See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/4053670

Modeling Ship Arrivals in Ports

Conference Paper · January 2004

DOI: 10.1109/WSC.2003.1261627 · Source: IEEE Xplore

CITATIONS READS

34 495

4 authors, including:

Rommert Dekker
Erasmus University Rotterdam
299 PUBLICATIONS   13,486 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Managing Reverse Logistics or Reversing Logistics Management? View project

Pro-active service logistics for capital goods – the next steps (ProSeLoNext) View project

All content following this page was uploaded by Rommert Dekker on 27 May 2014.

The user has requested enhancement of the downloaded file.

Proceedings of the 2003 Winter Simulation Conference
S. Chick, P. J. Sánchez, D. Ferrin, and D. J. Morrice, eds.

MODELING SHIP ARRIVALS IN PORTS

Eelco van Asperen

Rommert Dekker
Mark Polman
Henk de Swaan Arons

Rotterdam School of Economics

Erasmus University Rotterdam
PO Box 1738
3000 DR Rotterdam, THE NETHERLANDS

ABSTRACT The model used in this report focuses on the analysis

of ship waiting statistics and stock fluctuations under dif-
Ports provide jetty facilities for ships to load and unload ferent arrival processes. However, the basic outline is the
their cargo. Since ship delays are costly, terminal operators same: central to both models are a jetty and accompanying
attempt to minimize their number and duration. Here, simu- tankfarm facilities belonging to a new chemical plant in the
lation has proved to be a very suitable tool. However, in port Port of Rotterdam. Both the supply of raw materials and
simulation models, the impact of the arrival process of ships the export of finished products occur through ships loading
on the model outcomes tends to be underestimated. This ar- and unloading at the jetty. Since disruptions in the plant’s
ticle considers three arrival processes: stock-controlled, production process are very expensive, buffer stock is
equidistant per ship type, and Poisson. We assess how their needed to allow for variations in ship arrivals and overseas
deployment in a port simulation model, based on data from a exports through large ships.
real case study, affects the efficiency of the loading and In the case study two types of arrival processes were
unloading process. Poisson, which is the chosen arrival considered. The first type are the so-called stock-controlled
process in many client-oriented simulations, actually per- arrivals, i.e., ship arrivals are scheduled in such a way, that
forms worst in terms of both ship delays and required stor- a base stock level is maintained in the tanks. Given a base
age capacity. Stock-controlled arrivals perform best with re- stock level of a raw material or product, the time to fill up
gard to ship delays and required storage capacity. or empty the tanks and the tonnage of the next arriving
ship, a planned arrival time of this ship is calculated. The
1 INTRODUCTION second type of arrival process is based on equidistant arri-
vals of ships per ship type. In this article we add a third
In this article we investigate the impact of a number of kind of arrival process that was not considered in the origi-
different arrival processes for ships on the efficiency of nal case study: a Poisson process.
the loading and unloading process in a harbor. This study The subsequent arrival times are actually expected ar-
was performed using some data from a confidential case rival times but ships will seldom meet this schedule. For
study in the Port of Rotterdam. The tender of that case this reason early and late arrivals are modeled by distur-
study provided detailed data on the types and numbers of bances generated for the estimated times of arrival (ETA)
ships to be handled per year, but did not specify their tim- resulting in the actual times of arrival (ATA).
ing (hereafter referred to as the arrival process). The en- The three arrival processes will be compared in this ar-
gineering firm responsible for the tender evidently did not ticle, using data from the original case study. With respect
realize its importance. to the original case study, some simplifications apply. For
In the original case study, a simulation model was reasons of confidentiality, the diversity of ships has been
used to optimize and evaluate various scenarios for the skewed down, and their numbers modified. Still, the result-
jetty and tank layout for the loading and unloading process ing model is general enough to draw conclusions applica-
of raw materials and finished products. Due to unforeseen ble to many jetty simulation studies.
business events (including a takeover of the company) the After a literature review in Section 2 we continue in
plant was built six years later, and no feedback on how the Section 3 with a detailed discussion on the loading and
results were used has been given. unloading process: the layout of the jetty where ships

