SIMULATION OF A CONTAINER TERMINAL

THROUGH A DISCRETE EVENT APPROACH:
LITERATURE REVIEW AND GUIDELINES FOR APPLICATION
Armando Cartenì, Stefano de Luca
Department of Civil Engineering – University of Salerno

1

INTRODUCTION

Design and project appraisal of container terminals may be carried out
through two main approaches: optimization or simulation. Although the
approaches based on optimization models allow a more elegant and compact
formulation of the problem, simulation models are mainly based on discrete
event simulation (DES) models and help to achieve several aims: overcome
mathematical limitations of optimization approaches, support and make
computer-generated strategies/policies more understandable, and support
decision makers in daily decision processes through a “what if” approach.
Several applications of DES models have been proposed and simulation
results confirm that such an approach is quite effective at simulating container
terminal operations. Most of the contributions in the literature develop objectoriented simulation models and pursue a macroscopic approach which
gathers elementary handling activities (e.g. using cranes, reach stackers,
shuttles) into a few macro-activities (e.g. unloading vessels: crane-dock-reach
stacker-shuttle-yard), simulate the movement of an “aggregation” of
containers and therefore do not take into account the effects of container
types (e.g. 20’ vs 40’, full vs empty), the incidence of different handling
activities that may seem similar but show different time duration and
variability/dispersion (e.g. crane unloading a container to dock or to a shuttle)
and the differences within the same handling activity (e.g.
stacking/loading/unloading time with respect to the tier number). Such
contributions primarily focus on modelling architecture, on software
implementation issues and on simulating design/real scenarios. Activity
duration is often assumed to be deterministic, and those few authors that
estimate specific stochastic handling equipment models do not clearly state
how they were calibrated, what data were used and what the parameter
values are. Finally, no one investigates the effects of different modelling
hypotheses on the simulation of container terminal performances.
The focus of this paper is on the effects that different hypotheses on handling
equipment models calibration may have on the simulation (discrete event) of
container terminal performances. Such effects could not be negligible and
should be investigated with respect to different planning horizons, such as
strategic or tactical. The aim is to propose to analysts, modellers and
practitioners a sort of a guideline useful to point out the strengths or
weaknesses of different approaches.
Drawing on the model architecture proposed in a previous contribution by the
same authors (Cartenì, Cantarella and de Luca, 2005), a discrete event
simulation model is developed and applied to the Salerno Container Terminal
in order to deal with the following issues:

© Association for European Transport and contributors 2009

1

• analysis of the effects of different estimation approaches (sample mean
and random variable estimations) on estimating whole terminal
performance, hence on container terminal planning strategies. In
particular, analyses were made for different time horizons: long-term
planning interventions/investments, medium/short period, short-term or
real-time applications.
• Analysis of the effects of different hypotheses on the level of aggregation
of elementary activities (undifferentiated vs. container type model).
The paper is divided into four sections. In the first section (section 2) an in
depth literature survey is proposed. The aim is to go back over about thirty
year of container terminal simulation models, to highlight weaknesses points
of the existing approaches to handling equipment activities simulation, and to
propose a synthetic but complete outline of the models calibrated and of their
parameters. In section 3 a brief description of the discrete event simulation
model is reported. In section 4 results from model application are proposed
while the main conclusions are drawn in section 5.

2

Literature review

The existing literature reports approaches to either managing a container
terminal as a system and trying to simulate all elements or managing a subset of activities (simultaneously or sequentially following a predefined
hierarchy). The main contributions seek to maximize overall terminal efficiency
or the efficiency of a specific sub-area (or activity) inside the terminal. The
most widely followed approaches are based on deterministic optimization
methods, although recently a stochastic optimization model was proposed
(Murty et al., 2005). Such approaches schematize container terminal activities
through single queue models or through a network of queues. Following a
stochastic approach, both modelling solutions may lead to analytical problems
and/or unsatisfactory results if the probability distribution of activities involved
does not belong to the Erlang family (Nilse and Abdus-Samad, 1977; Ramani,
1996). Moreover, the resulting network could be very complicated and
theoretical solution might not be easy to obtain.
In such a context, an effective and challenging alternative approach for
container terminal system analysis may be represented by discrete simulation.
Simulation can help to achieve various aims: overcome mathematical
limitations of optimization approaches, allow a more detailed and realistic
representation of terminal characteristics, support decision makers in daily
decision processes through assessment of “what if” scenarios and make
computer-generated strategies/policies more understandable.
Simulation is not a new methodology in port operations. Several works have
been presented since the 1980s, most of them concerning port operations
management. Many of the proposed models do not focus on the details
regarding the model set-up, its calibration and its validation; but on the
application and/or the simulation of design scenarios. Moreover, although the
estimation of handling activity models should be one of the main issues of all
container terminal applications, this problem does not seem to be treated in
depth in most applications. While many contributions do not present any
information on handling activity models used, the remaining contributions
© Association for European Transport and contributors 2009

2

and Chung et al. Gambardella et al. the parameters estimated and on parameter values.carry out very simple approaches (deterministic) and/or give scant information on the estimation approach adopted. terminals were modelled more realistically through disaggregation of the main operations in several elementary activities. (1998) proposed a discrete event simulation model (based on processoriented paradigm) to simulate vessel loading/unloading. to simulate different application scenarios. El Sheikh et al. Mosca et al. In Agerschou et al. (1988) proposed a methodology based on a graphic simulation system to simulate the use of buffer space to increase the use of handling equipment and reduce total container loading time. Less attention was focused on modelling handling activities and/or model details. Kondratowicz (1990). within a general method for modelling seaport and inland terminals in intermodal freight transportation systems. Hayuth et al. The model was applied to the Italian container terminal of La Spezia (Italy). Lai and Lam (1994) examined strategies for allocation of yard equipment for a large container yard in Hong-Kong. (ii) to focus on the approaches. models and parameters used to model handling activities. with scant information on the data used and on the characteristics of the equipment used in the application. Comer and Taborga (1987) developed one of the first port simulation softwares (PORTSIM). the 1980s saw several works implementing the first simulation-based models. (1987) developed a simulation model for the ship-to-berth allocation problem. (1996) and Merkutyev et al. whereas empirical distributions are used for the remaining activities. Park and Noh (1987) used a Monte Carlo type simulation approach to plan port capacity. (1991) described a simulation program that models the transfer of containerized cargo to and from ships. and vessel interarrival and service time are modelled through exponential distribution functions. Tugcu (1983) proposed a simulation model for the port of Istanbul. The same case study was analyzed by Mastrolilli et al. In the same year. (1994) used a discrete event simulation to build a port simulator. both contributions proposing an application to the Riga Harbour Container Terminal. Starting from the pioneering work of Collier (1980) investigating the role of simulation as an aid to the study of a port as a system. © Association for European Transport and contributors 2009 3 . Key issues of the application of modelling and simulation were discussed in Tolujev et al. dealing with berth assignment and unloading operations. (1998). the phenomenon is modelled as a sequence of queues. but the main emphasis was on software and on hardware problems. and Hassan (1993) gave an overview of a computer simulation program used as a decision support tool to evaluate and improve port activities. The aim of our analysis is twofold: (i) to propose an extensive review of the main contributions in the literature. and much more attention was laid on real case studies. proposed an object-oriented model. the experimental data used. Silberholz et al. In the 1990s much effort was spent on simulating terminal containers: the number of applications based on simulation increased. (1992) used simulation to ascertain the efficiency of an automatic flatar system servicing a rail-mounted crane. on software issues and/or on model validation. (1983). TRANSNODE. Vessel arrival is simulated through Poisson distribution. The focus of most contributions was on developing practical tools to simulate terminal operations. In the same year.

