Journal of Soil Science and Environmental Management Vol. 1 (4), pp.

77-85, June 2010

Available online at http://www.academicjournals.org/JSSEM
ISSN 2141-2391 ©2010 Academic Journals

Full Length Research Paper

Application of queuing theory to the container terminal

at Alexandria seaport
M. E. El-Naggar
Civil Engineering Department, Faculty of Engineering, Alexandria University, Alexandria, Egypt.
E-mail: [email protected].
Accepted 29 April, 2010

This paper describes a methodology designed to support the decision-making process by developing
seaport infrastructure to meet future demand. In order to determine an optimum number of berths at a
sea port, the queuing theory is applied in the light of port facilities and activities. The aim is to avoid
inadvertent over and under-building. Within this methodology, the movements in port should firstly be
analyzed. The waiting time of vessels outside the port and in queue is calculated in accordance with the
considered queuing model. The theoretical functions representing the actual vessel arrival and service
time distributions are determined. For the economic considerations, cost estimate studies including
cost of port and waiting vessels are carried out. Finally, the optimum number of berths that minimizes
the total port costs can be decided. Both proposed mathematical and economical models are applied to
Alexandria port in Egypt.

Key words: Port capacity, port economy, port modeling, queuing theory.

INTRODUCTION

The port transportation system includes different physical inefficient and uneconomic due to the delay costs of
elements, e.g. berths, handling equipments, storage and waiting ships. Conversely, developing the port so that
traffic facilities. Although the capacity of any single ele- ships are never forced to wait also represents an
ment may be expressed as an absolute figure, such as uneconomic use of port resources.
the number of containers loaded per hour by a certain The ideal situation is one in which all berths are
crane, the aggregate capacity of the whole port cannot be occupied at all times and no ship is ever kept waiting.
so simply described. Each element can limit the overall This situation is impossible to achieve in practice
port productivity. because of the random arrivals of cargo ships and the
Port productivity can be viewed from two standpoints. variations in service time of ships of different sizes.
To ship operators, productivity implies the time needed at Therefore, decisions concerning port development can be
the port to serve ships, while at national level, port made by trading-off the cost of increasing the port
productivity can be defined as the amount of cargo capacity and the costs of both waiting and service times.
transported through the port during a certain time period. The purpose of this paper is to introduce a methodo-
Port development is often affected by operating policies logy which can be used to facilitate the decision-making
as well as by the traffic demand imposed in the port in process of port development. The proposed methodology
terms of the volume of cargo expected to be accom- covers two principal areas:
modated, the service time at the available berths within
which this volume should be handled, and the frequency a) Investigation of the pattern of ship traffic at a seaport
of ships arrivals. from the standpoint of queuing theory, and to use the
It would be possible to develop the port facilities so that findings to draw some hypotheses regarding its
its capacity is fully utilized at all times. In this manner, application to the overall operation in sea ports.
changes in demand have to be accommodated by forcing b) Determination of the optimum number of berths
ships to wait (at anchorage) until ships that arrived needed in a sea port that will minimize the total port
previously had been serviced. This policy would be usage costs.
78 J. Soil. Sci. Envirn. Manage.

tools for analyzing the behavior of waiting units (ships in

1
this case), for investigating the components of a multiple
operation system (Branislav and Nam, 2006). Thus,

Ships served at a berth

queuing theory may be adequate for studying ship
2
Ships waiting for a berth movement in sea ports.
Two basic elements are necessary for the application
nw
of queuing theory to a waiting line problem: an arrival
1 2 3 3
function and a service function. These functions should
first be modeled. Once the validity of these models is
tested, the different characteristics of the theoretical
S = number of berths
models, which describe the actual system with the accu-
nw= number of waiting ships
racy that may be realized in estimating future traffic, can
S then be determined.
To analyze the movement of ships in a sea port using
the queuing theory, the following conditions are assumed:
Figure 1. Ship queue at a seaport.
i) Ships arrivals and service times conform to the pattern
of random occurrences.
ii) Ships are processed on the “first-come first-served”
queue discipline.
iii) The queue length is unlimited, that is, if a ship arrives
and finds a long queue, it joins the waiting ships and
does not leave the port.

