Determining the Required Storage Capacity of an Import Dry Bulk Terminal

Delft University of Technology MATERIALS ENGINEERING _{Department Marine and Transport Technology} Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl

This report consists of 156 pages and 4 appendices. It may only be reproduced literally and as a whole. For commercial purposes only with written authorization of Delft University of Technology. Requests for consult are only taken into consideration under the condition that the applicant denies all legal rights on liabilities concerning the contents of the advice.

Specialization: Production Engineering and Logistics

Report number: 2012.PEL.7723

Title:

Determining the Required

Storage Capacity of an Import

Dry Bulk Terminal

Author:

Pieter Bot

Titel Het bepalen van de benodigde opslagcapaciteit van een import droge bulk terminal

Assignment: Master Thesis

Confidential: no

Initiator (university): ir. T.A. van Vianen Initiator (company): ir. J.A.J.M Dekkers

Supervisor: ir. T.A. van Vianen/ ir. R. van Duijn

Page | ii

Page | iii

Preface

This report is the written result of my graduation research of the study Mechanical Engineering, Section Transportation Engineering, at Delft University of Technology. This research has been conducted on behalf of the advisory group Heavy Industries & Logistics of Royal HaskoningDHV.

The subject of this research is to determine the required storage capacity of import dry bulk terminals with the use of a simulation model and an optimization algorithm. In this report, a DVD is attached, which contains the source code of the simulation model, a user guide for running the simulation model and all output files of the optimizations.

I would like to thank Teus van Vianen, Jan Dekkers, René van Duijn and prof. Lodewijks for their support during my research. Also, I would like to thank my colleagues at Royal HaskoningDHV who supported me with tips and advice.

Pieter Bot

Page | iv

Summary

Designing dry bulk terminals is one of the main activities conducted at Heavy Industries & Logistics, an advisory group at Royal HaskoningDHV. Due to the increasing demand of the dry bulk materials coal and iron ore, existing dry bulk terminals need to be expanded and new dry bulk terminals need to be designed. For this, the storage capacity is one of the main design parameters, as high investment costs are involved.

There are no clear guidelines available in literature about the required storage capacity of dry bulk terminals. Preliminary research of this study has shown that the annual throughput, which is generally used as the only guideline, does not have enough explanatory power. For these reasons, there is a need for a method that gains insight in the factors that determine the required storage capacity of a dry bulk terminal.

This focus of this research is to determine the required storage capacity of import dry bulk terminals that store their bulk materials identity-preserved. Due to stochastic influences, the storage level fluctuates at these types of terminals. For example, terminals have no or limited influence on the arrival times of ships. When ships arrive, service has to be provided within the pre-arranged time. In order to avoid all demurrage costs (and to handle the highest storage levels), a large stockyard is required. However, a large stockyard brings along high investment costs. Therefore, from an economical point of view, it might be better to accept some demurrage costs in order to have lower investment costs. In this research, it is assumed that the required storage capacity of an import dry bulk terminal is the storage capacity that involves the lowest costs; in other words, the storage capacity with the highest net present value (NPV).

First, in order to determine the required storage capacity, an extensive simulation model has been developed of an import dry bulk terminal in the discrete event simulation package TOMAS Delphi. In this simulation model is tried to capture all properties of the terminal as realistic as possible. The stochastic properties are included, like the inter-arrival times of ships (Erlang-2 distributed), the ship types, the ship capacities, and the storage time of the bulk materials on the stockyard (Erlang-2 distributed). In addition, all revenues and costs of the terminal are included in the terminal. As a result, the NPV is determined automatically for each simulation run. Moreover, it is possible for the user to adjust all parameter values (including the distributions of the stochastic variables), so many sorts of terminals can be configured.

Subsequently, the required storage capacity is determined for each configuration by optimizing the NPV of the simulation model: the required storage capacity is the storage capacity with the largest NVP. However, the most common gradient-based optimization algorithms are not suitable. When the NPV is optimized, two problems arise. On the one hand, the NPV is non-linear en contains thousands

Page | v of local maxima. On the other hand, the required storage capacity is influenced by the capacities of the quay cranes and the stacker-reclaimers. For these reasons, the “simultaneous perturbation stochastic approximation” (SPSA) algorithm with injected noise is implemented in the simulation model. This algorithm is able to efficiently optimize the NPV with respect to three variables simultaneously. Moreover, it can approximate the global maximum, as it contains a unique method to estimate the gradient.

The required storage capacity can be divided into two parts: a deterministic and a stochastic part. The deterministic part equals the average storage level and is determined by the multiplication of the annual throughput and the average time bulk materials spend on the terminal (Little’s Law). The stochastic part is the storage capacity a terminals needs due to its stochastic properties and can be expressed as a percentage of the annual throughput. The largest influences on the stochastic part of the required storage capacity are the annual throughput and the storage time of bulk materials.

In the base configuration, small terminals (10 Mton annual throughput) need 7.6% of the annual throughput as storage capacity to handle its stochastic properties. This is only 3.1% for large terminals (50 Mton annual throughput). A storage time of approximately two months is assumed in the base configuration. When the storage time decreases to one month, the stochastic part of the storage capacity decreases by 24%. On the other hand, an increase of the storage time to three months decreases the stochastic part of the storage capacity by 12%. Out of the stochastic variables, the inter-arrival times distribution has the largest impact on the stochastic part of the required storage capacity. The effects of the ship characteristics and irregular storage times are much smaller.

