Evaluation of the floaterm concept at marine container terminals via simulation

Simulation Modelling Practice and Theory 54 (2015) 19–35

Contents lists available at ScienceDirect

Simulation Modelling Practice and Theory journal homepage: www.elsevier.com/locate/simpat

Evaluation of the floaterm concept at marine container terminals via simulation M.A. Dulebenets a,1, M.M. Golias b,⇑, S. Mishra c,2, W.C. Heaslet d,3 a Department of Civil Engineering, Intermodal Freight Transportation Institute, The University of Memphis, 302A Engineering Admin. Bldg., 3815 Central Ave Memphis, TN 38152, USA b Department of Civil Engineering, Intermodal Freight Transportation Institute, The University of Memphis, 104B Engineering Science Bldg., 3815 Central Ave Memphis, TN 38152, USA c Department of Civil Engineering, Intermodal Freight Transportation Institute, The University of Memphis, 112D Engineering Science Bldg., 3815 Central Ave Memphis, TN 38152, USA d FedEx, 3101 Tchulahoma Ave, Memphis, TN 38118, USA

a r t i c l e

i n f o

Article history: Received 11 October 2014 Received in revised form 19 February 2015 Accepted 20 February 2015

Keywords: Maritime transportation Container terminals Simulation Floating quay cranes

a b s t r a c t International seaborne trade rose significantly during the past decades. This created the need to increase capacity of existing marine container terminals to meet the growing demand. The major objective of this paper is to evaluate the floaterm concept using simulation modeling and determine if it can improve terminal productivity. The main difference between floaterm and conventional marine container terminals is that, in the former case, transshipment containers are handled by off-shore quay cranes and stored on container barges. Two terminal configurations performance is compared (vessel handling times and equipment utilization) under normal and disruptive conditions. Computational experiments confirm preliminary expectations that the floaterm concept can enhance efficiency of marine container terminal operations under normal and disruptive conditions. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction The amount of cargo, transported by vessels, substantially increased over the last thirty years. According to United Nations Conference on Trade and Development (UNCTAD) overall seaborne trade, inlcuding containerized cargo, major bulk cargo, other dry cargo, and liquid bulk cargo, increased by roughly 3.6 billion tons from 2000 to 2013 and reached 9.6 billion tons [1]. International seaborne trade rose by more than 120% by weight from 1980 to 2008 primarily from increasing standards of living, fast industrialization, population growth, and competitive markets [2]. Containerized cargo increased by 6.6% in tonnage from 2012 to 2013, and the future growth was also projected for 2014 [1]. The Journal of Commerce [3] indicates that ‘‘seeking efficiency and economies of scale, the world’s container carriers are increasingly ordering megaships capable of handling more than 8000 20-foot-equivalent container units’’. To meet this growing demand, while facing capacity expansion limitations (e.g., lack of land, high cost of expansion, etc.), it is necessary to provide proper planning and ⇑ Corresponding author. Tel.: +1 901 678 3048; fax: +1 901 678 3026. E-mail addresses: [email protected] (M.A. Dulebenets), [email protected] (M.M. Golias), [email protected] (S. Mishra), [email protected] (W.C. Heaslet). 1 Tel.: +1 901 605 8737. 2 Tel.: +1 901 678 5043; fax: +1 901 678 3026. 3 Tel.: +1 901 224 6790. http://dx.doi.org/10.1016/j.simpat.2015.02.008 1569-190X/Ó 2015 Elsevier B.V. All rights reserved.

20

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

management of terminal operations. Deployment of larger vessels with higher capacity can further add constraints to seaport operations [4]. The port capacity can be increased by upgrading existing ports or constructing new facilities [5]. There are additional alternatives that do not involve construction, such as improvement of conventional equipment and productivity by introducing new forms of technology [6], information systems [7], and work organization [8]. Container terminal operations can be divided into quayside, storage yard, and landside operations. Quayside operations deal with berthing of vessels, stowage planning, quay crane (QC) assignment, and QC scheduling for (un)loading containers. Note that stowage planning is the only function not solely controlled by the terminal operator and receives significant input from the captain of the vessel. Storage yard operations include stacking and retrieval of inbound, outbound, and transshipment containers by gantry cranes (GCs) at yard blocks. Internal transport vehicles (ITVs) provide container transfer between the quayside and the storage yard. Landside operations consist of receiving or delivering containers by drayage trucks (DTs). There are three main quayside transfer processes at marine container terminals (MCTs): (a) vessel-to-yard (i.e., import), (b) yard-to-vessel (i.e., export), (c) and vessel-to-vessel (i.e., transshipment). Marine terminals operate as follows: once a vessel has entered the port, it is berthed at its assigned berth, and once moored, ship-to-shore QCs start (un)loading containers. ITVs (yard trucks, straddle carriers, automated guided vehicles, automated lifting vehicles, etc.) transfer containers between the quayside and pre-assigned blocks of the storage yard, where GCs arrange them either parallel or perpendicular to the berth. Import containers are delivered to the port by vessels, while export containers are drayed to the port by DTs through the gates (usually at least 24 h before the vessel calls at the port) or by rail (if on-dock rail access exists). Once a DT enters a terminal, it either travels to the assigned blocks in the storage area or to a receiving area where the container is (un)loaded. Transshipment occurs, when cargo, delivered by one vessel (usually called as mother vessel), is moved to another vessel (usually called as feeder vessel). Transshipment containers can be transported from vessel to vessel with or without temporary storage at the storage yard. Realizing efficient operations at MCTs remains a difficult task (most operations are formulated as mathematical programing models belonging to the NP class). Handling equipment and containers should be properly allocated for quayside, landside and storage areas. QCs should be assigned to particular berths, and their numbers depend on several factors (i.e., total number of QCs available; total number of vessels assigned to each berth; total number of containers to be handled for each vessel, etc.). Particular dispatching strategies of ITVs should be chosen to decrease or eliminate idle time of QCs and GCs (although the former always receive priority). Available GCs should be properly allocated between yard blocks and if more than one GC serves a yard block, particular safety policies should be taken into account to avoid clashing. There are also traffic congestion issues for large container terminals due to longer travel distances travelled by ITVs. The allocated equipment should be utilized in the most efficient manner (e.g., dual cycling of QCs and ITVs). To improve performance of MCTs by increasing quayside capacity with minimal capital investment a new concept (named floaterm) was proposed in early 2000 [9,10]. The floaterm concept includes two-sided operations (when a vessel is moored between the terminal berth and the crane barge as shown in Fig. 1A) and midstream operations (when a vessel is moored to the crane barge in the sea as shown in Fig. 1B). The floaterm concept was originally applied at the Ceres Terminal (Amsterdam, the Netherlands) in 2002 with throughput increasing by 24.6% from 2000 to 2005 [11,12]. No information was made available as to the role the floaterm concept played in the throughput increase. According to Ashar [9] and Liftech, Inc. [10] though the floaterm concept could significantly improve performance of seaports, decrease the size of the storage yard, reduce the number of handling equipment, reduce congestion, etc. To the best of the authors’ knowledge, no computational study exists (to date), describing and modeling the impact of the floaterm concept on MCT operations. In this paper we use simulation to compare operations (under normal and disruptive conditions) of a conventional MCT to a terminal with the floaterm concept and quantify (any) productivity gains, that may be realized by the latter. The rest of the paper is organized as follows. The next section presents a review of the literature on simulation modeling at MCTs. The third section describes the developed simulation models, and the fourth section presents a number of numerical experiments, conducted as part of this study. The fifth section discusses results from the study, and the last section provides conclusions and future research directions. 2. Literature review Operational, tactical, and planning level decision problems at MCTs have been the focus of interest of numerous researches and practitioners, especially during the last fifteen years [13–17]. In this paper the reviewed literature focuses on simulation modeling applications at MCTs (i.e., berth allocation, stowage planning, QC assignment and scheduling, storage and stacking, landside and quayside transport, vulnerability/resiliency of marine terminals, etc.). For reviews of the general literature on MCTs we refer to Steenken et al. [13], Stahlbock and Vos [14], Theofanis et al. [15], Bierwirth and Meisel [16], and Carlo et al. [17]. These articles include historical overviews of world seaborne trade, description of MCT structure and handling equipment, and classification of published literature by different topics, which is out of the scope of this paper. 2.1. Quayside operations Several simulation studies have been published on quayside operations (i.e., berth allocation/scheduling, stowage planning, QC allocation/scheduling, ITV scheduling and routing). Han et al. [18] addressed the problem of simultaneous berth and