1737
 Van Asperen, Dekker, Polman, and De Swaan Arons

unload raw materials or load finished products, the factory 3.1 The Jetty
which converts raw materials into products, the tanks that
hold raw materials or finished products, and the arrival of Central in the loading and unloading facility to be simu-
ships. In this article we focus on the impact of various lated are a number of mooring points. In this case there are
types of arrival processes on the efficiency of the loading four mooring points (mooring point 1 to 4) in a T-shaped
and unloading process. We therefore discuss the various layout (Figure 2). They differ in a number of aspects. One
arrival processes in more detail in Section 4. The imple- of these is the length of the ships that the mooring point
mentation model is the subject of Section 5, the experi- can handle. Mooring points 1 and 2 are suited to long
ments carried out with it and their results are discussed in ships; mooring points 3 and 4 can handle only short ships.
Section 6, and the conclusions are presented in Section 7. The mooring points also differ in their ability to load
and/or unload different materials (raw materials A or B,
2 A LITERATURE REVIEW and finished products C or D). For example, mooring
point 1 can handle A, B and C, and mooring point 2 can
Little has been published on the simulation of port facili- only handle products C and D. In the original case study,
ties, apart from some very scattered material. There is a several jetty layours were compared: here we consider just
nice book edited by Van Nunen and Verspui (1999) on one layout.
simulation and logistics in the port, but it is in Dutch only.
We briefly recapitulate the literature review on jetty design mooring point 1 mooring point 2
from Dekker (1999) in that volume. Well-known to insid-
ers are the reports from UNCTAD (1978) on the design of A B C C D
jetties. They report results from both queuing theory and
simulation applied to the capacity of jetties. The reports are A B C A B D
however difficult to obtain and they give yardsticks for
simple cases only. The other papers more or less describe mooring point 3 mooring point 4
that they have done a simulation study, without trying to
generalize their results. We like to mention Philips (1976)
and Andrews et al (1996), who describe the planning of a
crude-oil terminal, Baunach et al (1985), who deal with a
coal terminal, Van der Heyden and Ottjes (1985), Ottjes
(1992), and Ottjes et al (1994), who deal with the setup of Figure 2: Jetty Layout
the simulation programs for terminals. None of the papers
however deals explicitly with the arrival process. 3.2 Raw Materials and Finished Products

3 THE MODEL After being unloaded, raw materials are stored in tanks A
and B, from where they are withdrawn by the factory. Fin-
The model comprises the arrivals of ships, a jetty with a ished products are transferred to tanks C and D, to be
number of mooring points, storage tanks and a factory. loaded into ships.
These are briefly described in this section. Figure 1 pro-
vides a schematic outline of the model as a whole. 3.3 Tanks and Stocks

Tanks can be used for just one type of raw material or fin-
A ished product. The transfer of products from ships into
unloading
tanks, from tanks to the factory, and from the factory into
the tanks are continuous processes. In reality, there are
B several restrictions that affect actual tank operations, e.g.
J
e
no simultaneous pumping and running into and out of a
t
Factory
tank. We ignore these restrictions in our model, because
t
y they do not affect the comparison between the arrival proc-
esses (the original case study did model these restrictions).
C The same holds for stocks; for simplicity we allow the
loading stocks to take on any value (including negative values),
and neglect ship delays because of stock outs or lack of ul-
D lage (available tank space).
Figure 1: A Schematic Outline of the Loading and
Unloading Process, with Jetty, Tanks and Factory