(1998). whereas Weibull distribution seems to fit crane cycle time better. The main emphasis was on optimizing container placement in a terminal. yards and berths). HolguìnVera and Walton (1996) and Ramani (1996). a genetic algorithm approach was adopted. a seaport simulation model able to animate and visualize seaport processes and in the same year Bruzzone and Signorile (1998) developed a software tool to support terminal operators in making strategic decisions. Ramani (1996) designed and developed an interactive computer simulation model to support the logistics planning of container operations. yard crane and crane movements are simulated through a random variable made up by systematic and a random component. Since the end of the 1990s. Means and standard deviations are estimated for quay crane. yard crane and straddle carrier service time. Koh et al. As regards equipment characteristics. exponential for interarrival time of trailers. average values are used for handling equipment. the most important ports in the world have been modelled through discrete event simulation models. as well as vessel arrival and truck arrival. a simple application proposed. deterministic and stochastic distribution functions are considered: deterministic for trailer speed and for interarrival time of trailers and tractors. whereas distribution functions are used for crane operation time (Normal distribution). The proposed model relies on experimental data. The model is calibrated on experimental data and two approaches are carried out: a deterministic one based on empirical distribution and a stochastic one. The system is analyzed as a whole (gates. The same authors (Bruzzone et al. The model provides estimates for port performance indicators. it is based on Visual C++ and gives accurate results once validated on historical data. Interesting modelling details were proposed by Koh et al. (1994) developed an object-oriented approach using MODSIM simulation software. The same case study proposed by Choi and Yun (2000) follows an objectoriented approach. developing a model to simulate two different terminals located in Pusan. (1994). Yun and Choi (1999) develop an object-oriented simulation model using SIMPLE++ language and apply it to analyse the container terminal system used in Pusan. whereas speed of cranes and travel time of shuttle trailers are assumed deterministic. averages are used for cranes and trailer speed. using a model similar to that proposed in Gambardella et al. and greater interest is shown in the calibration of handling activities models. Their discrete event simulation model is based on data collected at the port of Kelang and specific analyses are carried out to identify the distribution functions for inter-arrival time of ships (Weibull distribution) and for © Association for European Transport and contributors 2009 4 . (1998) and proposing a calibration and a validation procedure of simulator parameters. vessels and service time of cranes. uniform for service time at the gates. It is not clear whether performance characteristics were estimated. the corresponding random parts are not clearly introduced. Tahar and Hussain (2000) deal with berth operation and crane allocation problems. 1999) investigated the effectiveness and benefits of a simulation approach as a decision support system for complex container terminals. yet no details can be found on the performance functions used. Nevins et al. Holguìn-Vera and Walton proposed a simulation model based on the next event approach. Gantry crane. While the systematic components are estimated using multiple regression. The simulation tool is generic and transferable to any other terminal. (1998) developed PORTSIM.

whereas maximum. The model is implemented in an EXTEND software package and applied to the port of Thessaloniki. (2002) use a port simulator developed in TAYLOR II software to investigate the effectiveness of two different operational systems applied to the terminal of Melbourne. Chin et al. handling activities are hypothesised as deterministic. finally. Kia et al. It is not clear whether the probability distributions were estimated or simply taken from the literature. A simulation tool is created in PASCAL programming © Association for European Transport and contributors 2009 5 . (2006) develop a simulation tool in JAVA programming language to simulate the port of Casablanca. Vessel interarrival is represented by an exponential distribution function (estimated). a higher order is adopted for more regular services and. interesting statistical analyses are presented on vessel arrival patterns (exponential distribution for interarrival times) and on vessel service time (k-stage Erlang distribution). the focus is on vessel arrivals and their distribution among the existing buffers and operators.service time at berths (distribution not mentioned). Shabayek and Yeung (2002) propose a discrete event simulation model employing the Witness program to analyse the performance of Hong Kong’s Kwai Chung container. Although the model encompasses all the operations that may occur in a terminal. a triangular distribution is used to assign the number of containers to cranes. Truck inter-arrival times follow an Erlang distribution. They stress the relevance of a stochastic approach and schematize the system as a three-phase queuing system with different numbers of servers in each phase. Bugaric and Petrovic (2007) simulate unloading services of bulk cargo vessels. Since no detailed disaggregate data are available. minimum or most probable values are estimated for speed and activity time of equipment involved. The focus is on the application and no details are given either on the models or data used. all the activities that occur inside the terminal are not explicitly simulated but aggregated in one variable represented by the vessel’s service time. While arrivals are simulated through a distribution function (k-stage Erlang). Although no details are reported on the model structure. The focus is on the architecture and on software issues. Legato and Mazza (2001) examine the vessel arrival-departure process. a first order Erlang distribution is applied for those services with the supposed larger variance. Developing a microscopic simulation model. (2002) evaluate the effectiveness of automated guidance vehicles. With the emphasis on terminal capacity. vessel loading/unloading and gate operations. three possible destinations and connections between them (by road and/or by rail). Parola and Sciomachen (2005) present a discrete event model to simulate the logistic chain of a system made by two ports. Sgouridis and Angelides (2002) develop a discrete event model to simulate the inbound container handling problem. developing a queuing network model through an object-oriented approach implemented in VISUAL SLAM language. the remaining operations are analysed in a very aggregate way and average values are considered (average handling capacity). Bielli et al. crane working time and truck waiting time by a truncated normal distribution. The model is implemented in ARENA software and is validated on historical data. The simulation is undertaken through WITNESS simulation software and the main emphasis is on vessel berthing.