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

21

Fig. 1. Two-sided and midstream applications of the floaterm concept.

QC scheduling with vessel stochastic arrival and handling times. They minimized vessel total service time (waiting and handling time) and weighted tardiness. The authors developed a simulation model to solve the problem by applying Genetic Algorithms (GAs). Arango et al. [19] studied the static discrete berth allocation problem at MCT of the Port of Seville. The objective of the model was to minimize the total service time of vessels. The authors developed a GAs based optimization module combined with the Arena simulation software. The optimization module was used to generate vessel to berth assignments and pass the information to the Arena software to simulate vessel handling operations. Qi and Song [20] considered the problem of optimal vessel scheduling in liner shipping routing, taking into account the impact of uncertainties. The objective aimed to minimize the total expected fuel consumption and penalties due to vessel delays. The authors introduced simulation-based stochastic approximation methods to solve the problem. Regarding stowage planning problems, Aye et al. [21] created a visualization and simulation tool for an automated stowage planning for large containerships. The stowage plan was generated by considering several factors that included type of cargo, visibility adjustments, weight limits, trim adjustments, etc. The objective was to maximize the vessel stability, while minimizing vessel dwell time at the port. Low et al. [22], Min et al. [23], and Xiao et al. [24] developed an automated stowage planning generator for large containerships. The system included three different modules: a stowage plan generator, a stability module, and an optimization engine. The stowage plan generator was capable to create a plan with reasonable crane intensity and number of re-handles (i.e., containers needed more than one lift by a QC). A stability adjustment module employed several algorithms to search for alternative container locations. An optimization engine began with a feasible stowage plan, modified it with the stability module, and optimized container allocation (e.g., minimize number of rehandles, obtain a reasonable QC intensity, appropriate weight distribution etc.). Canonaco et al. [25] proposed a queuing network model for the management of berth crane operations at MCTs. An event-graph methodology was employed to get a systematic representation of the terminal. The objective minimized the total waiting and service times of vessels. The authors developed a simulation model to solve the problem. Numerical experiments were conducted based on the operational data, provided by the Port of Gioia Tauro Port, Southern Italy. The simulation tool was found to be efficient for performing what-if scenario analysis. 2.2. Storage yard operations Various simulation studies have been published on storage and stacking of containers at the storage yard of container terminals. Saanen and Valkengoed [26] compared three stacking alternatives: (i) single RMG (rail mounted gantry crane), (ii) cross-over RMG, and (iii) twin RMG with cross-over RMGs (with the latter showing the highest productivity). Guo

