (brokers) who try to contract cargo, both
for the short and the long term.
Today, the scheduling of the vessels at Scaldis Reefer Chartering is still
done by hand. Since this process is time consuming and scheduling further
ahead than one voyage is too complicated to handle, this report
investigates the possibilities of implementing a computer based scheduling
application.
The investigation was started by creating a model of the scheduling
process and one of the future situation. These models showed that there is
a difference between 'soft' and 'hard' variables, each influencing the
outcome of the scheduling process. The hard variables can be formalised
and used by a scheduling application. The soft variables can either be
represented in the application by other hard variables, or the user of the
application has to make sure they are accounted for in any final solution.
Therefore, it is made clear that the user will always have to interact
with any future application, always being the one taking the final
decisions about the produced solution.
Based on the findings of the model formulation, a set of requirements has
been drawn up, with which any scheduling application should apply. These
requirements consist of three groups. The optimiser requirements demand a
quick calculation of large problems, solving as close to optimality as
possible. The system requirements mostly focus on user friendliness and
flexibility in the creation of solutions, while the user requirements call
for experienced schedulers who understand the methods underlying the
application.
In finding a possible application, a mathematical form of the scheduling
model had to be defined. It was then found that the branch and bound
algorithms are the most widely used solution methods for the kind of
models as defined. A general solution method cannot be used though, since
too much computing time would be needed. Specific algorithms have
appeared in literature each making use of a specific characteristic of the
problem formulation. The Elastic Set Partitioning algorithm, or X-System,
is one of them. The X-System has been found to be the only algorithm that
is probably capable of solving the scheduling problem, and has at the same
time been used on a commercial basis. Besides this, algorithms developed
for crew scheduling problems are thought to be useful, but more study is
needed to find out whether any specific application is availabie. The
development of a custom made application, based on algorithms available in
literature, is not practical, since too much time and money will be
involved.
Based on the layout of the so-called 'MOS' application, in which the
X-System is incorporated, a future layout of the Scaldis scheduling
software has been designed. Besides a central database, the set up will
consist of three main modules:
- ship schedule generator;
- cost calculator;
- optimiser.
The ship schedule generator calculates all possible combinations (ship
schedules) between vessels and voyages. The cost calculator calculates the
costs of all ship schedules and the optimiser finally calculates the
optimum set of ship schedules.
The design of the cost calculator proved to be one of the most important
factors of success to the program. The schedule costs, all costs for 'idle
time' and final position of the vessel, are based on stochastic, so
uncertain, variables, and will play a central role in the production of
practically acceptable schedules. More research should be carried out to
make good results possible.
The user will at last make the difference between a useful or useless
program. He will be the one who is abie to pre-assign variabies and thus
manipulate the results, in such a way that the produced schedules comply
with the wishes of the management.
Testing the MOS application proved to be impossible for most requirements,
except for the calculation speed, which is acceptable. Other requirements
can be complied with based on reasonable assumptions.
Only a very small fleet result improvement is necessary to pay off initial
investments. It is therefore reasonabie to expect good results from an
implementation of the X-System. Before implementing the X-System though,
it is recommended to check the companies references, since reasons for
doubt did appear.
Reports on Logistic Engineering (in Dutch)
Modified: 2001.01.22;
logistics@3mE.tudelft.nl
, TU Delft
/ 3mE
/ TT
/ LT.