Therefore, it is concluded that the required storage capacity can be determined by developing a simulation model in which the NPV is optimized. This way, the required storage capacity is the storage capacity that has the largest NPV. The stochastic influences have a smaller effect on the required storage capacity as the annual throughput increases. Depending on the type of terminal, 3 to 10% of the annual throughput is required as extra storage capacity for the stochastic effects in addition to the average storage level.

Page | vi

Samenvatting

Het ontwerpen van droge bulk terminals is één van de hoofdactiviteiten die worden uitgevoerd bij Heavy Industry & Logistics, een adviesgroep van Royal HaskoniningDHV. Door de toenemende vraag naar de droge bulk materialen ijzererts en steenkool worden huidige droge bulk terminals uitgebreid en moeten nieuwe terminals worden ontworpen. Daarbij is de opslagcapaciteit één van de belangrijkste ontwerpparameters, aangezien hier zeer hoge investeringskosten mee gemoeid zijn.

Er zijn geen duidelijke richtlijnen beschikbaar in de literatuur over de benodigde grootte van de opslagcapaciteit van een droge bulk terminal. Vooronderzoek van deze studie heeft aangetoond dat de jaarlijkse doorvoer, dat doorgaans als enige richtlijn wordt aangehouden, de grootte van de opslagcapaciteit van bestaande terminals niet voldoende kan verklaren. Daarom is er behoefte aan een methode die inzicht geeft in de factoren die de benodigde opslagcapaciteit van een droge bulk terminal bepalen.

Dit onderzoek heeft zich gefocust op de benodigde opslagcapaciteit van import droge bulk terminals die hun bulk materialen identiteit-behoudend opslaan. Door stochastische invloeden fluctueert het opslagniveau sterk bij dit soort terminals. Zo heeft de terminal bijvoorbeeld nauwelijks tot geen invloed op de aankomsttijden van schepen. Wanneer schepen arriveren moeten zij binnen een afgesproken tijd zijn geholpen. Wanneer terminals hier langer over doen moeten zij demurragekosten betalen aan de eigenaren van de schepen. Om alle demurragekosten te voorkomen (en de hoogste pieken van het opslagniveau op te vangen) is een zeer groot opslagterrein nodig. Een grote opslag brengt echter zeer grote investeringskosten met zich mee. Hierdoor is het vanuit een economisch perspectief beter om af en toe demurragekosten te accepteren om zo lagere investeringskosten te hebben. In dit onderzoek is dan ook aangenomen dat de benodigde opslagcapaciteit van een droge bulk terminal de opslagcapaciteit is waarbij de minste kosten gemoeid zijn; ofwel de opslagcapaciteit die resulteert in de hoogste netto contante waarde (NPV).

Om de benodigde opslagcapaciteit te bepalen is eerst een uitgebreid simulatiemodel ontwikkeld van een import droge bulk terminal in het discrete event simulatiepakket TOMAS in het programma Delphi. In dit simulatiemodel is getracht de eigenschappen van de terminal zo realistisch mogelijk te beschrijven. De stochastische eigenschappen van de terminal zijn meegenomen, zoals de aankomsttijden van schepen (Erlang-2 verdeeld), de scheepstypes, de scheepscapaciteiten, en de opslagtijd van de bulk materialen op het opslagterrein (Erlang-2 verdeeld). Ook zijn alle inkomsten en kosten van de terminal opgenomen in het model, waardoor voor elke simulatierun automatisch de NPV wordt bepaald. Daarbuiten is het gemakkelijk voor de gebruiker om de waarden van alle variabelen (inclusief de distributies van de stochastische variabelen) aan te passen, zodat vele soorten import droge bulk terminals kunnen worden geconfigureerd.

Page | vii Vervolgens is de benodigde opslagcapaciteit per configuratie bepaald door het simulatiemodel te optimaliseren naar de NPV: de benodigde opslagcapaciteit voor een configuratie is de opslagcapaciteit waarbij de hoogste NPV optreedt. De gangbare gradiënt-gebaseerde optimalisatiealgoritmen zijn echter niet toepasbaar. Bij het optimaliseren van de NPV van het simulatiemodel treden namelijk twee problemen op. Enerzijds is de NPV functie niet-lineair en bevat het duizenden lokale maxima. Daarnaast wordt de benodigde opslagcapaciteit beïnvloedt door de capaciteiten van de kade kranen en de stacker-reclaimers. Om deze redenen is het “simultaneous perturbation stochastic approximation” (SPSA) algoritme met geïnjecteerde ruis geïmplementeerd in het simulatiemodel. Dit algoritme kan efficiënt de NPV met respect naar drie variabelen tegelijkertijd optimaliseren en beschikt over een unieke methode om de gradiënt te schatten, waardoor het in staat is het globale maximum te benaderen.

De benodigde opslagcapaciteit valt op te delen in twee componenten: een deterministisch en een stochastisch deel. Het deterministische deel staat gelijk aan het gemiddelde opslagniveau en wordt berekend door de gemiddelde verblijftijd van de bulk materialen op de terminal te vermenigvuldigen met de jaarlijkse doorvoer. Het stochastische deel is de opslagcapaciteit dat een terminal nodig heeft door toedoen van de stochastische eigenschappen van de terminal. Het stochastische deel van de benodigde opslagcapaciteit kan worden uitgedrukt als percentage van de jaarlijkse doorvoer. De grootste invloeden op de grootte van het stochastische deel van de opslagcapaciteit zijn de jaarlijkse doorvoer en de opslagtijd van de bulk materialen.