22

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

et al. [27] addressed the problem of container yard block dynamic partitioning into non-overlapping zones to minimize total vehicle waiting time. Real time data driven simulation computed the dynamic workload of each crane in the proposed partitioning algorithm. Petering and Murty [28] assessed the effect of block length and GC deployment on the overall MCT performance. The authors applied a discrete event simulation to evaluate various scenarios (small terminal less equipment, small terminal more equipment, large terminal less equipment, and large terminal more equipment). Results of simulation runs indicated that block lengths between 56 and 72 slots yielded the highest QC and GC working rates. Petering [29] conducted a similar study using discrete event simulation to determine the effect of block width at a container seaport performance. It was found that the optimal block width ranged from 6 to 12 rows depending on the quantity of allocated equipment and terminal layout. Petering et al. [30] analyzed real-time GC control systems for MCTs. Several GC dispatching algorithms and yard truck availability criteria were considered. Numerical experiments showed that GCs should prioritize retrieving to stacking. A GC system should also account for ITVs heading to the block, and not only for those which have already been waiting for a service. Petering et al. [31] considered nine problems in strategic and tactical container terminal management. The paper demonstrated how the average QC rate (QCR) affected the capacity of the container terminal, the number of GCs and trucks within the terminal, the terminal operational system’s ability, and the overall facility in general. 2.3. Landside and quayside transport Cheng et al. [32] investigated the problem of automated guided vehicle (AGV) dispatching at MCT to reduce the total waiting time of AGVs. Numerical experiments indicated that the network flow method was more efficient than the greedy deployment scheme. Yang et al. [33] compared AGVs and automated lifting vehicles (ALVs), and evaluated their effect on automated container terminal operations performance. The authors developed a simulation model (with Visual Basic) to emulate terminal operations. ALVs resulted in less completion time for all jobs as compared to AGVs. Similar research was undertaken by Vis and Harika [34]. A simulation model, built using Arena, demonstrated that an additional 38% of AGVs were required as compared to ALVs to obtain the same container handling times. Esmer et al. [35] conducted a study to determine the optimal number of yard trucks for a Turkish Port based on the lean and green concept (i.e. minimize overall environmental damage due to emissions production from port operations). It was observed that reducing allocated trucks from 5 to 2 would significantly decrease the total environmental damage without affecting QC productivity. Garro and Russo [36] investigated a straddle carrier (SC) routing and dispatching problem at MCT. They minimized the total QC and SC idle time, and the total SC distance traveled using a simulation model of the terminal built with the Repast Simphony Toolkit. Petering [37] employed a fully-integrated discrete event simulation to evaluate various real-time dual-load yard control systems. It was found that the yard truck control system based on the most starving QC policy resulted in higher QCR. Terminal performance also improved, when all yard trucks were assigned to one dispatching pool. Dual loading increased QCR for the QCs handling one container at a time. Francesco et al. [38] considered the problem of empty container repositioning in marine container networks, taking into account possible man-made and natural disruptions. The simulation model accounted for uncertainty of various parameters (e.g., the number of empty container to be handled, the storage capacity for empty containers, the vessel capacity for empty containers, maximum number of empty containers to be handled). The objective was directed to minimize the total transportation cost, including loading, transfer, unloading, and storage. Uncertainty was assessed using a scenario-based stochastic programming approach. Computational experiments indicated that deterministic formulations were effective only in cases of correctly predicted disruptions. 3. Model description As revealed by the literature review, simulation is widely used for what-if scenario analysis and comparison of various resource assignment policies at MCTs. The scope of the past research, related to the floaterm concept, included theoretical discussions of its advantages [9,10], technical feasibility of the floating QCs application [12], analysis of changes in the stowage planning [12], and economical and operational feasibility of the floaterm concept [12]. The main objective of this study is to conduct a detailed comparative analysis of conventional and floaterm terminal types using simulation modeling under normal and disruptive conditions. In this section two simulation models: one for a conventional and one for a floaterm MCT (from now on referred to as CMT and FMT respectively) developed using the FlexSim simulation software package [39] will be presented. The simulation models will be used to estimate potential benefits of the latter terminal configuration. This section will describe in details the modeling assumptions of the quayside, yard, and landside operations for both terminals, including terminal layout, container types, handling equipment assignment, and characteristics of disruptive events. 3.1. Terminal and vessel characteristics In this paper both CMT and FMT are assumed to have 3 berths. The length of each berth is equal to 380 m, which allows mooring of Neo-Panamax vessels. The width of the apron area, connecting the quayside and the storage yard, covers 90 m. The main geometric characteristics of CMT and FMT are presented in Fig. 2A and B respectively. Note that the terminal layout

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

23

Fig. 2. Terminal layouts: (A) CMT and (B) FMT.

and dimensions for both CMT and FMT were based on information found in the available literature [28–31]. Container flow is illustrated in Fig. 3. At CMT three QCs are located on the quayside at each berth and handle all containers from each vessel. At FMT two QCs are located on the quayside at each berth and only handle export and import containers, while transshipment containers (from/to the feeder barges) are handled by one QC (at each berth), located on a crane barge [40]. During disruptive events QCs on crane barges are allowed to handle part of the import container demand. The capacity of each barge was assumed to be 200 TEUs. Once the barge is fully loaded, it is towed to the assigned feeder vessel by push boats. Setup time at the quayside is assumed to be 10 min (5 min for mooring 5 min for detaching). Setup time for the feeder barge (mooring to the crane barge and detaching from the crane barge) was assumed to be 10 min and can be modified in the model as needed. On-shore and off-shore QC productivity (QCP) was assumed to follow a triangular distribution [triangular (1.0, 1.5, 3.0) minutes per container move], which translates to a mean (nominal) value of 40 moves/h/QC [10]. The triangular distribution (upper and lower bound and mode) were chosen based on values found in the literature [28–31]. Workload between QCs for each vessel is equally distributed in both simulation models, as this policy increases productivity by minimizing vessel handling time [41]. It was further assumed that the stowage plan for each vessel satisfies stability conditions (e.g., stack weight limit, moment equilibrium between bow and stern and between the left and right side of the vessel).

3.2. ITV characteristics Two types of ITVs (yard trucks or YTs and automated lifting vehicles or ALVs) are assumed to carry containers between the quayside and the storage yard (see Fig. 2). Each terminal configuration can use only one type of vehicle (either YT or ALV). Usually, speeds of empty and laden YTs are 40 and 25 km/h respectively [28–31]. In this paper YTs speed was set constant and equal to 30 km/h. We assume that ALVs have the same speed = 30 km/h [33]. ITVs are assumed to carry one 20 foot (ft.) container but other container types can be introduced in both models (e.g., 20 ft., 40 ft., 45 ft., etc.) as well.

24

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

Fig. 3. Container flows at terminals.

Vessels are served by three gangs of ITVs (either YTs or ALVs), each dedicated to only serve all QCs at a particular berth. Several studies confirm that this multi-crane oriented (a.k.a. pooling) strategy, when ITVs are shared between QCs serving the same vessel, is more efficient [37,42,43]. Productivity of QCs with a multi-crane oriented strategy is approximately 20– 25% higher than the alternative strategy (where ITVs are not shared among QCs at the same berth), most likely due to the increase of QC and ITV dual cycling. 3.2.1. ITV deployment The ITV deployment strategy, used in this study, is depicted in Fig. 4 for both YTs and ALVs. The main differences between the two deployment strategies are: (a) QCs do not have to wait for an ALV to become available to unload a container, and (b) ALVs do not have to wait for a QC to pick up the container they are delivering to the quayside. Once a QC picks up a container from a vessel (Fig. 4A and B), it searches for the first available ITV to load the container. If more than one ITVs are available, the model will assign the container to the ITV closest to the QC at the given simulation time. If idling ITVs are not available, the QC will either wait for the first available YT or, for the ALV case, unload the container to the buffer area. If there is only one idling ITV, it will be assigned to the first available job (i.e., minimization of waiting time for QCs). A similar deployment strategy is applied for export/transshipment containers moved from the storage yard to the quayside (see Fig. 4C and D). The model computes distances between QCs, ITVs, and GCs based on the terminals’ road network. If a road network does not exist, the model estimates distances based on centroids. When a loaded ITV enters a yard block, it travels along the handling lane to the assigned GC (see Fig. 5). An empty ITV shuffles to the bypass lane. While YTs need to wait for a GC to pick up/place the container from/to their chassis, ALVs are capable of (un)loading the container from/to the handling lane without waiting for a GC service (which increases productivity). 3.3. Quayside and storage yard buffer areas Quayside and storage yard buffer areas of MCT serve three functions: (a) an area for cranes to operate on, (b) an area for ITV circulation, and (c) an area for drop-off/pick-up of containers (by QCs, GCs, and ITVs). Based on preliminary simulation experiments the optimal size of both buffer areas was determined and findings were similar to Vis and Harika [34]. Specifically, the buffer area size at quayside significantly affected QCP, but the buffer area size at the storage yard did not result in any substantial productivity difference. The buffer area capacity at the quayside and storage yard was set equal to three containers per QC and two containers per storage block, respectively. 3.4. Storage yard configuration The storage yard consisted of 30 and 15 yard blocks (10 and 5 blocks per berth) for CMT and FMT respectively. The storage yard size at FMT was set smaller as transshipment containers are stored on barges. Each storage area at CMT has separate yard blocks dedicated to import, export, and transshipment containers. The capacity of each block was assumed to be 600 TEUs (6 rows  5 tiers  20 bays). Length of each bay was assumed equal to 24 ft. (including 4 ft. of clearance space). GCs

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

25

Fig. 4. ITV deployment strategy.

