845

1 INTRODUCTION

The laboriousness (complexity) of container handling

operations is usually evaluated by the average

number of moves required for their performing. In

order to calculate the commercial and operational

characteristics one also needs to consider the time

needed to implement each movement [1-4]. When the

analytical approach is used, one cannot think in terms

of single operations, so the mean time of moves is

usually considered instead. Still, the number of moves

is not deterministic but stochastic value and should be

described by distribution functions. In order to get

ones, e.g. distribution density, which is important

parameter for container terminal design and

operational management, usually methods of Monte-

Carlo group are used [5]. These methods, actually, can

be considered as an attempt to introduce individual

values of different parameters which are defined by

the integral functions of each characteristic. In a way,

these methods of random generation of values are

directly reversed to the generalization provided the

step before.

In many cases this approach can be considered as

oversimplified, i.e. when a terminal exploits several

different types of container handling equipment for

stacking operations. Another problem appears when

Planning Simulation Experiments in the Tasks of

Studying the Operational Strategies of Container

Terminals

A.L. Kuznetsov, A.V. Kirichenko & A.D. Semyonov

Admiral Makarov State University of Maritime and Inland Shipping, Saint-Petersburg, Russia

H. Oja

Konecranes Finland Corp., Hyvinkää, Finland

ABSTRACT: There is a lot of scientific papers that consider the efficiency of different container yard’s handling

strategies. The methods of assessment vary from abstract mathematical calculations to aposterioric collection

and analyses of practical data. This article deals with a new way of stacking strategy’s estimation. It postulates

that investigations based on the extrapolation of existing strategies cannot be reliable due to rapid changes in

terminals’ environment and restricted amount of data that can be collected within limited periods of the

strategies’ application. The hypothesis of the study is that the only method that could provide robust

comparative study of container stacking strategies is the simulation modelling. In the same time, any model is

only a tool of analyses. The synthesis could be provided only by massive iterations of single simulation

experiments with controlled searching procedure in the space of parameters. Consequently, the stress of the

simulation study should be put not on the model itself, but on the way how to use it, i.e. on the experiments

planning. Only a rationally constructed machine of simulation scenario generation could provide adequate and

statistically reliable results. In order to demonstrate the importance of this ‘task setting’ machine, two example

strategies are selected for comparison: (1) allocation of containers to slots with the minimum height and (2)

stacking of containers in accordance with expected dwell time. It also shows that the objective function has a

great impact on the optimal strategy implementation. In the study described in the paper the number of moves

requested to select a container is opted as an optimization criterion.

http://www.transnav.eu

the International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 14

Number 4

December 2020

DOI: 10.12716/1001.14.04.08

846

one wishes to study technological and operational

changes which can manifest themselves and affect the

terminals’ efficiency only in a long-termed

perspective.

Generally speaking, the operational strategy of the

cargoflow handling consists of few particular tactics.

From mathematical point of view, most of these tactics

should be treated as heuristics. The “heuristics” is

used in operational research theory to describe

methods which are considered to be useful and

effective, but these features cannot be proved

mathematically [6-7]. As the result, their right for

existing is only proved by practice and/or simulation

modelling.

The heuristics constituting the operational strategy

are usually relevant to different interconnected parts

of general problem, that is why their individual

influence on the quality of operational decisions very

is difficult to separate and assess. The complex

interaction of different elements and unknown

sensitivity of the results to stochastic fluctuations of

the parameters cause high volatility of the final

evaluation. A single experiment cannot neither prove

any conclusions nor reveal the effects of such "weak”

influences. One example of these influences is the

‘operational temperature’ of a box in the stack. This

particular factor is selected in the paper to unfold the

suggested method of simulation experiments’

planning.

2 THE GOAL SETTING OА PARTIAL STRATEGY

STUDY

As it was mentioned above, the operational strategies

of container terminal are usually formed from the

number of heuristics and thus based on intuitive

perceptions. One of the most famous examples of this

perception is commonly practiced segregation of

containers on the “hot” and “cold” ones. This

segregation is based on the consideration of

containers dwell time.