In de basis configuratie hebben kleine terminals (10 Mton jaarlijkse doorvoer) 7.6% van de jaarlijkse doorvoer als opslagcapaciteit nodig om de stochastische effecten op te vangen. Voor grote terminals (50 Mton jaarlijkse doorvoer) is dit slechts 3.1%. In de basisconfiguratie is uitgegaan van een opslagtijd van ongeveer twee maanden. Wanneer de opslagtijd wordt verkleind tot één maand daalt de grootte van het stochastische deel van de opslagcapaciteit met 24%, terwijl het stochastische deel stijgt met 12% wanneer de opslagtijd wordt vergroot tot 3 maanden. Van de stochastische variabelen heeft de verdeling van de aankomsttijden van schepen de grootste invloed op het stochastische deel van de benodigde opslagcapaciteit. De effecten van verschillende scheepseigenschappen en onregelmatige opslagtijden zijn veel kleiner.

Er kan worden geconcludeerd dat de benodigde opslagcapaciteit kan worden bepaald door een simulatiemodel te ontwikkelen en daarin de NPV te optimaliseren. Op deze manier is de benodigde opslagcapaciteit de opslagcapaciteit die de minste kosten met zich mee brengt. De stochastische invloeden op de opslagcapaciteit worden kleiner naarmate de jaarlijkse doorvoer van een terminal groter is. Afhankelijk van het soort terminal, is er door stochastische invloeden 3 tot 10% van de jaarlijkse doorvoer aan extra opslagcapaciteit nodig ten opzichte van het gemiddelde opslagniveau.

Page | viii

List of Symbols



Scalar that determines how fast ak decreases



Scalar that determines how fast ck decreases

k



The p-dimensional perturbation vector at iteration ‘k’.

( )

f x



Gradient function x

k



 _{Scalars of unknown additive noise}



Annual throughput

μ Mean



Standard deviation

θ The p-dimensional explanatory variables of the algorithm

2

( )

k



Chi Square distribution with ´k´ degrees of freedom

k



Independently identically distributed injected noise term ak Gain sequence for value of iteration

ck Gain sequence for gradient

C

Required storage capacity

cranes

C

Combined capacity of quay cranes Cdem Sum of all demurrage costs

Deterministic Part

C

Deterministic part of the required storage capacity

Little's Law

C

Storage capacity due to Little’s Law

handling

C

Storage capacity due to handling times of stacking and reclaiming CINV Sum of all investments costs

SR

C

Combined capacity of stacker-reclaimers

Stochastic Part

C

Stochastic part of the required storage capacity

g(θ) True gradient

ĝ(θ) Estimated gradient L(θ) True loss function

k

q

Factor to influence the size of



_k

s Standard deviation

Tdep Depreciation time of the investments Thandling Handling time for stacking and reclaiming Trun Runtime of the simulation

Page | ix X Storage capacity needed to handle stochastic properties of terminal as a percentage

of the annual throughput

i

x

Observation ‘i’ of NPV

y(θ) Observed function

Z All standardized NPVs

zi Standardized NPV ´i´

List of Abbreviations

DSA Delft Systems Approach

EMO Europees Massagoed- Overslagbedrijf

JB Jarque-Bera test

IAT Inter-Arrival time IRR Internal Rate of Return NPV Net Present Value

K Kurtosis

OLS Ordinary Least Squares PDL Program Description Language PROPER PROcess PERformance model REV Sum of all revenues

RHDHV Royal Haskoning DHV

S Skewness

SA Stochastic Approximation

SE Standard Error

SPSA Simultaneous Perturbation Stochastic Approximation Std. dev. Standard deviation

Page | x

Contents

Preface………iii Summary………iv Samenvatting……….vi List of Symbols………viii List of Abbreviations………...ix