Fig. 5. ITVs in the yard block.

(un)load containers from ITVs from/to the assigned yard block based on the type of container (export, import, and transshipment). This particular terminal layout was chosen as it reduces the total distance traveled by ITVs and thus task completion time of ITVs, QCs, and GCs [44]. Import containers were allocated to the blocks, situated closer to the gates. Transshipment containers were placed to the blocks, located closer to the quayside. Export containers were allocated on the side blocks of each storage area. Exports are transported by DTs, passing through the terminal gates. DTs deliver export containers to assigned yard blocks, once the space is available (queuing occurred when the space was not available). Then GCs unload containers from DTs to the assigned yard blocks.

3.5. Storage yard handling equipment A group of rubber-tyred GCs is assigned to each storage area. Container stacking and retrieval times are assumed to follow a triangular distribution [28–31] with a nominal value of 20 moves/h [triangular (2.5, 3.3, 3.0)], including reshuffling time required by a GC to retrieve a container.

26

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

3.6. Optimal QCP determination The size of each ITV gang and GC group, required to obtain the optimal QCP for the two terminal types (CMT and FMT) under normal operating conditions, was determined based on simulation runs, where the number of ITVs was changed from 1 to 40 and the number of GCs from 1 to 30, both with an increment of one. Note that optimal and nominal QCP values differ as the latter is estimated based on the assumption of zero idle time for QCs. The optimal QCP will be less than or equal to the nominal productivity, as it depends on the volume of containers and resources (ITVs and GCs) allocated to serve QCs (i.e., a QC may have to wait for YTs to become available to pick up/drop-off containers). 3.7. Disruptive event assumptions Taking into account the growing international seaborne trade, it is important for port operations to exhibit resilience to potential man-made and natural disrupting events [45–47]. The scope of this research included comparison of the two terminal configurations not only under normal (as discussed previously), but also under disruptive operating conditions. In this paper two disruptive events were assumed for each type of container terminal:  Disruption A: 33.3% of on-shore QCs and GCs are not available for 12 h.  Disruption B: 50.0% of on-shore QCs and GCs are not available for 24 h. Note that damaged QCs and GCs will be available to handle containers at full capacity immediately after the end of each disruption. For each disruptive event the following assumptions were made as to their effect on the terminal operations:  Disruptions occur at the simulation time of zero (the beginning of each simulation run).  A disruptive event is assumed to affect the gate area, i.e., export containers will not be delivered to the terminal and import containers will not be picked up by DTs during the event.  ITVs are not damaged by the disruptive event. Even in the case where ITVs are potentially affected, they can be replaced (which may be difficult in the cases of damaged GCs or QCs) as terminal operators usually have more ITVs than required for daily operations (to account for downtime/maintenance and fluctuating demand).  In the cases of disruptive events, floating QCs will handle a portion of the import containers to compensate for quayside reduced productivity.  When import containers are handled by floating QCs, they will be placed on barges, and stored at the floating yard. Once vessels depart the port these import containers can be unloaded by QCs or mobile harbor cranes.  The number of container barges is sufficient to handle the import and transshipment containers.  Disruptive events affect only landside operations. Disruptions, causing breakdowns of seaside operations (e.g., tsunami), will result in a complete terminal (both CMT and FMT) shutdown. Storage yard operations will still be possible, however, vessels cannot be moored and served.  Disruptive events have deterministic features (i.e., fixed duration and start time of disruptive event, known quantity of damaged equipment). Analysis of stochastic disruptive events can be conducted using developed simulation models as well, but is left as future research.  Disruptive events do not affect the original berthing plan, i.e. each vessel will be served at the berth initially assigned. Evaluation of the floaterm concept for scenarios with a revised vessel berth plan is left as future research. 4. Computational experiments The goal of the computational experiments was to evaluate productivity (makespan of vessel service and QC moves per hour) of the two terminal configurations under normal and disruptive operating conditions. Twenty-four scenarios (shown in Table 1) were developed to model both CMT and FMT under normal operating conditions considering various: (a) container compositions, (b) numbers of on-shore QCs, and (c) numbers of floating QCs at FMT. Sixteen additional scenarios (shown in Table 2) were developed to analyze performance of both terminal types under disruptive conditions with various: (a) container compositions, (b) numbers of floating QCs at FMT, and (c) quantities of damaged equipment. Completion time for all vessel handling (i.e., makespan) was selected as the simulation stopping criterion which may result in final states of the simulation models under disruptive events that differ from normal operating conditions (i.e., import containers may be stored at the floating yard, when vessel service is completed). However, under disruptive events vessel completion time is the critical component of terminal operations, and as such, the selected stopping criterion does not limit the validity of the research and results, presented in this study. In most cases, the terminal operator will utilize available resources to move import containers, from the floating to the storage yard blocks, during low demand periods and once operations are back (or close) to normal. Ten replications for each scenario were used to estimate average values of the various performance measures (presented next). The number of replications was found to be sufficient, as the average standard deviation over all scenarios was less than 0.5% of the mean [48]. Simulation speeds averaged 170 min/s. Depending on the models’ complexity, the simulation

27

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35 Table 1 Scenario analysis under normal operational conditions. Scenario (1)

Terminal type (2)

ITV type (3)

Transshipment (% of total volume) (4)

#QCs (5)

#On-shore QCs (6)

#Off-shore QCs (7)