Modeling ship arrival

Probably the two most commonly encountered arrival

patterns of ships in a sea port are the random and
scheduled arrivals with considerable delays. Thus, to
predict the number of ships present in a port in a certain
time period (usually a day), the arrival pattern of ships
may be approximated by a Poisson function (Tadashi,
2003). In this way, the probability Pn of the arrival of n
Figure 2. Ships arrival distribution as poisson function, hypothetical ships in the port in a given time can be expressed as in
port. Figure 2:

( λ ) n −λ
Pn = .e
ANALYSIS OF SHIPS' MOVEMENT IN A SEAPORT n!
An important parameter measuring the performance of a Where
seaport is the delays that ships experience while waiting λ = average arrival rate of ships during the given time
to be processed. Two factors affect these delays: (a) the (one day, for example),
pattern of ships arrival, and (b) the berth time e = base of the natural logarithm (e = 2.71828...),
requirement for cargo handling. λn = the average arrival rate of n ships, and;
The arrival of a cargo ship in a port is often irregular, n! = the factorial of the ship number.
and when it arrives, it may be able to move directly onto a
berth or has to wait until a berth becomes empty, if all The distribution of ships arrivals with Poisson function
berths are occupied. The berth time needed to serve a can be calculated, only if the average arrival rate during
ship is also variable, as it depends on the amount of an entire period is known. The expected frequency Fn of
cargo which the ship carries and the capacity of the n ships in port in a given time T is:
present facilities for handling and storing cargo (Gokkup,
1995). Figure 1 shows ship behavior at a seaport.
The investigation of such random occurrences requires
Fn = T . Pn
a complex and detailed analysis. The concept of
“Queuing theory-waiting line problem” can successfully T is the considered time period of the port operation
be applied. Queuing Theory is one of the most useful (often expressed on an annual basis as 365 days).
 El-Naggar 79

Figure 3. Service time distribution as Erlang Function, hypothetical port.

Modeling of service time observed data.

The duration of ships at a berth for handling cargo may

be described as an Erlang-function (Son and Kim, 2004) QUEUING PHENOMENON
which is usually used to present service times that are
more regularly spaced in time than those represented by As the nature of the problem is defined, in this paper, as
the Poisson distribution. multi-channels (berths), with exponential arrivals
There are purely theoretical curves (Erlang-functions), (Poisson), and multiple exponential services (Erlang), no
each of which is based on the assumption that the feasible mathematical solution is possible (Zoran and
service time is split into two or more operating phases Branislav, 2005). The theoretical models available in the
following one another, and that the ship does not leave literature for multi-channel systems are inflexible for other
the berth until all phases are completed. “k” is the number than expo-nential distribution of arrivals and multiple
of “Erlang Phases” of ships service time distribution at a exponential service time distribution. For investigating
berth. Each function has a negative exponential queuing situa-tions of multi-channel systems, models are
distribution. As “k” increases, the total service times accessible only for the following two cases:
become more uniform, until finally with k = all service
times are identical. In the general case the total service Case : Exponential distributions for both arrivals and
time probability P0 is given in Figure 3. service times.
Case II: Exponential arrivals and a constant service time.
n = k −1 n
kb . ( kb ).
Po = e n =0 n!
An approximate method has recently been proposed
regarding the queuing model of case II (Wen-Chih et al.,
Where 2007). The essential parameters are derived as follows:
b = Average berth service time (in days),
k = Erlang number (k = 1, 2, 3, …., ), and; λ = Average arrival rate in ships/day (Poisson-
n = Counter. distribution),
µ = Average service rate in ships/day (Erlang-distribution)
Po = e − b . = 1/ average berth service time = 1/ b, and;
for k = 1,
Po = e −2 b.(1+ 2b ) . S = Number of berths.
for k = 2,
2 The ratio of the arrival rate to the service rate is usually
Po = e −3b.(1+3b+9 b / 2)
. known as the traffic intensity, thus:
for k = 3,