1

Introduction ... 1

2

Description of Dry Bulk Terminals ... 5

2.1 Dry Bulk Terminal Operations ... 5

2.2 Process Performance Model ... 7

3

Simulation and Optimization ... 13

3.1 Overview ...13

3.2 Simulation Model ...16

3.2.1 Short Summary of the Simulation Model ... 16

3.2.2 Input Parameters and Base Configuration ... 18

3.2.3 Program Description Language... 28

3.2.4 Verification of Simulation Model ... 39

3.2.5 Precision and Accuracy of Simulation Model ... 46

3.3 Optimization ...51

3.3.1 Introduction ... 51

3.3.2 Optimization techniques ... 54

3.3.3 Simultaneous Perturbation Stochastic Approximation ... 60

3.3.4 Verification of the SPSA Algorithm ... 68

Page | xi

4

Results... 73

4.1 Validation and Analysis of Storage Level in Simulation Model ...75

4.2 Determining the Design Parameters ...80

4.2.1 Quay Cranes and Stacker-Reclaimers Occupancies ... 82

4.2.2 Required Storage Capacity ... 83

4.3 Analysis of the Stochastic Part of the Required Storage Capacity ...89

4.4 Sensitivity Analysis Stochastic Variables ...93

4.4.1 Sensitivity of Inter-Arrival Times ... 93

4.4.2 Sensitivity of Storage Time ... 95

4.4.3 Sensitivity of Ship Characteristics ... 97

4.5 Sensitivity Analysis Demurrage Variables ... 100

4.5.1 Sensitivity of Demurrage Costs ... 100

4.5.2 Sensitivity of Discharge Rates ... 101

4.6 Case Description ... 103

4.7 Size and Sensitivity Stockyard Area ... 106

5

Conclusions and Recommendations ... 108

5.1 Conclusions ... 108

5.2 Limitations and Further Research ... 110

6

References ... 112

APPENDICES………116

A. Scientific Article……….1

B. Preliminary Research: Analysis of Existing Terminals……….8

C. Erlang Distribution in Delphi………..……….……….26

Page | 1

1 Introduction

Nowadays, coal is one of the main fuels for power generation. In a world population that increases by 80 million each year (US Census Bureau, 2012), the demand for energy increases as well. As a result, the last decades showed a tremendous increase of the global use of coal. As not only the world population, but also the world prosperity increases, the projections for the next decades show a growth for the global use of coal as can be seen in Figure 1.1.

Figure 1.1. Global coal import for 1980-2010 and forecast for 2015-2035 (US International Energy Information Administration, 2010).

Iron ore is mainly used for the production of steel. Due to the aforementioned reasons, the demand for steel has also increased over the last decades. Forecasts show that the global demand of iron ore will continue to increase for the next decades. Figure 1.2 shows the global steel production of the period 1980-2011 and a forecast for the period 2012-2020. The steel production has shown a steady growth in the last decades. According to various forecasts (T.A. van Vianen et al., 2010), this growth will continue for the next years.

Figure 1.2. Global steel production 1980-2011 (World Steel Association, 2010) and forecast for 2012-2020 (T.A. van Vianen).

0 500 1000 1500 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 Gl o b al C o al Im p o rt [M to n ] Year 0 500 1000 1500 2000 1980 1985 1990 1995 2000 2005 2010 2015 2020 Gl o b al Pr o d u ction o f Stee l [M to n ] Year

Page | 2 Coal and iron ore are acquired in mines and subsequently transshipped all over the world. Dry bulk terminals are used worldwide as a buffer of dry bulk materials between either international or intercontinental transportation and inland or domestic transportation or vice versa. Due to the high demand for energy and mineral resources many dry bulk terminals around the world are expanding and seriously increasing their capacity (Lodewijks et al., 2007).

Dry bulk terminals use a stockyard to store the bulk materials for a certain period of time. In many cases, import dry bulk terminals are located in highly populated areas. In such areas, the costs per square meter of land are high. As a result, terminals seek to use their stockyard as efficiently as possible. When expanding an existing dry bulk terminal or building a new dry bulk terminal, the required storage capacity is one of the main design parameters, as it determines a large part of the investment costs.

When the stockyard capacity is too small, arriving ships will not be able to unload their bulk materials. As a result, the terminal loses revenue (the bulk materials are unloaded somewhere else) or has to pay a fine (demurrage costs) to the ship owners. On the other hand, a surplus in storage capacity results in a waste of highly expensive space. Thus, from an economical point of view, the required storage capacity can be seen as the storage capacity where the costs are the smallest for the highest revenues.

This research will investigate the required storage capacity of dry bulk terminals, as this is such an important factor in the terminal design process. However, there is a lack of literature about determining the required storage capacity of the stockyard. One of the main design tools for dry bulk terminals is the report of UNCTAD (1985). However, here only calculations for the dimensions of a single stockpile are given. No information about the storage capacity or the storage size of the entire terminal is provided.

A way to determine the required storage capacity is to examine common industry practice: the so-called ‘rules of thumb’. The annual throughput of a dry bulk terminal determines to a great extent the size of the stockyard. As a result, rules of thumb use the annual throughput to determine the required storage capacity of an import terminal. For example, Royal HaskoningDHV determined that for large import terminals the required storage capacity should be approximately 1/12th _{of the annual} throughput for stevedore terminals and 1/8th _{for power plants. This is in the same range as the} estimates of Monrad and King (2007) and Schott and Lodewijks (2007), who approximate the stockyard capacity to be 10% of the annual throughput.

In order to verify the accuracy of such rules of thumb, a quantitative analysis of 48 existing terminals has been conducted. In this preliminary research, provided in Appendix B, the existing dry bulk terminals are classified by function and flow direction -and by the type of bulk they handle (Willekes, 1999). Regressions are conducted to explain the size of the storage area and of the number of

Page | 3 stockpiles located on the stockyard by the annual throughput and by terminal type. The analyses show that, in order to explain these characteristics, it is best to classify dry bulk terminals into import and export terminals. About two thirds of the variance between the characteristics of the stockyard of the 48 terminals was captured by this model. Consequently, one third of the variance could not be explained by the model. On average, import terminals require three times more storage area and four times more stockpiles per unit annual throughput than export terminals. This is a result of the fact that most import terminals need to stack the bulk materials identity-preserved: each shipload results in a separate stockpile.

As one third of the variance of the characteristics of the storage area could not be explained by the regressions of the preliminary research, no simple rules of thumb can be used to determine the size of the stockyard in the design process; it is not accurate enough. Clearly, there must be other factors that have a significant impact on the stockyard capacity.