# ITV (8)

# GCs (9)

S_1 S_2 S_3 S_4 S_5 S_6 S_7 S_8 S_9 S_10

CMT

YT

33.3 50.0 40.0 33.3 50.0 33.3 50.0 40.0 33.3 50.0

3 4 5 6

3 4 5 6 6 3 4 5 6 6

0 0 0 0 0 0 0 0 0 0

8 10 13 15 16 7 8 10 12 13

13 15 19 21 23 11 14 17 19 20

S_11 S_12 S_13 S_14 S_15 S_16 S_17 S_18 S_19 S_20 S_21 S_22 S_23 S_24

FMT

33.3 25.0 50.0 40.0 60.0 33.3 50.0 33.3 25.0 50.0 40.0 60.0 33.3 50.0

3 4

2 3 2 3 2 4 3 2 3 2 3 2 4 3

1 1 2 2 3 2 3 1 1 2 2 3 2 3

5 6 4 6 4 8 6 3 5 3 5 3 7 5

6 12 6 12 6 14 12 4 11 4 11 4 12 11

ALV

YT

ALV

3 4 5 6

5 6 3 4 5 6

Table 2 Scenario analysis under disruptive operational conditions. Scenario (1)

Terminal type (2)

ITV type (3)

Disruption (4)

Transshipment (% of total volume) (5)

#On-shore QCs (6)

#Off-shore QCs (7)

# ITV (8)

# GCs (9)

S_1⁄

CMT

YT

No disruption No disruption A A B B No disruption No disruption A A B B

33.3

6

0

15

21

50.0

6

0

16

23

33.3 50.0 33.3 50 33.3

4 4 3 3 6

0 0 0 0 0

15 16 15 16 12

14 15 10 11 19

50

6

0

13

20

33.3 50 33.3 50

4 4 3 3

0 0 0 0

12 13 12 13

12 13 9 10

No disruption No disruption A A B B No disruption No disruption A A B B

33.3

4

2

8

14

50

3

3

6

12

33.3 50 33.3 50 33.3

2 1 1 0 4

2 3 2 3 2

8 6 8 6 7

10 8 7 6 12

50

3

3

5

11

33.3 50 33.3 50

2 1 1 0

2 3 2 3

7 5 7 5

8 7 6 5

⁄

S_2

⁄

S_3 S_4⁄ S_5⁄ S_6⁄ S_7⁄

ALV

⁄

S_8

S_9⁄ S_10⁄ S_11⁄ S_12⁄ S_13⁄ S_14

FMT

YT

⁄

S_15⁄ S_16⁄ S_17⁄ S_18⁄ S_19⁄ S_20⁄ S_21⁄ S_22⁄ S_23⁄ S_24⁄

ALV

28

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

software package used in this study (i.e., FlexSim) allows for speeds up to 200,000 time units/s. The fact that the model speeds are low indicates high complexity. 4.1. Numerical data 4.1.1. Normal operating conditions Data for each one of the twenty-four scenarios used under normal operating conditions are shown in Table 1, where columns one through nine show: (1) scenario number, (2) terminal type, (3) ITV type, (4) percentage of transshipment containers, (5) total number of QCs, (6) number of on-shore QCs at each berth, (7) number of floating QCs at each berth, (8) number of ITVs for each gang, and (9) number of GCs at the storage area. For example, in the second scenario (S_2) four QCs (all located on-shore), ten YTs, and fifteen GCs are assigned to serve each vessel at each berth of CMT. The total demand for each vessel is 12,000 TEUs with an equal split between import and export containers. The quantity of transshipment containers varies by scenario (Table 1, column 4). For instance, in the first scenario (S_1) 4000 import and 2000 transshipment containers are unloaded from each vessel, and 4000 export and 2000 transshipment containers are loaded to each vessel. In the second scenario (S_2) 3000 import and 3000 transshipment containers are unloaded from each vessel, and 3000 export and 3000 transshipment containers are loaded to each vessel. The export/import split can be altered in the model (e.g., 70–30% instead of 50–50%), which shall change the amount of necessary GCs for serving yard blocks with export/import containers. The user may also assign different container compositions for each vessel, which will change the allocation of equipment at FMT (e.g., vessels with higher transshipment demand will probably require more floating QCs and vice versa). Note that sizes of ITV gangs and GC groups, required to obtain the optimal QCP, were determined based on simulation runs. An example of the procedure for estimating the necessary numbers of ITVs and GCs is presented in Fig. 6 for the case of 3 QCs at both CMT and FMT. Each graph provides the following information: (a) number of GCs (x-axis), (b) number of ITVs (y-axis), (c) obtained QCP (z-axis), (d) scenario number (top right edge), (e) optimal ITV and GC combination (depicted in the top left edge and labeled by ‘‘ ’’). For instance, in the first scenario 8 YTs and 13 GCs provided the optimal QCP = 32.66 moves per hour at CMT with 3 on-shore QCs (see S_1). Similar analysis was conducted for each scenario (see Table 1). It was found that on average CMT required 2.61 YTs per QC, 2.11 ALVs per QC, 3.89 GCs per QC for models with YT deployment, 3.44 GCs per QC for models with ALV deployment. As for FMT, 2.05 YTs per QC, 1.63 ALVs per QC, 3.58 GCs per QC for models with YT deployment, 3.00 GCs per QC for models with ALV deployment were required to obtain the optimal QCP. Thus, on average under normal operating conditions FMT required 21.4% less YTs, 22.7% less ALVs, 8.0% less GCs for models with YT deployment, and 12.9% less GCs for models with ALV deployment (savings are presented per QC). 4.1.2. Disruptive operating conditions Scenarios, used for productivity analysis of CMT and FMT under disruptive scenarios, are presented in Table 2, where columns one through nine show: (1) scenario number, (2) terminal type, (3) ITV type, (4) disruption, (5) percentage of transshipment containers, (6) number of operational on-shore QCs at each berth, (7) number of operational off-shore QCs at each berth, (8) number of ITVs for each gang, and (9) number of GCs at the storage area. Scenarios labeled as ‘‘No disruption’’ (e.g., S_1⁄, S_2⁄, S_7⁄, etc.) are identical to the ones used for modeling normal operating conditions (see Table 1, scenarios S_4, S_5, and S_9 respectively). Note that Table 2 presents the quantity of equipment, operational without breakdowns. For example, in the third scenario (S_3⁄) four on-shore QCs (two QCs are damaged and become available after 12 h), zero off-shore QCs, fifteen YTs, and fourteen GCs (seven GCs are damaged and become available after 12 h) are assigned to serve a vessel at each berth. The total demand for each vessel is 12,000 TEUs with an equal split between import and export containers as with normal conditions. The quantity of transshipment containers varies by scenario (Table 2, column 5). For instance, in the first scenario (S_1⁄) 4000 import and 2000 transshipment containers are unloaded from each vessel, and 4000 export and 2000 transshipment containers are loaded to each vessel. In the second scenario (S_2⁄) 3000 import and 3000 transshipment containers are unloaded from each vessel, and 3000 export and 3000 transshipment containers are loaded to each vessel. Next we present an analysis of the simulation results. 4.2. Makespan analysis The vessel service time makespan was chosen as the first performance measure to compare CMT and FMT operations under normal and disruptive conditions. 4.2.1. Normal operating conditions Fig. 7 presents the vessel service time makespan under normal operating conditions for all 24 scenarios. The x-axis label has three components: (a) scenario, (b) number of on-shore and off-shore QCs, and (c) percentage of transshipment containers. For example, in the upper left graph of Fig. 7 the first bar shows the makespan (122.5 h.) at CMT with YT deployment, where 3 on-shore and zero off-shore QCs serve each vessel, where transshipment containers are equal to 33.3% of the total demand (scenario S_1). FMT provided faster vessel service for all scenarios and on average, FMT makespan savings comprised 7.6 h. (or 9.5%) for YT deployment models and 0.5 h. (or 0.6%) for ALV deployment models. ALV deployment models outperformed YT deployment models in terms of makespan. However, FMT makespan savings were not substantial for cases