(2000) propose the uniform distribution function for quay crane and a triangular distribution function for yard gantry crane. yard). Koh et al. Choi (2000). Kia et al. Deterministic functions appear to have been used for gantry cranes. El Sheikh (1987). (2006). exponential for the transfer time in dock assignment while for vessel arrival time an empirical distribution function is used. a stochastic approach is unanimously proposed. Yun and Choi (1999). Merkuryeva et al. In particular. Lee and Cho (2007). lower and locate. (2000). Of the contributions introduced so far. hoist and traverse. AutoMod 11. (2007) set out to simulate the whole freight transport process in the Guadalquivir river estuary. KMI (2000) and Bielli et al. and all variables are generated using the Monte-Carlo method according to distribution functions obtained from an existing river terminal: normal for anchorage operations and for crane unloading times. Tugcu (1983). for a bulk cargo terminal. Bugavic and Petrovic (2007). Parola and Sciomachen (2005). Half of them adopt a stochastic approach and show estimated parameter values. Merkuryeva et al. (2008) report the estimated mean values. Choi (2000). With respect to crane speed. There is substantial heterogeneity regarding the level of aggregation of activities involved and how such activities are aggregated in a single macro-activity: El Sheikh (1987). little information on equipment characteristics and time duration is reported. Bielli et al. (2002) and Shabayek and Yeung (2002) analyse the entire time to load (unload) a vessel (vessel cycle time). Koh et al. Lee and Cho (2007) suggest the exponential distribution function for quay crane and a triangular distribution function for yard gantry crane operation time. As regards crane cycle time. only ten papers give information on the handling equipment models used. as already pointed out. all propose deterministic and aggregate models while only Yun and Choi (1999). Kia et al. KMI (2000). Thiers (1998). Yun and Choi (1999) propose the exponential distribution function both for quay crane and yard crane. (2006) follow a deterministic approach. © Association for European Transport and contributors 2009 6 . Merkuryeva et al. crane loading time to/from a vessel is analysed by Tugcu (1983). Most of the contributions deal with vessel loading/unloading operations. and Lee and Cho (2007).1 software is used and statistical models are proposed. Cortès et al. unlock and return). contrasting with the stochastic approach adopted by Yun and Choi (1999). propose normal random variables and report the estimated parameters. (2000). As regards vessel cycle time. Thiers (1998). Lee and Cho (2007) propose a model to simulate the effectiveness of a dynamic planning system for yard tractors utilizing real-time location systems technology.language. With regard to crane loading/unloading time. Parola and Sciomachen (2005) estimated a normal random variable but do not report parameter values. KMI (2000) and Legato et al. (1994) advise the use of a Weibull random variable. exponential for inter-arrival of vessels. (2002) and Shabayek and Yeung (2002) suggest using Erlang random variables whereas Choi (2000) proposes normal random variables for two crane types (quay. Despite a detailed description of operations and the software modules implemented. (1994) and Bugavic and Petrovic (2007) investigate the crane cycle time (time needed to: lock onto the container. Parola and Sciomachen (2005).

= 0. probability distribution and corresponding parameters are reported. 2 and 3.d.50 (min) normal not reported deterministic deterministic deterministic not reported not reported not reported Lee and Cho (2007) Bielli et al.) 45 (metres/min.d. (2006) Parola and Schiomachen (2005) Tugcu (1983) Thiers (1998) KMI (2000) - crane cycle time vessel cycle time - Koh et al.50 (min) mean = 1. = 0.77 mean = 37.33 mean = 7.57 (day) K = 10.6. = 5. the few models existing are hard to transfer to different case studies (due to the influence of path length. (2008) KMI (2000) deterministic deterministic deterministic KMI (2000) deterministic Choi (2000) deterministic Tugcu (1983) Koh et al. path winding. (2000) propose a triangular distribution function for the forklift.00 (min) s. Hence they are omitted in this survey.3] (hour) K = 117 45 (metres/min.00 (min) min. For each type of handling equipment and for each activity simulated.00 mean ∈ [9.08 (min) mean = 1. whereas Merkuryeva et al. A synopsis of the above analysis is presented in tables 1.26 (min) mean = 112. not much can be found in the literature: Sgouridis and Angelides (2002) use deterministic values for a straddle carrier.) mean = 1.60 (min) mean = 87.) 130 (metres/min. waiting time …). travel time. = 0. (1994) Thiers (1998) deterministic deterministic deterministic © Association for European Transport and contributors 2009 mean = 5.d. (2006) Yun and Choi (1999) deterministic exponential Merkuryeva et al.) 75 (metres/min.55 (min) s.) 45 (metres/min. Table 1 – Survey of handling models: Gantry crane (GC) Quay GC Yard GC crane operation time Yun and Choi (1999) Lee and Cho (2007) exponential Merkuryeva et al.85 (hour) K = 4. = 13.=2.d. Choi (1999) Legato et al.80 (min) s. traffic vehicle congestion inside the terminal and so on). (1994) weibull bulk cargo Bugavic and Petrovic (2007) normal Quay GC Choi (2000) normal Yard GC Choi (2000) normal Entire loading operation El Sheikh (1987) Erlang Entire unloading operation El Sheikh (1987) Erlang - Kia et al.00 (min. 134 (metres/min.50 (min) mean = 1.20 (day) K = 4.) 7 .89 (min) mean = 4.) 180 (metres/min. (2000) triangular deterministic mean = 1.00 (min) s. As regards shuttle performances (speed.00 (min.With respect to other handling equipment. (2000) uniform Bielli et al. (2002) Erlang - Shabayek and Yeung (2002) Erlang quay gantry crane crane speed hoist with full load hoist without load ship trolley store trolley yard gantry crane - mean = 0.) 55 (metres/min.d.41 (min) Yun.00 (min) 40’ loading : mean = 6.00 (min) 40’ unloading : mean = 4. 16.) max=4.00 (min) s.