So, besides the annual throughput and the terminal type, there are other factors that influence the required stockyard capacity. For example, the size of the ships and the storage time of the bulk materials differ for each arriving ship in the terminal and are thus stochastic. In simple analyses, these factors cannot be taken into account. Due to these stochastic factors, modeling dry bulk terminals in a deterministic setting will not capture the true characteristics. For example, Bugaric et al. (2009), determine the stockyard capacity by queuing theory. However, they also note that they were unable to capture any stochastic properties in this approach. As a result, the correctness of these outcomes is uncertain.

It is possible to determine the required storage capacity by neglecting all stochastic factors. Using simple calculations, the required storage capacity in a deterministic setting can be calculated by Little’s Law. The required storage capacity in a non-stochastic setting is then given by the multiplication of the storage time of the dry bulk materials and the annual throughput. However, neglecting the stochastic characteristics of dry bulk underestimates the required storage capacity.

In short, literature about the required storage capacity of import dry bulk terminals is scarce. Moreover, rules of thumb are insufficient to determine the required storage capacity of a terminal and analyses in deterministic settings cannot be used as well. As a result, the research question of this thesis is defined as follows:

“How can the required storage capacity of import dry bulk terminals be determined?”

As there are many different types of dry bulk terminals (see Appendix B), the scope of this study will be limited to import dry bulk terminals that store their bulk materials identity-preserved. In order to determine the required storage capacity of such terminals, a new method has to be developed.

Page | 4 To answer the research question, the following questions are defined:

 How does a dry bulk terminal work? (seeChapter 2)

 What factors influence the required storage capacity of dry bulk terminals? (See Chapter 3)

 How can the characteristics of import dry bulk terminals be modeled? (see Chapter 3)

 How can this model be used to determine the required storage capacity? (see Chapter 3)

 What are the effects of stochastic variables on the required storage capacity? (see Chapter 4)

 Can simple rules be derived to determine the required storage capacity? (See Chapter 4).

Outline of Thesis

First, in order to determine the characteristics of import dry bulk terminals, the Delft Systems Approach (Veeke et al. 2008) is used. This approach provides a clear overview of the functions that are fulfilled by a dry bulk terminal. The characteristics of the terminal are given in Chapter 2.

To capture the stochastic characteristics of the terminal, a simulation model is developed. The ideology of this study is that the required storage capacity can be seen as the storage capacity that has the perfect balance between investment costs and demurrage costs. Subsequently, a cost optimization must be conducted on the simulation model. The development of the simulation model and the optimization techniques are explained in Chapter 3.

Subsequently, the results are provided in Chapter 4. Variables can be adjusted in the simulation model. For example, the user can change the annual throughput, or the ship arrival distribution of the terminal. With the use of the simulation model and the optimization algorithm, the effects of parameter changes on the required storage capacity can be determined. Finally, the conclusions are discussed in Chapter 5. Limitations of this study and recommendations for further research are also provided here.

Page | 5

2 Description of Dry Bulk Terminals

This chapter will explain what import dry bulk terminals are, what they do and how their performance can be measured. This is necessary for the development of a simulation model. First, an overview of dry bulk terminals is given in Section 2.1. Here, also the scope of this study is provided. Subsequently, the processes of dry bulk terminals are investigated and modeled by the Delft Systems Approach in Section 2.2.

2.1 Dry Bulk Terminal Operations

Terminals are used to transship and store bulk materials. In this study, import terminals are assumed to be terminals where bulk materials arrive in large vessels and leave the terminal in smaller commodities. First an example of an import dry bulk terminal will be given, followed by the scope of this study.

Example of an Import Dry Bulk Terminal

By the aforementioned definition, the Europees Massagoed- Overslagbedrijf (EMO) in Rotterdam (see Figure 2.1) is considered to be an import dry bulk terminal. EMO will be described in order to gain more insight in import dry bulk terminal operations.

Figure 2.1. Photo of EMO in Rotterdam. (source: www.emo.nl)

At EMO, large quantities of coal, iron ore and other bulk material arrive in large ships. Subsequently, after a certain storage time on the stockyard, the bulk leaves EMO in smaller ships, trucks and trains.

a

b

e

d

Page | 6 During the research, a visit was made to EMO in order to gain insight into its processes and, more generally, the processes of an import dry bulk terminal. The main parts of the terminal are denoted by a letter in Figure 2.1. With the use of these main parts, in Figure 2.1, the working of a dry bulk terminal will now be explained.

Ships arrive at (a). The bulk materials on the ships are unloaded by large quay cranes. A photo of a quay crane is shown left in Figure 2.2. In huge grabs (85 tons safe working load), the bulk materials are unloaded and transported via a system of belt conveyors (b) (see Figure 2.2 right) to stacker-reclaimers on the stockyard.

Figure 2.2. Left: Quay cranes Right: System of belt conveyors

The stacker-reclaimers (c) stack the bulk materials of a shipload as a stockpile on the stockyard (d). This is shown in the left picture of Figure 2.3. The stockpile will remain on the stockyard for a certain period of time: the storage time (d). The owner of the bulk materials (customer) controls when the bulk materials will be removed from the stockyard. After the storage time, the bulk materials are reclaimed by the stacker-reclaimers (c). Finally, the bulk materials are removed from the terminal: this can be performed by trucks, trains or smaller ships (e).