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

Fig. 6. Procedure for estimating quantity of required ITVs and GCs.

Fig. 7. Makespan under normal operational conditions by terminal type, ITV configuration, and transshipment volumes.

29

30

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

when ALVs were employed as ITVs. The latter can be explained by the fact that ALVs are more productive than YTs, and were able to provide more efficient container handling at both CMT and FMT under normal operating conditions. FMT configuration also provided faster vessel service (than CMT) with less equipment for scenarios with higher transshipment volumes (see S_13, S_15, S_20, and S_22). 4.2.2. Disruptive operating conditions Several researchers quantified resilience of MCTs based on the difference in terminal productivity (e.g., vessel service time makespan) before and after disruptive events [44–46]. In this paper vessel service completion makespan was selected as the key performance measure to assess effects of the disruptive events. Fig. 8 presents the vessel service time makespan for all 24 scenarios. The x-axis label has two components: (a) ITV type, and (b) percentage of transshipments. For example, YT33.3% refers to the simulation model with YT deployment and 33.3% of all TEUs handled being transshipment containers. ALV deployment models outperformed YT deployment models in terms of makespan. For all scenarios CMT was affected more by the disruptive events (i.e., higher makespan). CMT YT deployment models were the most vulnerable to disruptive events with a makespan increase, as compared to normal operating conditions, averaging 10.6 h. (or 17.3%) and 19.1 h. (or 31.2%) for disruptions A and B respectively. CMT ALV deployment models were less affected by the disruptions with a makespan increase, as compared to normal operating conditions, averaging 7.1 h. (or 13.4%) and 13.7 h. (or 26.0%) for disruptions A and B respectively. FMT YT and ALV deployment models resulted in similar makespan increase (as compared to normal operating conditions), averaging 5.8 h. (or 10.8%) and 12.5 h. (or 23.4%) for disruptions A and B respectively. The latter results may be explained by the fewer number of ITVs and GCs used at the FMT model in scenarios with ALV deployment. For both disruptions (A and B) the FMT with YT deployment model provided substantially higher makespan savings when compared to ALV deployment, while scenarios with lower transshipment percentages showed smaller improvements. 4.3. QCP analysis QCP (on- and off-shore combined) was selected as the second performance measure of CMT and FMT operations under normal and disruptive conditions. QCP is important to terminal operators as their agreements with liner shipping companies usually contain a clause on container handling rates. 4.3.1. Normal operating conditions Fig. 9 shows QCP (moves per hour by QC) for all 24 scenarios, under normal operating conditions at both terminals. Labels on the x-axis show: (a) scenario, (b) number of on- and off-shore QC available at each berth, and (c) percentage of transshipment containers assigned to each vessel. For instance, in the upper left graph of Fig. 9 the first bar presents QCP (32.7 moves/ h) at CMT with YT deployment, where 3 on-shore and zero off-shore QCs serve each vessel with 33.3% of transshipment containers (scenario S_1). On average, CMT provided QCP of 32.6 moves/h for YT deployment models and 37.9 moves/h for ALV deployment models. As for FMT, the average QCP was 35.9 moves/h and 38.2 moves/h for YT and ALV deployment models respectively. 4.3.2. Disruptive operating conditions Fig. 10 presents QCP (moves per hour by QC) for all 24 scenarios, under disruptive operating conditions at both terminals. Labels on the x-axis show: (a) scenario, (b) number of on- and off-shore QC available at each berth, and (c) percentage of transshipment containers assigned to each vessel. For example, in the upper left graph of Fig. 10 the first bar denotes QCP (32.6 moves/h) at CMT with YT deployment, where 6 on-shore and zero off-shore QCs serve each vessel with 33.3% of transshipment containers (scenario S_1⁄). FMT exhibits significantly higher QCP for the cases, where YTs are employed.

Fig. 8. Makespan under normal and disruptive operational conditions by terminal type, ITV configuration, and transshipment volumes.

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

31

Fig. 9. QCP under normal operational conditions by terminal type, ITV configuration, and transshipment volumes.

The same trend does not apply to ALV deployment models, when similar QCPs were obtained at CMT and FMT. This can be explained by the fact that ALVs are more productive than YTs, and were able to provide more efficient container handling at both CMT and FMT under normal and disruptive conditions. 4.4. Storage of import containers at the floating yard Another performance measure, quantified by this study, was the volume of import containers unloaded to feeder barges and stored at the floating yard during the disruptive event. The amount of imports, stored on barges at the floating yard, was recorded at each simulation run and average values are presented in Fig. 11. Labels on the x-axis show: (a) scenario, (b) ITV type, and (c) percentage of transshipments. On average 49.0% more import containers were placed on barges under disruption B as compared to A. This can be expected as on-shore QCs are out of service for a longer time period during the former disruptive event, utilization of floating QCs for handling imports increases. Approximately 12% less import containers were stored at the floating yard for ALV deployment models as compared to YT deployment models, since more containers could be processed by QCs and GCs (ALVs do not have to wait for the container to be (un)loaded). Note that the strategy of storing imports at the floating yard was crucial for improving FMT productivity, as otherwise idling time of floating QCs along with makespan would substantially increase. 4.5. Economic analysis A 20-year cost analysis for CMT and FMT was conducted and based on the estimation of initial investments and operational costs for both systems. Investment costs included site development (e.g., clean and grub, civil site works, wharf construction, site electrical, yard lightening, gate site work, gate facility, maintenance and administration buildings, etc.) and equipment costs (i.e., on-shore QCs, off-shore QCs, ITVs, GCs, crane barges, and container barges). Site development costs for two terminal configurations were computed using guidelines, provided by Wilbur Smith Associates [49]. Equipment costs were calculated using brochures, released by manufacturers (Shanghai Zhenhua Heavy Industry Co., CNBM International Engineering Co., Kalmar Industries, etc.). Operational costs covered maintenance, insurance, QC gangs, ITV gangs, GC gangs, and push boat operators [12]. Note that investment and operational costs vary from terminal to terminal. The amount of necessary equipment, estimated by simulation models, was used as input for calculating associated costs (see Table 1).