1738
 Van Asperen, Dekker, Polman, and De Swaan Arons

3.4 Ships specifying, for example, annual amounts of raw product to

be delivered in equal batches every n weeks.
Ships (sea-going vessels, short-sea shipping vessels, and In our model, equidistant arrivals imply that arrivals of
inland barges) unload raw materials or load finished prod- ships within a ship type are assumed to be evenly spread
ucts. Each ship has four defining properties: the physical over the year. For example, per year, twelve vessels carry-
length (short or long), the cargo capacity (tonnage), the ing 6000 ton of product B arrive (see Table 1). With equi-
type of cargo it can handle (each ship can handle just one distant arrivals, this means a 1-month inter-arrival period
specific type of cargo), and (un)loading time (in hours). between such ships.
Ships are categorized based on the type of cargo they Both strategies actually yield a series of expected times
can carry. Loading or unloading can only be done at a of arrival (ETAs). However, in reality ships will seldom
mooring point that can handle a ship’s length and product. meet this schedule. For this reason disturbances to the ETAs
When a ship has arrived in the port, a suitable mooring are generated, modeling early and late arrivals resulting in
point is selected according to specified rules (e.g. if several the actual time of arrival (ATA) of each ship. Figure 3
suitable mooring points are available, select one of these shows the distribution of disturbances to the ETA of a ship:
randomly), including a priority scheme (see Section 5.3). all ATAs are within a margin of twelve hours before and
twelve hours after the corresponding ETA. Eighty percent
4 THE ARRIVAL PROCESS of these are within a margin of 2 hours before and 2 hours
after the corresponding ETA (these values were set together
Arrivals in client-oriented processes are often assumed to with shipping experts.)
be based on a Poisson process. The underlying assumption
is that the process cannot be controlled. Simulation lan-
guages and environments tend to offer this as a first-choice
80%
option for the specification of arrival processes. As men-
tioned above, we have looked at three scenarios:
1. Stock-controlled arrivals; 10% 10%

2. Equidistant arrivals; -12 -2 2 12

3. Arrivals according to a Poisson process. Figure 3: Distribution of Disturbances to Ex-

Stock-controlled arrivals aim at maintaining a target pected Times of Arrival
base stock level of raw material and finished product in the
tanks. For the loading process, this implies that the arrival The third arrival process considered in this article is a
time of the next ship is planned to coincide with the mo- straightforward Poisson process: within each cargo type,
ment that, through production, there is sufficient stock in ships arrive with exponentially distributed interarrival times.
the tank to load the ship without dropping below base stock In reality, the arrival of a ship is known, sometimes
level. In this calculation, the parameters are the loading days beforehand, to the plant. This can be used in a moor-
time of the present ship, the cargo capacity and loading ing point allocation system based on priorities. This is fur-
time of the next ship, and the production capacity of the ther explained in Section 5.3.
factory. Setting the appropriate base stock level for a tank
involves an estimation of the tendency of ships to arrive 5 THE IMPLEMENTATION MODEL
ahead of schedule (see below), this being the only threat to
maintaining base stock level. The simulation model has been implemented in Enterprise
For the unloading process, maintaining base stock lev- Dynamics, a simulation package for discrete-event simula-
els in the raw materials tanks is achieved by planning the tion. The implementation model, see Figure 4, comprises
next ship’s arrival to coincide with the moment that, various types of atoms (atoms are the Enterprise Dynamics
through extraction of raw material during production, base equivalents of objects). Some of the atoms implement the
stock level will be reached. In this calculation, the parame- simulation’s logic, others hold the simulation’s data (ta-
ters are the cargo capacity of the present ship, and the rate bles), define the types of experiments or store the output
at which the factory extracts material from the tank. Here, (e.g., graphs).
the danger of stock dropping below base stock level comes
from ships arriving late. 5.1 The Logic
For each product, the order in which the ships of differ-
ent types arrive can be determined in several ways. Here we The Generator atom is responsible for generating ship arri-
made a random selection with stratification for each ship type vals. Upon arrival a ship proceeds along the atom Arrival
to make sure that a fixed number of ships arrive per year. Route (the vertical atom in the middle) to one of the four
Equidistant arrivals model a situation in which loading mooring points that suits its length and cargo type (see
and unloading ships arrive at regular intervals. This regu- Section 3.4). If all suitable mooring points are occupied,
larity could be the consequence of year-based contracts the ship will wait in one of the queues (Queue 1, 2, 3 or 4).