Through the choice of k, a service time function may be λ

σ =
described as anything from the purely random µ
exponential type (k = 1) to the completely regular
constant service time type (k = ), the value of k should In this case, it can be noted that the average waiting time
be selected and tested to provide the best fit to the before service wk is given by (Erlang function) (Wen-Chih
80 J. Soil. Sci. Envirn. Manage.

et al., 2007): = 8760 h), and R = average rate of cargo transfer

between ship and berth (in tons per hour). Then,
=
= =

Where,
w1 is the correction of the average constant service time Thus, the gross berth time available is “Smin. T”.
obtained by selecting an Erlang-function with constant k
number. w1 can be calculated from the following function, Then, let β (berth utilization) equal the % of berth usage
(Wen-Chih et al., 2007): throughout the period T.
β = (berth time required)/ (berth time available), or
λS Q
w1 = n =
S −1 1 λ S min .( R.T ) .
( S − 1)! µ ( S −1) .( S .µ − λ2 ) . + λS .( S .µ − λ )
n =1
n! µ
In this manner, the calculated number of berths is based
on average values; regardless of the random arrivals of
Where, n = actual number of ships present in a port in a ships and the variation in berth service times.
certain time period.
Thus the average time that a ship spends in seaport ts
can be determined as follows: Optimum capacity

t s = b + wk If the number of berths in a port is S, the total cost spent

in the port during a certain period, C equals the sum of
From the above analysis of delays in the queue, two different types of costs: cost related to berths and
computation can readily be made of the average length of cost related to ships present (Jan and Robert, 2002).
queue, that is,for average number of ships waiting for a Thus, it can be expressed as (Figure 4):
berth nw, the appropriate expression is:
C = Cb. T .S + C sη s
η =λ
In which,
The average number of ships ns present in port with S C = total cost of a port with S berths during the period T,
berths in a certain time period can be determined using usually one year = 365 days, (in L.E.),
the following formula: Cb = average cost of a berth; that is, construction and
maintenance costs (L.E./day/berth),
η =η +η cs = average delay cost of a waiting ship (L.E./day/ship),
and;
Where, ns = average number of ships present in port.
nb = average number of n ships served at S berths
= S × berth utilization factor Accordingly, if the amount of cargo that must be dealt
with at a port during the period T is given as time
= S. (λ/(µ.S) = σ. planning target, then such number of berths S becomes
the optimum that minimizes the total cost C. Therefore, C
Thus, it is seen that the traffic intensity, σ defined in the is a proper measure to examine the optimality of a port
queuing theory equals the average number of ships system.
served at berths nb. Now, both sides of the above equation are divided by
“Cs.t” in order to decrease the number of the parameters
involved. Thus,
ANALYSIS OF PORT CAPACITY
Rs = C / (CS.T) = ) (cb/cS).S + ηs = (rbs. S) + ηs
Minimum capacity
In which,
The minimum number of berths Smin needed in a seaport rs = ratio of the total annual cost for port to annual ship
to handle a certain amount of cargo can be calculated cost, and
using the following procedure: rbs = berth-ship cost ratio.
Assuming that S is optimum, then the following
Let Q = the total amount of cargo (in tons) handled in a optimization condition must be held:
port section in a time period T (for example, T = one year rs < rs + 1, and rs < rs-1
 El-Naggar 81

handled in the port in that year was 4.326 million tons (Egyptian
Maritime Data Bank, 2008).
Alexandria port is constituted of an old and complicated layout
with short quays and too narrow or too long piers. A large number
of quays has limited drought less than 8.0 meters, and only a lower
number of berths is capable to receive ships with more than 130.00
meter length. The number of berths available for general cargo in
the port is 32 berths.