32

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

Fig. 10. QCP under normal and disruptive operational conditions by terminal type, ITV configuration, and transshipment volumes.

Fig. 11. Import containers stored on barges during disruptions.

Results of the economic analysis are presented for 24 considered scenarios (only normal operating conditions) in Fig. 12. The total costs were estimated per berth of CMT or FMT for a given quantity of QCs. For instance, the blue4 line with abbreviation CMT_YT(33.3) at the top left graph of Fig. 12 indicates that the total expenses (including investment and operational costs) for CMT with YT deployment and 33.3% of transshipment comprise $454.3 million at the end of the considered time horizon (i.e., at year 20). Results of the economic analysis indicate that ALV deployment models have a higher capital investment but lower operational costs than YT deployment models. Pay back periods for ALV deployment models did not exceed 2 years, assuming 80% QC utilization (7008 operational h/year) for both FMT and CMT. CMT with YT deployment was found to be the most expensive alternative for all scenarios. FMT site development costs were lower than CMT site development costs (mainly due to 4

For interpretation of color in Fig. 12, the reader is referred to the web version of this article.

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

33

Fig. 12. Economic analysis.

larger size of the storage yard), but higher equipment investment costs (mainly due to the cost of a crane barge at FMT, which could comprises 10–12 million USD). FMT average savings over 20-year horizon comprised $66.7 million for YT deployment models and $15.0 million for ALV deployment models. FMT operational costs were lower than CMT operational costs for cases with high percentage of transshipment containers. Advantages of FMT over CMT substantially decreased with the demand reduction for transshipment containers. Note that FMT might be a cost effective alternative at import/export MCTs even under normal operating conditions, when the terminal operator is not able to increase capacity of the storage yard to serve incoming volumes of import containers (e.g., land restriction, substantial costs for constructing new yard blocks and purchase of equipment, etc.). Along with the container composition, there are several other factors that should be taken into account to assess benefits of FMT adequately, e.g. terminal layout, technical characteristics of equipment, QC utilization, geometrical characteristics of vessels to be served, features of potential disruptive events, preferences of terminal operators and liner shipping companies. Additional simulation and cost analyses are recommended to decide on implementation of the floaterm concept at the given MCT. 5. Conclusions and future research avenues In this paper the floaterm concept was evaluated as means to increase productivity of MCT operations and improve their resilience. Two simulation models were developed to compare performance of a conventional marine container terminal to one that has adopted the floaterm concept under normal and disrupted operating conditions. From the analysis significant

34

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

savings in the makespan of vessel completion time were observed under both operating conditions for FMT as compared to CMT. Results also suggested that benefits of the floaterm concept increased with transshipment volumes. The latter observation should be expected as the main purpose of the floaterm concept is to relieve landside operations from handling of transshipments containers, while at the same time act as a buffer storage area for import containers, when disruptive events limit the (un)loading capacity of on-shore QCs. Finally, the floaterm concept improved QC productivity under YT deployment, but did not exhibit any significant QC productivity improvement under ALVs were employed as ITVs. Research outcomes indicated that FMT demonstrated substantial cost and vessel service makespan savings for scenarios with YT deployment. Although FMT with ALV deployment did not significantly outperform CMT in terms of vessel service makespan, significant cost savings were observed for the majority of the cases evaluated. Even though simulation, as a modeling tool, offers a number of advantages, the models developed herein inherit a number of limitations common amongst marine container terminal simulation models found in the literature. These limitations include: (a) capturing ITV interference [30], (b) implementing optimal ITV deployment strategies, (c) accounting for terminal congestion, and (d) modeling different storage yard strategies and areas for hazmat, overweight, oversized, and refrigerated containers (a.k.a. refeer). These drawbacks can be addressed as part of future research, and do not reduce the validity of the research outcomes presented herein. Addressing these limitations will most likely increase the estimated benefits of the floaterm concept under normal and disruptive operating conditions. Appendix A List of abbreviations: AGV ALV CMT DT FMT GA GC ITV MCT QC QCP QCR RMG SC YT

Automated guided vehicle Automated lifting vehicle Conventional marine terminal Drayage truck Floaterm marine terminal Genetic algorithm Gantry crane Internal transport vehicle Marine container terminal Quay crane QC productivity QC rate Rail mounted gantry crane Straddle carrier Yard truck

References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18]

UNCTAD, Review of Maritime Transport 2013, United Nations Conference on Trade and Development, New York and Geneva, 2013. N. Umang, M. Bierlaire, I. Vacca, The berth allocation problem in dry ports, Ecole Polytechnique Federale de Lausanne, 2011. The Journal of Commerce, Mega-ships, , (accessed 02.08.13). M.C. Mourão, M.V. Pato, A.C. Paixão, Ship assignment with hub and spoke constraints, Marit. Policy Manage. 29 (2002) 135–150. R. McCalla, Global change, local pain: intermodal seaport terminals and their service areas, J. Transp. Geogr. 7 (1999) 247–254. A. Ballis, J. Golias, C. Abakoumkin, A comparison between conventional and advanced handling systems for low volume container maritime terminals, Marit. Policy Manage. 24 (1997) 73–92. L. Henesey, Enhancing Container Terminal Performance: The Multi Agent Systems Approach, Licentiate thesis, Blekinge Institute of Technology, Karlshamn, 2004. A. Paixão, P. Marlow, Fourth generation ports – a question of agility, Int. J. Phys. Distrib. Logist. Manage. 33 (2003) 355–376. A. Ashar, Long-term trends in container shipping – the revised Forth Revolution, 2012, , (accessed 14.02.13). Liftech, Inc., The floaterm concept: reducing terminal congestion with waterside cranes, AAPA Seaports Magazine, Fall 2007. Port of Amsterdam, Port Statistics, , (accessed 07.07.14). B. Pielage, J. Rijsenbrij, W. Van den Bosch, H. Ligteringen, J. Van Beemen, Floating Cranes for Container Handling, Port Research Center, Rotterdam, Delft, 2008. D. Steenken, S. Vos, R. Stahlbock, Container terminal operation and operations research – a classification and literature review, OR Spectrum 26 (2004) 3–49. R. Stahlbock, S. Vos, Operation research at container terminals: a literature update, OR Spectrum 30 (2008) 1–52. S. Theofanis, M. Boile, M. Golias, Container Terminal Berth Planning, in: Transportation Research Record: Journal of the Transportation Research Board, No. 2100, Transportation Research Board of the National Academics, Washington, D.C., 2009. C. Bierwirth, F. Meisel, A survey of berth allocation and quay crane scheduling problems in container terminals, Eur. J. Oper. Res. 202 (2010) 615–627. H. Carlo, I. Vis, K. Roddbergen, Seaside operations in container terminals: literature overview, trends, and research directions, J. Prod. Res. 49 (2013) 6199–6226. X. Han, Z. Lu, L. Xi, A proactive approach for simultaneous berth and quay crane scheduling problem with stochastic arrival and handling time, Eur. J. Oper. Res. 207 (2010) 1327–1340.