) st.00 (min. = 0.00 (min. gates.41 (min.Table 2 – Survey of handling models: Straddle carrier (SC) Handling activity speed shuttle loading/unloading time spreader movement turning container spotting Model used deterministic Characteristic parameters inside yard: 110(met.30 (min. © Association for European Transport and contributors 2009 8 . variables define the state (dynamic characteristics) of each entity and may change over time and can further be classified as deterministic or stochastic. dev. Such entities do not usually move containers but can control/manage entities that handle containers and can thus change their attributes.) 3 MODEL The proposed approach schematizes a container terminal (CT) as a discrete event system and models its functioning through a simulator.) 0. they may be characterized by parameters and variables. Apart from the above-described entities other entities can be considered.) References Shouridis and Angelides (2002) Table 3 – Survey of handling models: Forklift (FL) Handling activity loading/unloading time Model used Characteristic parameters 20’ loading mean = 4.). − Handling equipment is a resident and active entity and may be characterized by parameters. dev.60 (min. use one more reach stacker) or by sub-models that change entity attributes. Resident entities remain part of the system for long intervals of time.00 (min./min. − Physical locations are resident and passive entities. − Containers are transient and passive entities. etc. As for containers. A discrete event system can be defined as an interacting set of entities/objects that evolves through different states as internal or external events happen. variables and an activity. The change in such attributes may be driven by simple heuristic rules (e. [b] Demand characterization and estimation. [c] Supply characterization and calibration./min. yard.) st.) outside yard:250 (met. Parameters define static (stationary) characteristics that never change. Entities/objects may be physical. trying to optimize overall terminal performance in real time.) deterministic 0. transient entities enter into and depart from the system several times. if there are more than four trucks waiting for a reach stacker. the containers and all those physical locations relevant to CT operations (dock. Entities can be characterized by parameters and/or variables. = 0. and can be resident or transient.g. (2000) 20’ unloading mean = 3.) deterministic deterministic deterministic 0. In discrete event modelling the model is defined once the case study is defined and three main tasks should be carried out. In a CT entities represent the handling equipment.41 (min.02 (min. conceptual (information flows) or mathematical..) 1. [a] Identification of the terminal’s logical and functional architecture.) triangular References Merkuryeva et al.

Ac.m.t. c.t. B-RS c. B-S EMPTY YARD Truck Br. FL-T c.m. Ac (full) Customs YES c.t.7 GATE OUT YES c. B-RS Shuttle c.m.t. RS-C c. The architecture is proposed in Figure 1.m.m. loading (yard – berth) and unloading (berth – yard).14 UNLOAD Shuttle c.m.m. The basic activities occur simultaneously and interactively. C-R Wait Stor.m. C-V YES Wait Buffer NO NO Shuttle c. YES YES Road Network c. RS-S YES YARD BERTH Vessel Figure 1 – Model architecture ID. B-C c. V-C c.t.t.m. S-C LOAD c.13 Rail Br. c. Road Network GATE c. T-RS c.16 c.t.t. which amounts to 45 kTEUs/ha.12 c.t.9 NO YES Wait Rail Destination Truck Br.t. FL-T Wait c. RS-S c. all the typical activities of a container terminal were explicitly simulated. C-V c.t. trailers and reach stackers. Ac. C-T c.t. C-B c. from Shuttle to Fork Lift from Forklift to Truck from Crane to Rail from Crane to Truck j container movement (j∈[1…16]) bureaucratic activities storage container waiting time customer activities rail bureaucratic activities truck bureaucratic activities © Association for European Transport and contributors 2009 9 . RS-B c. loading/unloading cranes. Cartenì and de Luca.t. B-S ACTIVITY DESCRIPTION from Truck to Reach Stacker from Reach Stacker to Shuttle from Shuttle to Crane from Reach Stacker to Berth from Reach Stacker to Crane from Berth to Crane from Crane to Vessel from Vessel to Crane from Crane to Berth from Berth to Reach Stacker from Berth to Shuttle ID.t.1 Stor. more details may be found in (Cartenì et al. Ac (empty) NO Load NO YES BUFFER Stor. Wait c.15 c. Apart from vessel arrival and berthing (not relevant to our case study) and apart from truck arrival.45 MTEUs per year in less than 10ha (100. T-RS c.11 Stor.m.t.4 Vessel c.t.Case study In this paper the Salerno Container Terminal (SalCT) is analyzed.m. ACTIVITY DESCRIPTION c.t. Stor.000 m2). Truck Br.t.t. Wait Cust. RS-B c. RS-S Consolidation c.t.t. C-T c.t. Ac.m.5 c.8 c. S-FL c. c. c. and is both small and very efficient: it handles close to 0.Ac. S-FL Stor. SalCT is a major private container terminal operator in southern Italy.m.t.t. container yards. Model architecture Three different macro-activities were taken into account: import. The Salerno Container Terminal (SalCT) can be divided into three subsystems: enter/exit port gates (land-side). V-C NO NO Empty Tranship.0 EXPORT YARD c. RS c. RS-C c. RS-S c. B-C c. yard tractors. B-RS YES NO NO c.t.10 IMPORT YARD c.t. S-C c. C-B c. Container handling equipment comprises storage cranes. Rail Br.m. 2005.t. and berths (sea-side).m.m.m.t.t. Ac. C-R c.t. S-C Cust.6 c. Terminal container Rail Network c. Ac.t. Ac c. and can be grouped into four main operations: receiving (gate – yard).t.j Br.m..t.3 Firm c. delivery (yard – gate). 2009).2 GATE IN Br. export and transhipment.