1739
 Van Asperen, Dekker, Polman, and De Swaan Arons

An important table is ArrivalTimes, which, for each

category of ships, provides the ETA and the ATA (which is
derived by disturbing the ETA according to the distribution
function outlined in Figure 3). The Generators table con-
tains the arrival process to be used for each ship type.
The Ships table contains specific ship data such as
type, size, length, type of cargo, loading time, and the
number of ships of this type arriving annually.
The table JettyProducts describes which type of prod-
ucts can be handled by which mooring points, whereas the
table JettyLengths holds the lengths of the mooring points.
The base stock levels of the tanks are stored in table Tanks.
Table Factory specifies the number of tons of raw materi-
als to be processed, and finished products to be produced,
both on a yearly basis. Table SimulationSettings holds
some data concerning the distribution function used for
Figure 4: Implementation of the Simulation Model disturbing ETAs.
The AnnouncedShips, WaitingTimes, and TankLevel
Raw materials are unloaded and transferred to either atoms are used to store data collected during the simula-
Tank A or B, from which they are withdrawn by the Fac- tion run. The Graph Tank atoms visualize the tank levels
tory atom. The factory stores finished products in Tank C over time.
and D, from which they are withdrawn to be loaded into
ships. After loading or unloading the ships leave the sys- 5.3 Priority of Ships
tem via the Sink atom (Sink is the standard Enterprise Dy-
namics atom at which entities leave a simulation model). The priority scheme used in the selection of a suitable
It is worthwhile mentioning that the stock of the tanks mooring point for a ship is based on the following:
is not modeled as a continuous variable, but is updated at • There are only two priorities (high and low);
discrete intervals (every two hours). As stated before, for • Long ships have high priority, short ships have
this study, we assume that the process is not limited by the low priority.
capacities of the tanks. As a consequence, we can model Ships with low priority do not get assigned to a moor-
storage by using tanks with unlimited capacity and with the ing point when a high priority ship is known to arrive
possibility to contain negative stock. This simplification within the next 48 hours and would be assigned to the
does not affect the simulation’s objective. same mooring point within the time frame that the low pri-
The arrival and queue atoms contain specific pro- ority ship will still be busy with (un)loading.
gramming code refining their default (i.e. as defined in En-
terprise Dynamics) logic. The others are custom developed 6 EXPERIMENTS AND RESULTS
to perform dedicated tasks. Finally, the atom Initialize con-
tains code to be executed prior to each simulation run. The implementation of the model outlined in the previous
Global parameters for the simulation experiment are section has been used to carry out experiments. While it is
set using the Experiment atom. This atom contains several capable of generating results on a variety of topics, and on
PFM atoms (Performance Measure), each defining one many levels of detail, we focus on the ones relevant to our
output variable of interest. (Most of the PFM atoms are not objective: assessing the impact of using different arrival
shown in the figure.) For example, PFM1 till PFM4 pro- processes on ships’ waiting times and stock levels. In do-
vide the differences between the highest and lowest stock ing so, some level of detail is maintained, in that we make
data of the tanks. a distinction between waiting times for high priority and
for low priority ships.
5.2 Data
6.1 Simulation Parameters
The remaining atoms are mainly tables providing data for
the simulation process. An important reason for using tables For each of the three arrival processes, a ten-year simulation
is that they can easily import input data from an external re- run is conducted. With the equidistant and stock-planned ar-
source such as a spreadsheet into the model, and export the
rival processes, year-based stratification is applied to ship
simulation results to another spreadsheet for later analysis.
arrivals (for the Poisson process this is not possible). This
Spreadsheets as a source of input data and storage mecha-
means that the total number of ships of each type is fixed per
nism for simulation results are easy to maintain and provide
year (see Table 1), and aligned with the factory’s production
more flexibility (e.g. in modeling the arrival processes).

1740
 Van Asperen, Dekker, Polman, and De Swaan Arons

capacity, yielding a constant base stock level over time (i.e. 6.2 Results and Analysis
stock does not structurally increase or decrease over one
year, hence not over ten years). Year-based stratification is Tables 2 and 3 show the relevant simulation outcomes. Ta-
consistent with reality, in the sense that many contracts con- ble 2 contains the waiting statistics for ships, per arrival
cerning transport of raw material and finished products are process, each divided into separate results for high and low
based on specified quantities per year (often to be shipped priority ships. Table 3 reports on the maximum and mini-
in, for example, monthly batches). mum stock levels reached for each of the arrival processes,
The simulation starts in a steady-state situation, with the both in raw material and finished product tanks. Standard
tanks filled to base stock level. This eliminates the need for a deviations values are based on a comparison of the out-
warm-up period, which has consequently been omitted. comes for each of the ten years.