Figure 2.3. Left: A stacker-reclaiming is stacking a stockpile Right: Shiploader

a

b

c

d

Page | 7 Scope of Study

Note that the aforementioned description of EMO was highly simplified. In this study, it is not possible to cover all processes of dry bulk terminals. Therefore, some assumptions are made about the dry bulk terminals that are considered in this study:

 Use of stacker-reclaimers. At EMO, the same machines are used for the stacking and reclaiming of dry bulk materials: the stacker-reclaimers. However, there are also terminals that have separate machines for the stacking (stackers) and for the reclaiming process (reclaiming). In this study, only stacker-reclaimers will be considered.

 Terminal influences. In this study, import stevedore dry bulk terminals are considered. It is assumed that these terminals have no influence on the arrival time of a ship and on the storage time of bulk materials on the stockyard. The ship owners determine when a ship arrives at the terminal, and the owners of the bulk materials determine the storage time of the bulk materials on the stockyard.

 Identity-preserved storage. In this study, the bulk materials of a shipload will never be stacked onto an existing stockpile. The bulk materials of each shipload will be stored in separate stockpiles.

 Bypassing and blending. Bypassing is a process in which the bulk materials are transported directly to other modalities. Thus, the bulk materials are not stored on the stockyard. Both bypassing and blending will not be considered in this study.

2.2 Process Performance Model

Before it is possible to develop a simulation model of an import dry bulk terminal, all processes and relations between the processes must be structured. Therefore, the Delft Systems Approach (DSA) is used (Veeke et al. 2008). The DSA is a useful way to structure and gain insight in processes. Dry bulk terminals are transshipment and transport systems, consisting of different subsystems that enable a division of functions according to place, time, personnel and means (Schott and Lodewijks, 2007). As a dry bulk terminal is such a complex system in which aspects, which are subsets of the relations, can be distinguished, the process performance (PROPER) model (Veeke et al. 2008) of the DSA will be used. The PROPER model contains all relations between the primary process of the terminal (providing service: the transshipment of bulk from ships via the stockyard to outgoing modality) and the other functions of the terminal. With the use of the PROPER model, the development of a simulation model is less complex.

Page | 8 A simple form of the PROPER model of an import dry bulk terminal is shown in Figure 2.4.

Standards Results Performance Requirements

Environment

Handle service

Terminal control

Ships Bulk materials on ship Equipment Handled ships Bulk materials on outgoing modality Used equipment

Figure 2.4. The PROPER model of an import terminal. The thick arrows represent material flows, while the thin arrows represent the flows of information.

The PROPER model of Figure 2.4 consists of an operational subsystem (handle service) and a control subsystem (terminal control). The black box of the operational subsystem contains all processes that fulfill the primary function of the terminal and includes three aspects of the terminal model. Each aspect represents the transformation of a flow. Three types of flows can be identified: the transformation of the ships, the transformation of the bulk materials on the ship and the transformation of the equipment.

The terminal control function in Figure 2.4 coordinates the three transformations of the black box. This is done by generating tasks out of each arriving ship and by assigning the necessary equipment.

The terminal receives requirements from its customers and the authorities (the environment) and may deliver its performances to them. Without requirements and performances it is impossible to determine how well one terminal performs in contrast to another terminal. This is important in this study, as eventually many different capacities of the stockyard will be compared to determine the best design. Therefore, the objectives of the terminal must be clear. A method to define the most important objectives of a terminal is by the ‘six main process criteria’ of Bikker. These criteria are also known as the ‘key performance indicators’. For the design of a terminal, the following criteria can be identified:

Page | 9

 1. Effectiveness: the terminal must guarantee a certain annual throughput.

 2. Productivity: the handling costs of the bulk material for the terminal should be as low as possible.

 3a. Product flexibility: the terminal must be able to handle both coal and iron ore in different quantities.

 3b. Process flexibility: the terminal must be able to handle different types of vessels.

 4a. Control of external targets: the terminal must be able to guarantee a competitive service rate for unloading ships and must meet all governmental demands.

 4b. Control of internal targets: the terminal must be profitable and must be able to generate these cash flows on the long term. This means that the revenues must exceed all costs of the terminal.

 5. Quality of work: the terminal must be a safe environment for employees.

 6. Innovative potential: the performance of the terminal must be controlled at all times.

The financial performance of a terminal (taken into account by point 4b) is considered to be the most important performance indicator in this research. Customers will always try to use the best services for the lowest price. So, when the service rates of a terminal are too high, the customer will either use the service of a cheaper terminal or will negotiate the service price. As a result, terminals that are able to make a profit on the long-term with their services have a competitive service rate.

Furthermore, in this study it is assumed, that if a terminal is profitable on the long-term, the other performance indicators are met as well. For example, if the main activities of a terminal turn a profit on the long-term, it automatically has a competitive service rate (otherwise there would not be customers and the required annual throughput would not be obtained) and the terminal is a safe environment for its employees (otherwise the terminal would have been forced to close, and there would not be any profit). Also, a good performing terminal has a healthy balance in the capacities of its equipment. A lack of storage capacity and capacity of the equipment results in waiting times for the vessels. However, an over-capacity is a waste of capital.

As the financial performance is such an import performance indicator of the terminal, it makes sense to link the financial performance of a terminal to the objective of this study: “determining the required storage capacity of an import dry bulk terminal”. Thus, the required storage capacity is obtained by the best financial performance of the terminal.