© Association for European Transport and contributors 2009 10 . results of estimation (sample means and probability function parameters) are reported for each handling equipment and for each activity. container flows were subdivided by origin and destination zone and were arranged in origindestination matrices. reported in table 6. the demand flows were characterized over space. For each macro-operation (import. export. In table 7 means and standard deviations related to Gamma distribution are reported. In such a classification the different entities involved must be characterized by their geometrical characteristics (if physical points) and by the corresponding performance supplied (time duration and/or transport capacity). concern loading activities from shuttle to vessel or from dock to vessel. Storage capacity was estimated for quays and yards. …). The same random variable seems to be the best approximation for loading and unloading activities that involve 20’ and 40’ (full or empty) containers. 2009). 2009). gates. the analysis is mainly focused on full containers. Different O-D matrices were estimated for each container type (20 feet vs. reach stacker (RS).. In the following tables. operations are set up by elementary handling activities. transhipment). and unloading activities from vessel to dock. Since most Salerno Container Terminal loading/unloading activities concern full containers. Details on the pursued estimation methodologies and/or comments on estimation and calibration results may be found in Cartenì and de Luca. gantry crane (GC). 40 feet. The results. for each operation macro-origin and macrodestination zones were identified. (details in Cartenì et al.Demand characterization Demand is represented by single containers. in a container terminal macrooperations. 2005. usually corresponding to quays. averages and probability distribution functions were estimated for handling equipments time duration. Handling equipments involved were: mobile harbour crane (MHC). Supply characterization As introduced in the previous sections. Some results on empty containers are proposed only for activities that systematically involve empty containers. time and type. 40’ and 20’x20’. MHCs operating in the Salerno Container Terminal are three Gottwald HMK 260 mounted on rubber-tyres and are mainly devoted to loading/unloading containers to/from berthed vessels. The following container types were considered: undifferentiated containers. Macrooperations are set up by operations. As regards spatial characterization. each demand flow was characterized by its distribution over time. Cartenì and de Luca. yards. operations and handling activities may be distinguished. 20’. empty. Statistical analysis for undifferentiated containers shows that the Gamma distribution function is always statistically significant. full vs. In particular.

856 0. The first movement is performed by the hoist.p. The analysis is focused on full containers.p.214 0. n.398 0.183 n.402 0.387 1.101 0. Each activity was analyzed distinguishing undifferentiated containers from 20’ and 40’ containers.657 0. The analyses carried out concern loading and unloading to the shuttle/truck.385 1. further distinguishing the tier.p. The second is the trolley gear. 0.252 0.485 n. © Association for European Transport and contributors 2009 11 . which allows the hoist to be positioned directly above the container for placement. which raises and lowers the container.121 n.825 σ 0.139 µ 1. 1.183 µ 1.690 0.p. 1. 1.326 1.139 0.193 0.102 0.230 0.188 0.387 1.288 0.p.216 0. which allows the entire crane to be moved along the working area.p.221 µ 1.084 0. In tables 10 and 11 means and standard deviations are reported for each activity.678 1. n.389 2. n.632 n.856 σ 0.263 0. n. σ 0.257 0. n.476 0.768 0.426 0.214 0.494 0.p. σ 0.835 0.p.326 GCs operating in the Salerno Container Terminal are four rubber-tyred gantry cranes used both for movement/storage of containers and for loading of shuttles/trucks.549 1. averages and standard deviations were estimated for trolley speed and crane speed. n.372 0.214 20’ empty 40’ full empty 2 x 20’ full full µ 1.340 1. µ 1.p.971 0.084 0.405 1.p. Moreover.244 0.083 0. 1.862 0. and loading and unloading to the stack (sometimes called pile).375 2. the Gamma distribution function proved the best solution for all analysed activities.101 n.971 σ 0. 1.435 0.871 0.263 20’ empty 40’ full empty 2 x 20’ full full µ 1.332 0.407 n.485 n. The third is the gantry.p.238 0.p.p.p.441 n.387 0.405 0. loading time from stack is reported.867 0.926 0.514 0.p.340 n.083 0.216 µ 1.366 1.926 0.227 0.230 µ 2.386 n. This crane type usually consists of three separate movements for container transportation. 0.385 0. n.316 0.867 σ 0. As regards undifferentiated containers.221 n. since these activities are the most frequent in the Salerno Container Terminal. µ 1.933 σ 0.121 0.825 0.862 0.664 0.871 0.366 0.p. Finally.272 0.835 σ 0.562 n.664 σ 0.366 Table 7 –MHC statistical results: parameters of Gamma distribution function activity loading unloading loading from dock loading from shuttle unloading to dock undifferentiate d µ σ 1.444 0.350 0.p.933 0.Table 6 –MHC empirical results activity loading unloading loading from dock loading from shuttle unloading to dock undifferentiate d µ σ 1.p. n. 1.386 1.690 0.389 0.p.102 0.768 σ 0.p.188 µ 2. Similar results were achieved on analysing activities for each container type and each tier number.214 0.

a.a.a.a. n. n.597 12.tier 3 0. n.a. n.618 0.tier 5 0.673 0.011 0.188 0.a. loading (from stack) .246 0.774 0.245 n.a. n.031 0. n. n. n.tier 4 0.659 0. Table 10 – GC statistical results: parameters of Gamma distribution function activity 20’ undifferentiated 40’ full full µ σ µ σ µ σ loading (from stack) 0.a.a. n.236 0.tier 1 1. n.760 0.339 n.097 0.766 0.687 0.325 0.tier 2 0.a.a.509 loading (from stack) .312 n. n. n.703 0.261 0.076 12.431 1.a.663 49.tier 4 0.a.380 0.a.a. n.a.a. activity 20’ undifferentiated 40’ full µ σ µ full σ µ σ trolley speed (with container) 11.a.a.613 0.tier 3 0.a. n.256 loading (from stack) .406 0. n.101 0.a.a. n.Table 8 – GC empirical results (minutes) activity 20’ undifferentiated 40’ full σ 0.353 1. n.a.a. unloading (to stack) .250 0.240 n. unloading (to stack) . n.142 12.a. n.653 4.753 0.592 0.671 0.222 0.586 n. unloading (to stack) .658 0.a.a. n.609 29. loading (from shuttle) activity trolley speed (with container) free trolley speed crane speed µ full µ σ σ 12.683 0. n.625 0.275 11.449 1.383 loading (from stack) .699 0.tier 2 0.623 0. n. n. n. n. n.352 n.498 © Association for European Transport and contributors 2009 4.a.367 0. loading (from stack) .a.tier 1 1.584 0.530 12 .a.a.a. n. n.706 0. unloading (to stack) .a. n.tier 5 0.311 0.a.416 13.a.916 σ 20’ full undifferentiated µ µ 40’ full µ σ µ σ 6. n.a.060 0.tier 4 0.307 n.712 0.280 0.457 unloading (to stack) 0.672 0. n. n.515 - - - 6.tier 1 1.340 loading (from stack) .a.tier 3 0.640 0.a. n. n.390 loading (from stack) .892 - - - - 4.460 1. unloading (to stack) .752 0.571 0. unloading (to stack) .a.583 0. unloading (to stack) .231 n.019 0.270 0.376 0.309 n.290 0.434 1.721 0.236 n. unloading (to shuttle) 1. unloading (to stack) . n.tier 3 0.348 1.647 0.740 4.tier 2 0.a.668 0.402 loading (from stack) 0.a. n.tier 4 0.169 0. n.261 0.309 n. n.713 0.638 0.614 0.355 0. n. n.283 0.tier 5 0.361 loading (from stack) .606 0.415 unloading (to stack) .769 0.636 0. n.a.a.a.399 unloading (to stack) .a.256 n.a.401 loading (from stack) .a. crane speed 11.769 0.605 0.203 free trolley speed 46. n.888 0.374 0.tier 2 0.a.022 0.658 0.758 0.a. n.422 unloading (to stack) 0.303 0.a.a.tier 5 0.508 30.202 5.025 0.a.331 0.a. n.741 0.a.a.560 0. n.352 n.243 4.561 loading (from stack) .tier 1 1.a.a.a.a.323 0.308 n.902 - n.a.