Table 1: An Example of the Variety of Ships 6.2.1 Waiting Times

Shiptype Size Length Product (Un)load Number
(tons) (type) (hours) per year
From Table 2, it can be observed that the choice for a par-
1 Barge 1,500 short A 6 196 ticular arrival process has significant impact on the number
2 Vessel 2,000 short A 8 48 of waiting ships and the number of hours spent waiting by
… … … … … … these ships. With Poisson arrivals both numbers are higher
8 Vessel 6,000 short B 26 12 than those observed with equidistant and stock-controlled
… … … … … …
arrivals. This holds for both high and low priority ships.
Clearly, the lack of a mechanism to keep ships apart,
10 Vessel 2,000 long C 14 126
whether it be equidistant or stock-controlled arrival plan-
… ning, allows for clusters of ships arriving within a small
13 Vessel 10,000 long D 44 14 time frame, causing queues.
14 Vessel 20,000 long D 56 8 Table 2 reveals a noticeable difference between the
outcomes of equidistant arrivals and stock-controlled arri-

Table 2: Ship Statistics per Arrival Process (Means over a 10-Year Period)
Ship Priority
Low High
Mean St. dev. Mean St. dev.
Poisson Arrivals:
Total number of ships 1,174 31 205 16
Percentage of ships that had to wait 39.5% 3.2% 18.9% 3.4%
Average waiting time of ships that had to wait (hours) 9.1 1.5 14.1 1.6
Equidistant Arrivals:
Total number of ships 1,163 0 208 0
Percentage of ships that had to wait 28.7% 0.7% 9.2% 1.5%
Average waiting time of ships that had to wait (hours) 7.2 0.2 9.8 1.2
Stock-controlled Arrivals:
Total number of ships 1,163 0 208 0
Percentage of ships that had to wait 14.2% 0.9% 8.5% 1.4%
Average waiting time of ships that had to wait (hours) 3.8 0.3 10.0 2.3

Table 3: Stock Levels Ranges per Arrival Process (Means in Tons over a 10-Year Period)
Tank
A B C D
Mean St. dev. Mean St. dev. Mean St. dev. Mean St. dev.
Poisson 88,342 21,937 46,741 13,252 41,470 15,041 116,812 39,839
Equidistant 10,756 273 11,265 342 3,381 283 27,474 574
Stock-controlled 6,702 474 5,893 296 2,945 340 15,552 682

1741
 Van Asperen, Dekker, Polman, and De Swaan Arons

vals. For both low and high priority ships, the stock-
controlled arrival process ‘outperforms’ the equidistant ar-
rival process.
The explanation for this is manifold. For one, stock-
controlled arrivals are more efficient overall since they tend
to keep ships of identical cargo types apart, whereas equidis-
tant arrivals keep ships of identical types apart. With multi-
ple ship types per cargo type this is an advantage.
Furthermore, simulation-specific factors have to be
taken into account. Consider the arrival rates of the indi-
vidual ship types. Here, care has been taken to avoid intro-
ducing unrealistic queuing situations. With equidistant ar-
rivals, for example, special measurements seek to prevent
the scheduling of arrivals for multiple ship types in such a
way, that they all coincide several times a year. Not all
such mechanisms are that obvious though, especially when Figure 5: Level of Tank D During One Year with Pois-
related to another simulation-specific aspect: the jetty lay- son Process
out. The combined effects of these factors are still subject
to further research.
However, the observed differences in waiting time sta-
tistics among arrival processes, whatever their causing fac-
tors, clearly demonstrate the need for careful arrival proc-
ess modeling, which is this article's primary objective.
Obviously, arrival process modeling requires a careful look
at the real situation, involving expert input on many sub-
jects. Only then are simulation results valid, and can they
be used in corporate decision-making. Alternatively stated,
providing only the numerical data from Table 1, and
throwing in a Poisson process, is simply insufficient, ren-
dering any subsequent decision (for example on expensive
alternative jetty layout to reduce waiting times) ill founded.