In the practice of container yard operations, the

“hot” containers are the ones that dwelled in the stack

long enough to be selected from the stack soon.

Otherwise, recently arrived containers are the “cold”

ones and expected to dwell there for some time.

“Cold” containers arrive after “hot” and could block

the access to the “hot” ones, so they have to be shifted

to other positions in order to clear a target “hot”

container. An example of a stack where “cold” and

“hot” containers marked by different colors reflecting

their current dwell time is represented by Fig. 1.

Figure 1. An example of graphical representation of

containers’ status

When a request for container selection occurs, the

target container can be blocked by other ones.

Traditionally, the blocking container is replaced to

closest position with minimum height [8-9]. From the

other hand, the higher the temperature of blocking

container, the higher is the probability that it will soon

leave the terminal. Consequently, “hot” containers

could be moved to higher positions, where they

would not be blocked by other containers. At the same

time, it can take a longer time to place a container to a

position chosen by “operational temperature” then to

closest one with minimal height, since it could require

longer travel distance. This could lead to the decrease

of the positive effect gained from the number of

moves reduction due to increase of mean move time.

The results can be proved or denied only by very

large numbers of experiments. However, envisaging

the subtle character of this effect, these experiments

must be implemented simultaneously for each variant

of container stacking strategy in absolutely identical

operational conditions.

Implementing this or that technological

innovations in the operational practice of a real

terminal, we will not be able to make any evidential

conclusions: the registered changes of parameters

could be influenced by many controllable and

uncontrollable factors, including pure statistical errors

caused by limited data base of real world

experiments. If, for example, we would read over

every arriving container surah of Quran or sprinkle

with Holy Water or apply semantical networks for

improving the allocation, to evaluate the effect of

these techniques by the operational data in the end of

the year will be impossible. In order to do so we had

to gain the parallel results from these three hypothetic

variants and the forth basic variant, where we would

operate without proposed measures, and do it in

absolutely identical external operational environment

(mainly on the same patterns of external transport

flows). Moreover, the closer would be the monitoring

results, the longer period of observation it will take.

Obviously, the practical experiments with real

terminals are inacceptable, both for ‘singularity’ of the

results and thus low statistical reliability, and simply

for costs of these experiments. On the other hand,

equally inapplicable turn out to be also common

simulation techniques.

The methodological problem responsible for this

contradiction is that every task of the study of a

partial strategy requires the development and

implementation of specific simulation models [10].

847

These partial models should be synchronized and

coordinated by input data, together creating a unified

basis for adequate and efficient study of various

operational strategies. This paper offers a variant to

solve this problem.

3 THE MODEL OF SCENARIO GENERATION

The import containers, arriving by sea and departing

by land, move opposite to export containers, arriving

by land and departing by sea. In both directions this

movement is not continuous and ceaseless; it assumes

delays of containers on the terminal (dwelling or

storage) for certain time (Fig. 2).

Figure 3. Graphical representation of container flows and

storage

Every crossing of the terminal perimeter by cargo

flows shown by this figure assumes certain physical

handling of containers, mainly discharging from

arriving and loading on the departing vehicles. The

relevant facilities constitute the dedicated cargo front.

Depending on the capacity of the vehicles handled at

this front, the discharging and loading procedures

deal with different sizes of compact cargo parties. For

the automobile transport they encounter couples of

containers, for rail transport – hundreds and for sea

transport – several thousands

Obviously, all containers arrived to the terminal

sooner or later leave it, so the algebraic sum of all

inbound flows taken with the sign plus and all

outbound ones with the sign minus is zero. In the

same time, the typical cargo parties of different

transport modes cause different “packs” of containers

in inbound and outbound flows. For example, an

ocean ship delivered several thousand containers to a

port should be discharged and charged as soon as

possible to minimize its idle operation time, i.e. cargo

operation time for port operators and ship turnaround

time for ship owners.

A cargo party of a ship could consist of hundreds

railroad train parties, receiving and dispatching of

which should be done so that the railroad routes’

capacity is not exceeded. Similarly, the truck traffic

should be spread evenly in order to meet the

throughput capacity of the connecting road network.