n.205 0. the results are showed.055 0.170 0. n.tier 3 0.a.a.a.a. n.215 0.a.086 n.a.098 n. n. n.272 unloading from shuttle/truck 0.155 0. Table 12 – RS empirical results activity undifferentiated µ σ 20’ 40’ full full µ σ µ σ loading to shuttle/truck 0.a.119 stacking time 0.144 0. n.a.a. n. n.355 0.157 n.250 0.307 0. As regards RSs speed.a.tier 5 0.056 n. They are used both to transport containers in short distances very quickly and to pile/storage them in various rows. n.a.tier 4 0. Moreover. stacking time . n.a.188 unloading from shuttle/truck 0.201 0.a.a. n. For the stacking time.148 n.tier 2 0.087 stacking time 0.a.167 0. n. n.186 0.311 0. For the mentioned activities Gamma random variable fits the data better due to best values of the validation test.056 0.a.a. n. n.074 0.304 0.a. stacking time . unloading from shuttle/truck and stacking.153 0.a. n. © Association for European Transport and contributors 2009 13 .a.344 0. In table 13.a.tier 1 0. n.a. n.334 0. n.tier 5 0.114 0.542 0.071 n.a. n. n.a. n.a.a. n.542 0.a. n.tier 1 0. stacking time . up to five. Table 13 – RS statistical results: parameters of Gamma distribution function 20’ activity undifferentiated µ σ 40’ full µ full σ µ σ loading to shuttle/truck 0. n. stacking time .a. stacking time . stacking time .118 n.357 0.a.140 n.a. the authors suggest to estimate the time duration of these activities directly.200 0.The RSs operating in the Salerno Container Terminal are eleven and are equipped with a twin-lift spreader able to move two full 20’ containers.365 0. Each activity was analyzed distinguishing undifferentiated containers from 20’ and 40’ containers.a.146 n. stacking time .a.a. The analysis is focused on full containers since in Salerno Container Terminal the main activities are related to full containers.a. stacking time .238 0. The analyses carried out concern: loading to shuttle/truck.260 0. n.288 0.tier 2 0.a.a. the time duration for each tier. n.185 0. stacking was analyzed distinguishing the tier number.a. n. n.a.a.a.tier 4 0. n.a.a. n.a.186 0.212 0. n.a. stacking time .a.tier 3 0. n.a. but it was not possible to distinguish containers typology.a.a.077 n.062 n. was computed. stacking time .164 n. n.236 0. n. n.

Sample mean values are used to estimate handling equipment time duration and there is no distinction between containers type. handling equipments time duration is modeled as a random variable and there is no distinction between containers type. 2 x 20’ full. were implemented: – Sample Mean Undifferentiated (SMU) model. The discrete event simulation model was developed in Witness® software. Handling equipment time duration is modeled as a random variable and containers type are explicitly taken into account: 20’ full and/or empty. – Sample Mean Container Type (SMCT) models. To test the applicability of the model architecture proposed for all the cited kinds of application. o shuttle waiting time. o reach stacker idle time.g. The time associated to each single activity is the realization of a random variable. 40’ full and/or empty.00 GB RAM. the values used for calibration are obtained by determining the average of 25 simulations. while local indicators are used to evaluate the benefits of medium/short-term investment and for real time applications.. 2 x 20’ full.4 SIMULATION RESULTS To plan investments for a container terminal several project scenarios need to be compared through performance indicator estimation. time spent moving a container from quay to vessel or from shuttle to stack). the implemented model was validated with respect to performance indicators coherent with those measured by the terminal monitoring office and summarized above: global performance indicators • terminal operation time: daily time required to bring all terminal activities to a close. Global indicators are generally used to evaluate the benefits of long-term investments.00 GHz. o quay/yard crane idle time. four different models based on four different handling equipment models. or local if referring to a single container (disaggregate indicators). 40’ full and/or empty. These indicators could be global. Starting from the model architecture proposed in the previous section. – Random Variable Container Type (RVCT) models. Sample mean values are used to estimate handling equipment time duration and containers type are explicitly taken into account: 20’ full and/or empty. • container indicator. – Random Variable Undifferentiated (RVU) model. 2. if referring to the container terminal as a whole (aggregate indicators). local performance indicators • handling equipment indicators. o vessel loading and/or unloading time. o shuttle transfer time. o reach stacker stacking time. the simulations were carried out on Intel Core 2 CPU 2. o gate in/out waiting time. o container operation time: time required to move a container with handling equipment (e. © Association for European Transport and contributors 2009 14 .

for more than 80% of the business day simulated.3% 4. while for random variable estimations the absolute estimation error is always lower than 10% and for the RVCT this value is lower than 5% for 75% of the handling equipment simulated. Regarding the handling equipment indicators. the great variability of the phenomenon observed produces absolute estimation errors for sample mean estimations lower than 10% only for 25% of the handling equipment simulated. results in terms of global indicators show an average absolute percentage error of about 9% for the sample mean model. whereas in using the random variable model the percentage estimation error is about 3%. in all other cases the estimation errors are about 30%.9% 11. Results obtained using random variable handling models are significant: average absolute percentage errors for handling equipment indicators are more than 6% with the RVU handling model.50 0. the latter are below one minute.6% 3. for over 80% of the observations the absolute estimation error is © Association for European Transport and contributors 2009 15 .0% 11. the absolute percentage error is lower than 10%. Results in terms of global indicators show an average absolute percentage error of more than 10% for the SMU handling model. when only using the RVCT handling models the absolute percentage estimation error is acceptable (>11%). The former require about 20 minutes. The use of sample mean handling models does not produce very good results in terms of local indicators. Using the Container Type models.5% 28. In figure 3 and 4 a sensitivity analysis is proposed.2% In figure 2 cumulative absolute percentage estimation error variation is reported.30 absolute percentage estimation error local indicators global handling indicators containers equipment 10. whereas in using the RVU handling model the percentage estimation error is lower than 5%.1% 13. cumulative absolute percentage error variation are reported (figure 2).5% 2. Table 14 – Average absolute percentage estimation error: estimation sample handling model SMU SMCT SRVU RVCT simulation time (minutes) 0. With respect to container indicators.6% 6.2% 30. absolute percentage error (table 14). By contrast.60 18. If the aim of the simulation is to estimate container movement.0% 29. The results in terms of simulation time point out that random variable models require a computational time much greater than sample mean ones. average percentage estimation errors exceed 13% for handling equipment indicators and are about 30% for container indicators. and lower than 15% for over 60% of the observations. and about 3% with RVCT handling models.8% 8.70 21. With respect to the global indicator the results described above are confirmed. the only suitable handling model is the RVCT one: only for this model is the absolute estimation error of the container operation time lower than 10% for 45% of the observations.In the following tables and figures analyses on computational time (table 14).