Data base

The daily “log books” of the traffic department of the Alexandria Port
Authority include (among others) the arrival time of each ship at the
pilot vessel. In addition, detailed information concerning the move-
ment of each ship in the port is also available in the so-called “ship
log sheets”. Every sheet is a ship report, and it contains the
Figure 4. Total usage cost, hypothetical port. following data:

a) Ship name, nationality, type of cargo, and total tonnage.

b) Berth occupancy, including berth changes during the period in
Thus, rs will be adopted hereafter as a measure to port.
determine the optimum number of berths. c) Date and time of arrival, berthing, and quitting the port.
From the preceding information the procedure can be
standardized as follows when given the data Q, R, cb, cs, Ships arrivals
λ, µ, k:
If the distribution of ships arrivals can be predicted reliably, port
Step 1. Calculate the minimum number of berths from the planner can proceed with great confidence in making development
plans that may avoid over-building or under-building the port
Q facilities.
S min = The actual pattern of ship arrivals at the port of Alexandria is
equation,
R.T . compared with the theoretical function prognosticated
Step 2. Determine the value of traffic intensity σ as; mathematically by Poisson distribution of random occurrences. The
application includes a specific analysis of the number of ships
λ
(σ = ). present, day by day, over a period of one year (from July 1, 2007 to
µ June 30, 2008).
The number of ships present in the port, each day, was trans-
Step 3. Compute the value of berth-ship cost ratio rbs cribed from the port “log books” and then summarized to obtain the
from the given data cb and cs. number of days, that various number of ships were present during
Step 4. For each number of berths, with S greater than the period studied. The theoretical distribution, Poisson is
the minimum value, estimate the number of ships present computed.
Table 1 compares the predicted distribution with the actual one.
in port ns, and predict the ratio rs. The average arrival rate was 5.68 ships per day. Table 1 shows a
Step 5. The number of berths which satisfies the good agreement between actual and predicted distributions. The
optimization condition (rs < rs+1, and rs < rs-1) is optimum. number of days that various numbers of ships are predicted to be
Step 6. Compute the average berth utilization, β: present in the port is in agreement with the actual distribution on
σ 336 days of 360 days, that is, on 92% of days.
β = . To judge whether the observed frequencies of ship arrival
distribution is compatible with the predicted theoretical frequencies,
Step 7. Summarize the queuing results (average number Chi-square is computed, and the result, x2 = 20.0 with 10%
of ships present in the port, average number of ships at probability, indicates a good fit. From the statistical standpoint, pro-
berths, average number of waiting ships, average waiting bability values between 5 and 95% designate good fit from which it
is concluded that this theoretical distribution is plausible (Tadashi,
time). 2003). Figure 5 demonstrates the goodness of fit between actual
and predicted distributions.

APPLICATION OF THE PROPOSED METHODOLOGY TO

ALEXANDRIA SEAPORT Berth service times

The foregoing methodology is applied to investigate the movements Information giving the date and time of arrival at a berth and the
of ships in Alexandria Port and to predict the future capacity. The date and time of departure from the berth were obtained from the
application is restricted to general cargo ships, excluding full- “ship log sheets”. A total of 315 observations, including those
container, bulk, and RO/RO ships which have particular berths at general cargo ships which were tied up at the berths between July
the port. 1, 2007 and June 30, 2008 were randomly selected to be analyzed.
Alexandria Port is the major port in Egypt. About 40.80 million A class interval of 15 hours was selected for such analysis.
tons passed through the port in the year 2007/2008, that is, 36% of Search for a suitable model for the distribution of the durations at
the total volume of the foreign trade. The amount of general cargo berths led to an Erlang distribution giving K = 3. The mean time
82 J. Soil. Sci. Envirn. Manage.

Table 1. Comparison of actual versus predicted ship arrival distribution.