The unevenness of the inbound and outbound

flows caused by the different sizes of cargo parties

and requirements to handle them in different periods

of time necessitates the existence of certain ‘buffer

stock’ of containers at the terminal. The physical

storage of these containers is performed by the

container yard of the terminal. The container yard is

characterized by its capacity, handling technology and

equipment parameters, which are defined by the

characteristics of container flows.

These dependences reveal themselves both directly

and indirectly, through many internal ties (e.g. by the

limited size of the technologic resources shared for

different operations). Their specific interdependences

make analytical (algebraic) techniques inadequate to

provide the accuracy needed for the design and

planning activity, thus forcing to use mathematical

modeling, particularly simulation.

The general structure of the model complex

(environment) designated to serve as a tool of the sea

port operation analyses is given by Fig. 3.

Figure 3. General structure of the sea terminal model

environment

The model shown by this figure consists of two

generators of cargo flows (sea side and land side), the

sea cargo front (SCF), the land cargo front (LCF) and

the container yard (CY).

The parameters of container flows include annual

volumes, the sizes of cargo parties and time

characteristics (unevenness and densities). The

parameters of cargo fronts describe their throughput

capacities expressed via the productivity and number

of technological equipment allocated for handling.

The parameters of the container yard include one-time

storage capacity and productivities of reception-

dispatching operations.

The features of the model, specifically the

characteristics of elements’ performance, should be

studied not qualitatively, but quantitatively, as well as

the time characteristics of the technologic processes

running inside it (utilization coefficients, delays and

refuses of servicing, queue lengths, waiting times etc.).

It is important to note that all input and output

parameters are not determined, but random values

and are represented by their a-priori and a-posteri

distribution functions.

Well known is the fact that the simulation is an

instrument of analyses and not of syntheses. Any

simulation enables to assess the qualitative

characteristics of the simulated object depending on

its input parameters’ values. Since all these

parameters are random values, it is necessary to run

serial experiments in numbers which would enable to

guarantee the required degree of the statistical

reliability.

Besides the quantitative parameters set by

deterministic and random values, the handling of

848

cargo flows passing through the terminal depends on

the different strategies of containers’ allocation on the

yard aiming on the reduction of selection time,

minimization of idle moves, and maximization of the

utilization of the technological resources. The

selection of the appropriate strategy also could be

done using dedicated logical reference machines. This

feature allows to use this simulation tool not only for

accessing the values of the technological parameters,

but for comparative study of operational strategies

and tactics.

When using the simulation model not as the

design tool, but as an instrument for operational

planning, the sea side and land side generators would

be replaced by real-time information of vehicles

arriving and leaving within the planning interval of

time, as well as the actual data on the state of the

technological processes running at the terminal’s

elements, i.e. cargo fronts and yards. In this case the

models serve as a tool for the sensitivity analyses

(‘what happens if …’) in order to seek for rational

distribution of the technological resources between

different operations.

4 RESULTS AND DISCUSSIONS

The single run of the model external generator shown

in Fig. 3 provides the detailed and full annual time

schedules for every transport mode connected in the

node. These time schedules contain the exact arrival

and departure times for every container passing

through the terminal within the simulation term.

Arriving containers should be allocated in the stack of

container yard, and in due moment (determined by

the generated time schedules) would leave the

terminal. As it was said above, in this paper the

investigated function of the operational strategy is the

shifting of blocking containers not in the lowest

positions (slots), but in spots determined taking into

account the operational temperature (the rest of dwell

time). The criterion for the estimation is the

laboriousness of selection. For this purpose, during

the simulation the total number and number of moves

per box are calculated. The computation is being done

by simulation of every physical motions and moves of

technologic equipment operating in the container

yard area.

The functional simulation model of container yard

operation in the structure of the modelling

environment on Fig. 3 is represented by the

rectangular in the center marked as ‘CY’. The arrow

denoting the model parameters includes both

geometrical characteristics of the container yard area

and technological features of the container handling

system, qualitative and quantitative. The

discriminating parameter of the investigated function

of the strategy are also included into this set: in one

case it is the minimal height strategy, in the other –

minimal operational temperature.