M.A. Dulebenets et al. / Simulation Modelling Practice and Theory 54 (2015) 19–35

35

[19] C. Arango, P. Cortes, J. Munuzuri, L. Onieva, Berth allocation planning in Seville inland port by simulation and optimization, Adv. Eng. Inform. 25 (2011) 452–461. [20] X. Qi, D. Song, Minimizing fuel emissions by optimizing vessel schedules in liner shipping with uncertain port times, Transp. Res. Part E 48 (2012) 863– 880. [21] W. Aye, M. Low, H. Ying, H. Jing, L. Fan, Z. Min, Visualization and simulation tool for automated stowage plan generation system, School of Computer Engineering, Nanyang Technological University, 2010. [22] M. Low, X. Xiao, F. Liu, S. Huang, W. Hsu, Z. Li, An Automated Stowage Planning System for Large Container ships, School of Computer Engineering, Nanyang Technological University, 2010. [23] Z. Min, M. Low, W. Hsu, S. Huang, F. Liu, W. Aye, Improving Ship Stability in Automated Stowage Planning for Large Containerships, School of Computer Engineering, Nanyang Technological University, 2010. [24] X. Xiao, M. Low, F. Liu, S. Huang, W. Hsu, Z. Li, An efficient block-based heuristic method for stowage planning of large containerships with crane split consideration, School of Computer Engineering, Nanyang Technological University, 2010. [25] P. Canonaco, P. Legato, R. Mazza, R. Musmanno, A queuing network model for the management of berth crane operations, Comput. Oper. Res. (2008) 2432–2446. [26] Y. Saanen, M. Valkengoed, Comparison of three automated stacking alternatives by means of simulation, in: Proceedings for the 2005 Winter Simulation Conference, 2005. [27] X. Guo, S. Huang, W. Hsu, M. Low, Simulation-based dynamic partitioning of yard crane workload for container terminal operations, School of Computer Engineering, Nanyang Technological University, 2008. [28] M. Petering, K. Murty, Effect of block length and yard crane deployment systems on overall performance at a seaport container transshipment terminal, Comput. Oper. Res. 156 (2008) 1–15. [29] M. Petering, Effect of block width and storage yard layout on marine container terminal performance, Transp. Res. Part E 45 (2009) 591–610. [30] M. Petering, Y. Wu, W. Li, M. Goh, R. Souza, Development and simulation analysis of real-time yard crane control systems for seaport container transshipment terminals, OR Spectrum 31 (2009) 801–835. [31] M. Petering, Decision support for yard capacity, fleet composition, truck substitutability, and scalability issues at seaport container terminals, Transp. Res. Part E 47 (2011) 85–103. [32] Y. Cheng, H. Sen, K. Natarajan, C. Teo, C. Tan, Dispatching Automated Guided Vehicles in a Container Terminal, in: Singapore-MIT Alliance Program, National University of Singapore, 2003. [33] C. Yang, Y. Choi, T. Ha, Simulation-based performance evaluation of transport vehicles at automated container terminals, OR Spectrum 26 (2004) 149– 170. [34] I. Vis, I. Harika, Comparison of vehicle types at an automated container terminal, OR Spectrum 26 (2004) 117–143. [35] S. Esmer, I. Cetin, O. Tuna, A simulation for optimum terminal truck number in a Turkish port based on lean and green concept, Asian J. Shipp. Logist. 2 (2010) 277–296. [36] A. Garro, W. Russo, EASYABMS: a domain-expert oriented methodology for agent-based modeling and simulation, Simul. Model. Pract. Theory 18 (2010) 1453–1467. [37] M. Petering, Development and simulation analysis of real-time, dual-load yard truck control systems for seaport container transshipment terminals, OR Spectrum 32 (2010) 633–661. [38] M. Francesco, M. Lai, P. Zuddas, Maritime repositioning of empty containers under uncertain port disruptions, Comput. Ind. Eng. 64 (2013) 827–837. [39] FlexSim, FlexSim Simulation Software, , (accessed 02.07.14). [40] M. Dulebents, M. Golias, W. Heaslet, Evaluation of the floaterm concept via simulation, in: 93rd Annual Meeting Transportation Research Board, 2013. [41] L. Song, T. Cherrett, W. Guan, Study on berth planning problem in a container seaport: using an integrated programming approach, Comput. Ind. Eng. 62 (2012) 119–128. [42] N. Park, B. Dragovic, J. Kim, Dynamic equipment deployment at a container terminal: transfer system based on real-time positioning, J. Mech. Eng. 55 (2009) 83–94. [43] Q. Zeng, Z. Yang, L. Lai, Models and algorithms for multi-crane oriented scheduling method in container terminals, Transp. Policy 16 (2009) 271–278. [44] N. Mohseni, Developing a Tool for Designing a Container Terminal Yard, Master Thesis Project, Delft University. [45] H. Gajjar, T. Wakeman, H. Saloum, Development of Economic Resiliency Assessment Framework for Marine terminals, Stevens Institute of Technology, 2008. [46] A. Rose, D. Wei, Measuring total economic benefits of USCG marine safety programs, Create Homeland Security System, 2010. [47] K. Barker, R. Pant, H. Baroud, T. Landers, Interdependent, multi-regional impacts of inoperability at inland waterway ports and network, in: Maritime Risk Symposium, Piscataway, New Jersey, 2011. [48] A. Pritsker, C. Pegden, Introduction to Simulation and SLAM, John Wiley & Sons Inc, New York, 1979. [49] Wilbur Smith Associates, Inc. Latin American Trade & Transportation Study, , (accessed 26.11.13).