Arrival rate Actual number of days Predicted number Minimum (A) or

(Ships/day) (A) of days (B) (B)
0 1 1 1
1 6 7 6
2 18 20 18
3 27 37 27
4 49 54 49
5 73 62 62
6 60 59 59
7 61 47 47
8 37 34 34
9 15 21 15
10 9 12 9
11 5 6 5
12 2 3 2
13 2 2 2
Total 365 365 336

spent at a berth was found 5.58 days for the 315 observations. The
standard deviation of the distribution was computed and found to be
± 1.43 days. Figure 6 presents the frequency and the cumulative
distributions of the observed data and compares the values of the
cumulative distribution with those of the Erlang function having K =
3.
A Chi-square test was also performed to test the goodness of fit
between the observed frequency distribution and the postulated
Erlang function, and a value X2 = 14.87 for 42% probability was
found. Comparison with other Erlang functions (K = 1, K = 2, and K
= 4) indicates that K = 3 is the best choice for this distribution
function. Figure 6 also shows the observed data points and a plot
of the selected function.

Optimum number of berths

Figure 5. Frequency Distribution of Ships Arrivals, Alexandria Port
2007/2008 To establish the optimum number of berths needed for general
cargo handling at Alexandria port in the year 2017, applying the
proposed procedure, the following input data are used:

i) Due to the further development of the Egyptian ports, particularly

the Dekheila port, the annual general cargo tonnage to be handled
at the berths of Alexandria port will be only about 4.00 million tons
at the target year (tonnage in year 2007/2008 = 4.326 million tons)
(Egyptian Maritime Data Bank, 2008).
ii) The average arrival rate of general cargo ships will be 5.68 ships
/day, assuming that the average ships load equals 2084 tons (the
present value).
iii) The average rate of cargo handling at a general cargo berth R =
373.5 tons per day (the existing rate).
iv) The average cost of a berth cb = 2000 per day (approximately
$600 per day), based on the development program of the
Alexandria port (Egyptian Maritime Data Bank, 2008).
v) The average delay cost of a general cargo ship cs = $ 6000 per
day.

The calculations are carried out as follows:

Figure 6. Frequency distribution of berth service time, Smin = 4000 000/(373.5 × 365) = 29.34 = 30 berths
Alexandria port 2007/2008. λ = 5.68 ships/day
 El-Naggar 83

Figure 7. Determination of optimum number of berths, , Alexandria port, case study.

Figure 8. Determination of Optimum Number of Berths, due to different cost

ratio, rbs, Alexandria Port, Case Study.

µ = 1/5.58 = 0.18 ships/day changing the traffic intensity and/or the cost ratio rbs.
= 5.68/0.18 = 29.35 The optimum number of berths corresponding to a suitable cost
rbs = 600/6000 = 0.10 ratio rbs values (from 0.10 to 0.30) is noted in Figure 8. It can also
be seen that a 33-berths set is the optimum port capacity in case of
Figure 7 shows the relationships between traffic intensity and the traffic intensity values varying between 27.58 and 29.60.
cost ratio rs for a proper number of berths (from S = 29 to S = 34) Table 2 shows the calculation of the costs of idle berths and idle
The optimum port capacity is 33 berths. In this instance, rs = 34.34, ships for 33 berths in view of the expected frequency (number of
and the total port costs C = $75.200 million (the development of 33 days per year). It also presents the combined costs (vacant berths
berths plus the annual maintenance costs). and ships) in case of port size 31, 32, 33 and 34 berths. The cost
comparison indicates that the total port cost is least when there are
The average berth utilization = 29.35/33 = 0.89 33 berths. This conclusion confirms the result previously obtained
by applying the proposed methodology.
Queuing results:

Average number of ships present in port ns = 31.04 RESULTS CONCLUSIONS

Average number of ships served at berths nb = 2935
Average number of waiting ships nw = 1.69 This paper presents a methodology proposed to predict
Average waiting time per ship wk = 0.32 days
the optimum number of berths required in a sea port to
The relationships in Figure 7 are prepared as design curves derived meet the future traffic volumes. The methodology is
to determine the optimum number of berths for Alexandria port by based on the hypothesis that the number of berths can
84 J. Soil. Sci. Envirn. Manage.

Table 2. Cost calculation in case of 33 berths, and the comparison of the resulting value with those for 31, 32 and 34 berths.