The difference in the number of movements

received as the investigated property on Fig. 3,

enables to make judgment on the advantage or

disadvantage of the strategy over random allocation.

In order to make the statistically reliable

conclusions every experiments were repeated many

times for different variants of container flows

structures and volumes. This enables to exclude any

statistical fluctuations and influence of different

patterns of external transport arrival.

Fig. 4 represents the results of the simulation set

undertaken for the study of the selected function of

operational strategy.

Рисунок 3. Результаты экспериментов

The basic theoretical complexity of selection is

computated by the simple combinatoric formula ,

the lines showing the complexity of minimal height

and minimal temperature strategies show the saving

of 0.5 moves per container in favor of the latter.

Though this gain turns out to be big enough in the

year term, to reveal this fact by any traditional

singular simulation would not be possible.

5 CONCLUSIONS

1 Basic components of container terminal

operational strategies are the heuristics, or the

methods whose efficiency cannot be proved

theoretically.

2 In order to include a heuristic into operational

strategy of container terminal, it is necessary to

prove its practical usefulness, which cannot be

done only by imitational simulation.

3 The construction of the scenarios of influencing

conditions needed to investigate a strategy

requires to take into account many real factors, like

volumes and structures of cargo flows, sizes of

cargo parties, irregularity of the traffic, throughput

capacity of elements etc.

4 In order to receive the statistically reliable results,

the experiments should be done in large numbers,

both for stochastically close versions and different

variants, all in the same environment and with

identical values.

5 The paper offers the methodic for controlled

generation of scenarios, turning the simulation tool

from the instrument of analyses into the

instrument of synthesis.

6 As an example, showing the efficiency of the

proposed technique the strategic function of

container allocation by the current rest of the dwell

time is selected.

849

REFERENCES

[1] Kuznetsov A.L., Kirichenko A.V., Semenov A.D.: Box

Selectivity in Different Container Cargo-handling

Systems. TransNav, the International Journal on Marine

Navigation and Safety of Sea Transportation, Vol. 13,

No. 4, doi:10.12716/1001.13.04.12, pp. 797-801, 2019

[2] Kuznetsov A.L., Kirichenko A.V., Semenov A.D.:

Evaluation of Sinking Effect in Container Stack.

TransNav, the International Journal on Marine

Navigation and Safety of Sea Transportation, Vol. 14,

No. 1, doi:10.12716/1001.14.01.19, pp. 159-162, 2020

[3] Kuznetsov K., Kirichenko A., Semenov A., Borevich A.,

Optimization Strategies of Container Terminals,

Scientific Journal of Gdynia Maritime University, No.

115, pp. 42-54, 2020

[4] Kuznetsov A., Kirichenko A., Izotov O. The Influence of

the Storage Strategy on the Complexity of the Container

Selection Procedure //IOP Conference Series: Earth and

Environmental Science. – 2018. – Vol. 171. – No. 1. – pp.

1755-1315.

[5] Hammersley J. Monte Carlo methods. – Springer Science

& Business Media, 2013.

[6] Sanen Y. A., Dekker R. Intelligent stacking as way out of

congested yards part 2 //Port Technol Int 2007. – 2007. –

Vol. 32. – pp. 80-85

[7] Borgman B. Online rules for container stacking / B.

Borgman, E. van Asperen, R. Dekker //OR spectrum. —

2010. — Vol. 32. — Is. 3. — pp. 687–716.

[8] Dekker R., Voogd P., Van Asperen E. Advanced methods

for container stacking //Container terminals and cargo

systems. – Springer, Berlin, Heidelberg, 2007. – pp. 131-

154.

[9] Böse, Jürgen W., ed. Handbook of terminal planning.

Vol. 49. Springer Science & Business Media, 2011

[10] Dragović B. Simulation modelling in ports and

container terminals: literature overview and analysis by

research field, application area and tool / B. Dragović, E.

Tzannatos, N. K. Park // Flexible Services and

Manufacturing Journal. — 2017. — Vol. 29. — Is. 1. —

pp. 4–34. DOI: 10.1007/s10696-016-9239-5.