It is possible to zoom in on the PROPER model of Figure 2.4. This is achieved by opening the black box of the three product transformations, which results in Figure 2.5.

Page | 10

Performance

Terminal control

Standards Results

Operate: transship bulk

from ship via stockyard to

outgoing modality

Bulk materials on ship Bulk materials on outgoing modality FUNCTION CONTROL

Handle service

Ships Handled _ships

FUNCTION CONTROL

Use: use equipment

Equipment Used equipment FUNCTION CONTROL Task Assign Requirements

Figure 2.5. PROPER model of the terminal where the three flows (ships, bulk on ship, equipment) are distinguished.

In contrast to the three flows of Figure 2.4, in Figure 2.5, five flows can be identified:

 Ship flow: each arriving vessel produces a new order. To handle this order, the ship must be unloaded. This order process concerns the customer service.

 Bulk materials flow: all primary processes happen in this bulk materials flow; the shipload is unloaded from the ship by quay cranes. In some cases the shipload is temporarily stored on the quay (this only happens if no stacker-reclaimers are available). Consequently, the shipload is transported to the stockyard by a system of belt conveyors and stacked as stockpiles. Eventually, after a certain storage time on the stockyard, the bulk materials are reclaimed and leave the terminal. The intended result of this flow is that the required annual throughput will be achieved.

 Equipment flow: the primary processes can only be executed with the use of equipment. This flow contains the allocation and capacities of the cranes, the system of belt conveyors, the stacker-reclaimers and the stockyard.

 Task flow: the task flow is located between the “handle service” function and the “operate” function. A task is defined as the transshipment of the quantity of bulk materials that will become

Page | 11 one stockpile on the stockyard. The lead time of all tasks is the time of arrival of a ship in the system until its departure. One order (ship) may result in multiple tasks as the bulk on the ship may contain bulk for more than one stockpile.

 The assignment flow: the assignment flow is located between the “operate” function and the “use” function. The assignment flow is used to assign equipment to each task.

To fully understand the bulk and equipment flow in Figure 2.5, the two flows are considered in more detail in Figure 2.6.

Unload bulk from ship to quay Bulk on ship Bulk on outgoing modality Function Control Service Ships Handled ships Use quaycranes Equipment Used equipment Task: transship bulk Progress: transshipped bulk Stack bulk on

stockyard Reclaim bulk

Use stacker-

reclaimers Use stockyard

Use stacker- reclaimers

Terminal control

Operate: transship bulk from vessel via stockyard to outgoing modality

A s s ig n R e le a s e A s s ig n R e le a s e A s s ig n R e le a s e A s s ig n R e le a s e FUNCTION CONTROL Stock-yard Results Standards Requirements Performance Function Control Transportation to stockyard Use system of belt conveyors A s s ig n R e le a s e

Use: use equipment

Figure 2.6. PROPER model of the terminal, where the bulk and equipment flow are considered into more detail.

Figure 2.6 reveals how the product and resources flow behave as explained in the previous text. A blue triangle denotes a buffer. Two buffers for the bulk materials are shown in Figure 2.6: besides the stockyard buffer, there is also a small buffer at the quay where bulk is temporarily stored.

In practice, it is not always possible to transport the shipload to the stockyard immediately. In case of maintenance or in case of adjusting the system of belt conveyors, the shipload cannot be transported. Also, the shipload will not be transported if no stacker-reclaimers are available: if a stacker-reclaimer is reclaiming a stockpile, it may finish the reclaiming, before it is used to stack the shipload. Thus, there are many situations in which the shipload must be stored temporarily at the quay. However, this

Page | 12 situation is undesirable as it requires double handling: terminal management will always try to avoid this.

Example

In order to show that terminals sometimes perform the undesirable double handling, by storing bulk materials for a very short time on the quay, Figure 2.7 is provided. Here, a photo is shown of EMO where a shipload is stored on the quay.

Figure 2.7. Photo of situation where a shipload is temporarily stored on the quay as buffer. (source: EMO, KIVI Niria presentation 27 September 2012)

In situations as shown in the figure, the bulk materials are stored until the system of belt conveyors and stacker-reclaimers are available.

Figure 2.6 shows that equipment is assigned for each process of the bulk materials. First, the shiploads are unloaded. After unloading, they are transported to the stockyard. This transport occurs either directly, or via a short storage at the quay (as discussed before). After arriving at the stockyard, the bulk materials are stacked. Subsequently, after the storage time of the stockpile, the stockpile is reclaimed and leaves the terminal.

This PROPER model of Figure 2.6 will be used to construct a simulation model of the terminal. The models have provided a clear description of the most important processes of a terminal. Furthermore, the simulation model must be able to show its performance. For this, some of the main performance indicators of Bikker will be used. Extra attention must be given to the most important performance indicator: the terminal’s financial situation.

Page | 13

3 Simulation and Optimization

In this chapter, the methods to determine the required storage capacity of an import terminal will be explained. In order to do so, a simulation model and an optimization algorithm will be used.

First the structure of the PROPER model of Chapter 2 will be translated into a simulation model for the program Delphi. The input parameters of this simulation model can be adjusted. A simulation run for each set of input parameters, a configuration provides information about the performances of the terminal. Special attention is given to the financial performance of the terminal by providing a net present value (NPV). The NPV can be seen as the sum of the costs (investments and demurrage) and revenues. Thus, if the value of the maximum storage capacity is changed in the simulation model, the investments and demurrage costs vary and a different NPV is obtained.