Arrival Required Over-building Under-building

Berth
rate Predicted frequency (in days) number of Number Berths- Number
utilization Ships-days
(ships/day) berths of berths days of ships
* *
µ F =λ÷µ 33 – F . (33 – ) λ-X λ-X)
F . (λ
0 1 0.00 0 33 33
1 7 0.17 6 27 189
2 20 0.34 12 21 420
3 37 0.51 17 16 592
4 54 0.68 23 10 540
5 62 0.85 28 5 310
6 59 1.00 33 0 0 0 0
7 47 1 47
8 34 2 68
9 21 3 63
10 12 4 48
11 6 5 30
12 3 6 18
13 0 7 0
Total 365 2086 274
Cost in Million $ (using cb = 600, cs = 6000) 1.25 1.64
Total costs for 33 berths in Million dollars 2.89
Total costs for 31 berths in Million dollars 3.15
Total costs for 32 berths in Million dollars 3.08
Total costs for 34 berths in Million dollars 2.93
X* = Number of available berths × maximum berth utilization / average service time = 33 × 1/5.58 = 6.00

be increased as long as the marginal cost of berths (con- Thus, there is no doubt that ships arrive at Alexandria
struction and maintenance) is less than the delay costs of port in accordance with a random pattern and that the
waiting ships. degree of accuracy compares favorably with the accuracy
The “Queuing theory” has been employed to derive the that may be realized in estimating future traffic.
number of waiting ships and the average ship delays. The application to Alexandria port verifies the antici-
The usage of queuing theory is subjected to the following pated benefit of using the suggested methodology to
two assumptions: evaluate the port size in the best interests of both ship
operators and the port authority. The evaluation is settled
i) Ships arrivals at a sea port can be described as a on the premise that maximum port efficiency results when
negative exponential distribution, and, the total port cost is minimum, that is, the cost of vacant
ii) Berth service time yields to a multi-exponential berths over a substantial period plus the time cost of
function. ships waiting for a berth during the same period.

The employment of the queuing theory to study the

movements of general cargo ships at Alexandria port was REFERENCES
profitable. The observed pattern of ships arrivals appears
Gokkup U (1995). Application of queuing theory on the design of fishing
to agree with Poisson’s law of random distribution. In Harbors, Transactions on the Built Environment. 11: 711-719.
addition, the berth service time for 315 ships was found Branislav D, Nam KP (2006). Modeling of ship-Berth-Yard Link
to conform most closely to an Erlang distribution with K = Performance and Throuput Optimizatin, IAME 2006 Conference,
3. The usage of an approximate model of queuing theory Melbourne, Australia,
Tadashi Y (2003). Optimizing the handling Capacity in a Container
led to acceptable results. The criterion for acceptance of Terminal for Investigating efficient Handling Systems, J. Eastern Asia
this model was the reasonable agreement achieved Society for Transportation Studies, 5: 597–608.
between the computed and observed values of average Son JH, Kim MH (2004). An analysis of the optimal number of servers in
waiting time and average number of waiting ships in distributed client/servers, Environ. Decision Support Syst., 36: 297-
312.
queues at berths. Zoran R, Branislav D (2005). Optimal Number and Capacity of Servers
in Queuing Systems, Inf. Manage. Sci., 16(3): 1-16.
Wen-Chih H, Tu-Cheng K, Sheng-Chieh W (2007). A Comparison of Jan OJ, Robert E (2002). Unification of Accounts and Marginal Costs for
Analytical Methods and Simulation for Container Terminal Planning, Transport Efficiency Swedish Seaport Case Study: Price-relevant
200 J. Chinese Institute of Indus. Engineers, 24(3): 200-209. marginal cost of Swedish seaport services, ITS, University of Leeds,
El-Naggar 85

Funded by the European Commission, 5th Framework – Transport

RTD.
Egyptian Maritime Data Bank (2008). Annual Statistical Report, National
Planning Institute, Cairo.