6.2.2 Stock Levels

Figure 6: Level of Tank D During One Year with Equi-
Table 3 shows ten-year stock level statistics in terms of the Distant Process
difference between minimum and maximum levels
reached. Poisson arrivals allow for the broadest stock fluc-
tuations. In fact, since Poisson arrivals constitute an uncon-
trolled process, stock range values are theoretically un-
bounded. This is not the case for stratified equidistant and
stock-controlled arrivals. However, with equidistant arri-
vals, considerable fluctuations are still observed, necessi-
tating high base stock levels to avoid stock outs.
Figures 5, 6 and 7 show example stock behavior over
time for product D over a one-year period (notice that the
scale of the vertical axis varies). Figure 6 shows that fluc-
tuations are such, that the initial stock level for product D
(2000 tons) does not suffice to avoid stock outs.
The stratified arrival processes are aligned with pro-
duction in such a way, that stock does not structurally grow
or shrink over a one-year period. Any difference between
stock levels at the start or the end of a year are due to ships Figure 7: Level of Tank D During One Year with
still being loaded and unloaded at the end. This does not Stock-Control Process
hold for Poisson arrivals, as is evident from Figure 5.

1742
 Van Asperen, Dekker, Polman, and De Swaan Arons

Figure 7 clearly shows the typical stock fluctuation Finally, we intend to consider yet another arrival proc-
pattern for stock-controlled arrivals. Peak levels are ess, a hybrid one, with planned arrivals for the larger vessels
reached whenever large ships are scheduled to arrive for and equidistant or Poisson arrivals for the smaller barges.
loading. In fact, the largest available vessel (see Table 1) More information on this study is available online via
comes in to load product D eight times a year, which ex- http://www.few.eur.nl/few/research/eurf
plains the eight peaks in the Figure. ew21/m&s/article/jetty/. The website contains
Notice that in the case of product D, stock fluctuation graphs showing the levels of all tanks over a one year pe-
is almost completely determined by the size of this large riod and a video that shows a simulation run.
vessel, which makes it easy to determine the required
tank capacity. ACKNOWLEDGMENTS
So, again, the choice of arrival process is an important
factor in simulation outcomes. For example, should the The authors would like to thank the students Nees Jan van
simulation be part of a cost-benefit analysis to the acquisi- Eck, Arthur Oink, Gerard Seedorf and Ludo Waltman for
tion of additional tankage, then its results are of no value their solid implementation of the simulation model in En-
without realistic arrival process modeling. terprise Dynamics and for carrying out the experiments.
They are also grateful to Stef Kurstjens for his work in the
6.2.3 Jetty Utilization original case study.

The jetty utilization varies little over the three arrival proc- REFERENCES
esses. This is due to the fact that with all three processes,
roughly the same number of ships is generated. In fact, year- Andrews, S., F.H. Murphy, X.P. Wang, and S. Welch.
based stratification with equidistant and stock-controlled ar- 1996. Modelling crude oil lightering in Delaware Bay,
rivals causes ships to be generated in identical numbers and Interfaces 26(6): 68-76.
types. Differences in jetty utilization follow from differences Baunach, G.R., E.S. Wibberley, and B.R. Wood. 1985.
in end-of-year situations among simulation runs. Simulation of a coal transhipment terminal: Batam Is-
land, Indonesia. Math. Comp. Simul. 27: 115-120.
7 CONCLUSIONS AND Dekker R. 1999. Simulation of jetty and storage activities
FURTHER RESEARCH for oil and chemicals. In: SimLog, Simulation and lo-
gistics in the harbor, ed. J. Van Nunen and L. Verspui,
The importance of carefully modeling the arrival processes 105-116. Eburon. Delft (in Dutch).
is clearly demonstrated in this article. The often-used Pois- Enterprise Dynamics. 2003. Documentation material. Incon-
son process has by far the worst performance of the three trol Enterprise Dynamics. More information is available
processes discussed, both in terms of the waiting times and online at <www.enterprisedynamics.com>.
in terms of the required storage capacity, whereas the stock- Heyden, W.P.A. van der, and J.A. Ottjes. 1985. A decision
controlled process performs best overall. Although these re- support system for the planning of the workload on a
sults were obtained in a specific case, we think that they are grain terminal. Decision Support Systems. 1: 293-297.
general enough to be appropriate for many port and jetty Nunen, J. van, and L. Verspui, 1999. SimLog, Simulatie en
simulation studies. As soon as there is some sort of control logistiek rond de haven, (translation: SimLog, Simula-
over arrivals, it should be incorporated in the model. tion and logistics in the harbor), Delft, Eburon (in
Obviously, the challenge in managing logistical proc- Dutch).
esses will be to determine which arrival processes can be Ottjes, J.A. 1992. Modelvorming en simulatie van logistieke
actually realized. This requires close collaboration between systemen. Manual i76C, Delft University (in Dutch).
production, logistics and the sales or marketing functions Ottjes, J.A., S. Hengst, and W.H. Tutuarima. 1994. A
within a company. If such cooperation is lacking, a market- simulation model of a sailing container terminal ser-
ing department might buy or sell large quantities to meet vice in the port of Rotterdam. In: Proceedings of the
sales targets, causing serious disruptions in planned arri- Conference on Modeling and Simulation, ed. Guasch
vals, yielding costly delays. and Huber, 876-880.
There are various directions in which future research is Philips, O.O. 1976. Optimization models for a crude oil
planned. First, the role of the jetty’s layout needs to be ex- storage export system, Ph.D. thesis, Penn. State Univ.
plored, specifically the impact of limited length of the in- UNCTAD. 1978. Port development: a handbook for plan-
dividual mooring points, and the restrictions on the avail- ners in developing countries. New York UN,
ability of piping for specific products. TD/B/C.U/175, chapter II: the break-bulk berth group.
Also, the effects of using the allocation scheme for as- p. 108-128.
signing ships to mooring points, including the priority
scheme based on 48-hour lookahead, requires further study.