© Association for European Transport and contributors 2009 16 . the simulation results show an expected decreased benefit effect. SMCT and RVU) produce unacceptable absolute estimation errors for the purposes of the simulation.g.lower than 30%. Terminal operation time Handling equipment indicators 100% Percentage of handling equipments Percentage of business days 100% 80% 60% 40% 20% 0% 80% 60% 40% 20% 0% 0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% Absolute percentage estimation error Absolute percentage estimation error SMU / SMCT SMU / SMCT RVU RVCT RVU 100% RVCT Figure 2 – Cumulate absolute percentage estimation error: terminal and handling equipment Container operation time 100% Percentage of containers 80% 60% 40% 20% 0% 0% 20% 40% 60% 80% 100% Absolute percentage estimation error SMU / SMCT RVU RVCT Figure 3 – Cumulate absolute percentage estimation error: containers To evaluate the model’s sensitivity. In the Salerno container terminal there are critical activities: the gate-in procedure and vessel loading time. -10% of service time produces about -35% of truck waiting time.. the truck waiting time (queue waiting time plus service time) percentage variation was estimated with respect to the service time percentage variation. percentage variation in equipment number or equipment performance) were estimated. the percentage performance indicators variations with respect to the input variables percentage variations (e. in particular. In figures 3 and 4 some representative results obtained with the RVCT handling models estimated are reported (results related to the other performance indicators are analogous). With respect to gate-in activity (figure 3). As can be seen in figure 2. the other models (SMU.

GATE IN service time % var..0% Vessel loading time % var.0% -80. Hence a crane number increase does not produce a great reduction in terms of vessel loading time. Instead. -80% -60% -40% -20% 0% -20% -40% -60% GATE IN waiting time % variation 0% -80% Figure 3 – An example of the model’s sensitivity: gate-in waiting time variation measured against gate-in service time variation With respect to vessel loading time (figure 4). Shuttle lag time % var. a shuttle lag time reduction produces a significant vessel loading time variation (e.0% -60. one of the critical points of the Salerno terminal is the low arrival frequency of the shuttles near the vessels.0% -100. This phenomenon occurs till crane capacity is reached.20% of service time produces about -55% of waiting time while -30% of service time produces more than -70% of waiting time. -30% of shuttle lag time produce about -30% of vessel loading time). % var. + 50% crane numb. 0% 20% 40% 60% 80% 100% 0.g.0% Crane numb.0% -40. Shuttle / Crane number % var. then only a combined increase of crane number with the shuttle lag time could further improve loading performance. -20. Figure 4 – An example of the model’s sensitivity: vessel loading time variation measured against shuttle/crane number variation © Association for European Transport and contributors 2009 17 . Shuttle lag time % var.

gantry cranes. the remaining contributions carry out very simple approaches (deterministic) and/or give scanty information: on the estimation approach pursued. its calibration and its validation. not much can be found in the literature. With respect to crane speed. Half of them adopt a stochastic approach and show estimated parameter values. with particular attention to estimation results of handling activity models for three handling equipment (mobile harbour cranes. modellers and practitioners a sort of a guidelines useful to point out the strengths or weaknesses of different approaches. As regards shuttle performances (speed. few models exist and are © Association for European Transport and contributors 2009 18 . The aim was to suggest to analysts. Moreover. waiting time …). (iii) the segmentation of container type.). If on the one hand. (c) the simulation of the effects of different hypotheses regarding (i) the approach to estimate handling activities time duration (sample mean vs random variable estimation). With respect to other handling equipment. travel time. As regards vessel cycle time. all existent papers propose deterministic and aggregate models. Most of the contributions deal with vessel loading/unloading operations. on experimental data used. no one investigates the effects that different hypotheses on handling equipment models calibration may have on the simulation of container terminal performances. models calibrated and corresponding parameters. 40 feet. Guidelines were presented through: (a) a preliminary in depth literature survey. In this paper a discrete event simulation model was proposed and applied to the Salerno container terminal in order to address some of the open issues introduced above.g. There is substantial heterogeneity regarding the level of aggregation of activities involved and how such activities are aggregated in a single macro-activity. empty. (b) the description of the developed discrete event models. most of the existing papers are only focused on the application and/or on the comparison of design scenarios and do not pay great attention on the model set-up. whereas other one propose a triangular distribution function for the forklift. many contributions do not present any information on equipment handling models used. a stochastic approach is unanimously proposed. Literature review allowed a comprehension of what has be done in the last twenty five years in container terminal simulation field and gave detailed information on approaches pursued. full…. software used. on parameters estimated and on parameters value. vessel loading vs explicit simulation of elementary activities sequence). reach stackers) and for different container type (undifferentiated. With regard to crane loading/unloading time both deterministic approach and stochastic one have been pursued.5 CONCLUSIONS In literature numerous efforts may be found in the field of simulation of a container terminal. 20 feet. Among the contributions proposed only ten papers give information on the handling equipment models used. (ii) the level of aggregation of handling activities (e. someone propose deterministic values for a straddle carrier. Such effects could not be negligible and should be investigated with respect to different planning horizons. such as strategic or tactical.