1743
 Van Asperen, Dekker, Polman, and De Swaan Arons

AUTHOR BIOGRAPHIES

EELCO VAN ASPEREN graduated in Business Computer

Science at Erasmus University Rotterdam in 1993. From
1991 to 2000 he was a member of the IT support group at
the Erasmus University Rotterdam, for the department of
Computer Science and from 1995 for the Faculty of Eco-
nomics, focusing on the design and implementation of large
scale computer facilities. Since January 2000 he is an assis-
tant professor at the Department of Computer Science of the
Faculty of Economics at Erasmus University Rotterdam. His
research focuses on simulation with applications in logistics.
You can reach him by e-mail at <vanasperen@few.eur
.nl>; and his web address is <http://www.few.
eur.nl/few/people/vanasperen/>.

ROMMERT DEKKER is a full professor in operations re-

search at the Econometric Institute of Erasmus University
Rotterdam. He obtained his Ph.D. in operations research at
the State University of Leiden, and his M.Sc. degree in in-
dustrial engineering from Twente University of Technology.
He worked with Shell for seven years on reliability and re-
finery logistics. His current research interests are: mainte-
nance and logistics (inventory control, spare parts, ports,
containers and reverse logistics). He has applied simulation
models in various logistical problems. His e-mail address is
<rdekker@few.eur.nl>; his web address is <http:
//www.few.eur.nl/few/people/rdekker>.

MARK POLMAN graduated in Business Computer Sci-

ence at Erasmus University Rotterdam in 1993. Since
January 2000 he is an assistant professor at the Department
of Computer Science of the Faculty of Economics at Eras-
mus University Rotterdam. Previous research areas include
machine learning and communication modeling in distrib-
uted systems. His research focuses on discrete-event simu-
lation. His e-mail address is <polman@few.eur.nl>;
his web address is <http://www.few.eur.nl/few/
people/polman>.

HENK DE SWAAN ARONS is an associate professor at

the Department of Computer Science of the Faculty of Eco-
nomics at Erasmus University Rotterdam. He graduated in
Applied Mathematics at Delft University of Technology in
1972. In 1991 he obtained his Ph.D. degree in computer sci-
ence at Delft University of Technology. The thesis was
mainly concerned with the design, applicability and applica-
tions of expert system tools. His research focuses on dis-
crete-event and continuous simulation, with the emphasis on
economical applications. His e-mail address is <deswaan
arons@few.eur.nl>; his web address is <http://
www.few.eur.nl/few/people/deswaanarons/>.

1744

View publication stats