In this context. In such a context sample mean handling models may be used. © Association for European Transport and contributors 2009 19 . it is important to have a simulation model efficient. since the aim is to simulate whole container terminal performance with respect to longer simulation time horizon (for instance 365 days). benefits from medium/shortterm investments and/or benefits from real time strategies should be estimated. path winding. both sample mean and random variable estimation can be pursued. and Normal. different planning horizons (strategic vs tactical) were investigated with respect to two different modelling hypotheses: sample means vs. since realistic micro-simulation of single containers movement does not sensibly improve simulation results. From a statistical point of view. In particular. only the Random Variable Container Type handling models can be used. random variable handling models. In strategic planning horizon. Discrete event model description and calibration results enriched the existent state of art. Finally. Gamma random variable led to better goodness fit for all handling activities and for all container type involved. The application of discrete event model allowed to draw some operational guidelines on the best approach to simulate handling equipment activities and/or on the best approach to simulate terminal performances with respect to different planning horizons. highlighted a family of distribution functions suitable to simulate handling equipment time duration and allowed to define the best performing distribution functions for each handling equipment and for each container type. If the focus is to simulate handling equipment time duration. Sample mean models could be used for estimating handling equipment indicators. but greater average absolute percentage errors must be accepted (11% to 13%) with respect to random variable handling models that allow absolute percentage error varying from 3% to 6%. In tactical planning horizons. it is important to have a simulation model efficient and realistic in the simulation of single container movement. Gamma and Weibull distribution functions turned out statistically significant to interpret handling activities time duration. terminal operation time). If the focus is to simulate single container trip time (container indicators). gave some insights on the best calibration approach (Moment.g. Maximum Likelihood).easily transferable to different case studies (due to the influence of path length. traffic vehicle congestion inside the terminal and so on). easy to implement and realistic in the simulation of aggregate terminal performance indicators (e. while it is about 30% for the other approaches. Maximum Likelihood estimation approach seemed to be the most performing one. Obtained results showed that only the use of random variable handling models allows satisfactory simulation results. In this case average absolute percentage estimation error is about 11%. The whole set of distribution functions (and of their parameters) allowed to implement different simulation models as activities level of aggregation changes and as container type changes.

. Computers and Operations Research. Baker. 71(2) 107-116. E. Springer. A. Wilson.E. © Association for European Transport and contributors 2009 20 . R. (1998) Simulation and genetic algorithms for ship planning and shipyard layout. Rizzoli. Of Civil Engineering. L. C.. M.. 141-144. (1980) Simulation as an aid to the study of a port as a system. London. (2009) Simulation of the Salerno Container Terminal. University of Salerno. eds).. C. de Luca. Y. U. 443-450. Simulation. (2008) A macroscopic model of a container terminal based on diachronic networks. P.. Golden.. J.. The Journal of the Operational Research Society.. P. Ship Operation Automation III. Silberholz. Cartenì. S. Ein-Dor. A. Computers and Industrial Engineering. For a complete list of all references. Borovits. (1987) A port simulation model for bulk cargo operations. (1998) Assessment of simulation models based on trace-file analysis: a metamodeling approach. the reader can refer to the technical paper by Cartenì and de Luca (2009). (1991) Using simulation to study the impact of work rules on productivity at marine container terminals. 48 (6) 236-246. Bruzzone. B.M. Gambardella. G.. A. 433 – 452. Schedule-Based Modeling of Transportation Networks: Theory and Applications (Nuzzolo. (1988) A simulation analysis for a transtainer-based container handling facility. G.REFERENCES Only the most relevant contributions are cited. Signorile. S. Zaffalon M. S. A. de Luca. B.. A. I. McDowell E... A.H.. 71(2) 74--83. K. PTRC. W. Simulation Today. A... Simulation. El Sheikh. D. 18 (5). (1983) A Simulation Study on the Determination of the Best Investment Plan for Istanbul Seaport The Journal of the Operational Research Society. M. L.. 14 (2) 113-125. Harding. Cartenì.. Y. 38 (8) 673-681. Tolujiev et al. (1998). Simulation. Proceedings of the European Transport Annual Meeting: Transportation Planning Methods. S. Collier. S. D. ACM. Randhawa S. Cantarella. I. Noh. Paul.. Cartenì. Tuǧcu. (2005) A simulation model for a container terminal. R. S. Park.E. D. N. Dept. (1987) A Microcomputer-Based Simulation Study of a Port. A. E. R. Chung. (1990) Computer simulation of seaport container terminal. 51-6. technical paper. 34 (6) 479-487.. de Luca. Cantarella. Balmer. Simulation and planning of an intermodal container terminal.

Simulation. animation. 39 (3) pp. R. (1983) A simulation study on the determination of the best investment plan for Istanbul seaport. G. Hussain. J.M.. Tugeu. 13 (1) 14-20. A. Novitsky. G. 71 (2) 96-106.A. M.. and visualization of seaport operations. Shabayek. F. Decision Support Systems. Merkuryev. C.. (2005) A decision support system for operations in a container terminal. Liu. Legato. 235–248. (2001) Berthing planning and resources optimisation at a container terminal via discrete event simulation. P. E. 20 (3). V. J. Simulation. Nevins. Y. (1988) Decision support system for container terminal planning. J. (1998) Simulation.. Ginters.S.Y. R. 59 221-230. Journal of the Operational Research Society. 66 (5) 291-300. 133 537-547. Janssens. Van Hee.. G. (2000) Simulation and analysis for the Kelang container terminal. Simulation.. R. E. K. Love. E. European Journal of Operational Research. Tolujew. Linn R. J. Merkuryeva. (1998) A modelling and simulation methodology for managing the Riga Harbour container terminal.. Y. Viktorova. Yeung W. W. 140 1–11. R. (1996) An interactive simulation model for the logistics planning of container operations in seaports. Yun. L. Bragen. Choi. Murty. © Association for European Transport and contributors 2009 21 . J. M. Simulation.Kozan.W. 34 262–272... E. Wan. 71 (2) 84--95. Pronins. K. R.. Thiers. K. Y.. Production Economics. Blumel. (1998) A port simulation model as a permanent decision instrument. Transportation Planning and Technology. M. K. Logistics Information Management.. Rina.. European Journal of Operational Research. S. G.. (1999) A simulation model for container-terminal operation analysis using an object-oriented approach.. Wijbrands. M. 309-332. (1997) Comparison of analytical and simulation planning models of seaport container terminals. Int... 71 (2) 117--125. M. Ramani..J. J. European Journal of Operational Research. (2002) A simulation model for the Kwai Chung container terminals in Hong Kong.. 34 (6) 479--487. Tahar. K. Macal.