The required storage capacity is considered to be the storage capacity that corresponds with the highest NPV. To obtain the required storage capacity, the NPV of the terminal must be optimized with respect to, among others, the storage capacity. The used optimization technique is the Simultaneous Perturbation Stochastic Optimization (SPSA) algorithm.

The outline of this section is as follows. First, in Section 3.1, an explanation is provided why simulation should be used instead of deterministic methods. This paragraph also explains what simulation package is used and what methods are used to determine the required storage capacity. Roughly two steps can be identified in the method: the development of the simulation model, and the use of the SPSA algorithm. The characteristics of the simulation model will be explained in Section 3.2. Subsequently, the use of optimization techniques will be discussed in Section 3.3.

3.1 Overview

In a deterministic setting, in which calculations can be made with the use of queuing models (see for example Bugaric et al., 2009) or in the Microsoft Excel program, the true behavior of a terminal cannot be captured. This is because these methods are not able to capture the stochastic properties of a dry bulk terminal. These stochastic properties can be seen as uncertainties in the model. In a dry bulk terminal the following uncertainties can be identified:

 Inter-arrival times of ships. Terminals do not know exactly when a ship will arrive. The ‘customer is king’ principle applies to terminals: the ships determine when they arrive and when they must be unloaded.

Page | 14

 Ship types. Different types of ships arrive at the terminal. Besides the space each type of ship uses at the quay, larger ships can be unloaded by more quay cranes than smaller ships.

 Ship capacities. Each ship has a different capacity. Therefore, each ship has a different unloading time and the bulk materials on the ship require a different amount of storage on the stockyard.

 Storage time of the bulk materials. The stockpiles on the stockyard have different storage times. Some stockpiles will be reclaimed after a few days, while other stockpiles will be stored for several months.

In a deterministic setting, the averages of these uncertainties are taken. However, this system is flawed. For example, the required layout for a terminal where a ship arrives every 24 hours is completely different than for a terminal where two ships may arrive at the same time, while no ships arrive on the following day. Stochastic variables cause ‘peak moments’ and terminals should be able to deal with these ‘peaks’. In order to do so, terminals must have larger capacities for the quay cranes, for the stockyard and for the stacker-reclaimers. Thus, the required storage capacity consists of the storage capacity in a deterministic scenario plus the additional storage needed to handle its stochastic properties.

TOMAS Package

In order to capture the stochastic properties in a model, the dry bulk terminal must be modeled in a discrete-event simulation environment. One way to do this is by using the discrete simulation package TOMAS (Tool for Object oriented Modeling And Simulation, www.tomasweb.com) in the Delphi program. In TOMAS, the user is able to simulate all processes from scratch as TOMAS has all the basic needs for a process oriented simulation. Also, TOMAS contains visual control, queues and statistics. This way, the program offers a unique flexibility for the user. For these reasons, TOMAS Delphi will be used as the simulation environment for the dry bulk terminal.

Stochastic variables have a random value, instead of a pre-defined value. In TOMAS, this random value is drawn from a certain distribution. TOMAS has a good random number generator to model these stochastic variables by a multiplicative random generator. For each distribution the user must provide a seed, which is 32-bit integer. The generator has a cycle length of 232_{-1, which guarantees} an incredible amount of random numbers.

In order to support the modeling of stochastic behavior, TOMAS provides a Uniform, Exponential (also known as the negative exponential distribution), Normal and Table distribution. However, for the arrival pattern of ships, the use of an Erlang-2 distribution is the most accurate (UNCTAD, 1985). According to the terminal EMO, this distribution also models the storage time of the bulk on the stockyard best. As this distribution is not available in the standard TOMAS, it is added to the source

Page | 15 code of TOMAS. The addition of the Erlang-k distribution to the source code is given in Appendix C. This distribution is also used for the storage time of the bulk materials.

Net Present Value Ideology

The simulation model must be able to capture the processes and the stochastic properties of an import dry bulk terminal. In Chapter 2, it was explained that the financial situation of a terminal is the main performance indicator for this study. Therefore, the user must be able to change certain parameters in the simulation model, resulting in different values of the revenues and costs of the terminals. The concept of the NPV will be explained in Chapter 3.2.2.1 and is defined as all future revenues of the terminal (incoming cash flows) minus all its future costs (outgoing cash flows). The total costs of the terminal consist of demurrage costs and investment costs. So, when, for example, the value of the storage capacity in the simulation model is changed, different values for the investment and demurrage costs occur. Subsequently, this results in a different NPV. The ideology of this study is that the required storage capacity is the storage capacity that corresponds with the highest NPV.

Example: NPV Ideology

Due to the stochastic properties of a terminal, the storage level of a terminal varies over time. A hypothetical situation is given in the figure below, where the storage levels are shown for one year.

Figure 3.1. Representation of a storage level (y-axis) over time (x-axis).

The figure shows that the average storage level is approximately 5.5 Mton in this situation. However, the maximum and minimum storage levels are 6.5 Mton and 4.5 Mton respectively. So, what should be the required storage capacity? If the storage capacity is allowed to be 6.5 Mton, high investments have to be paid for a situation that happens approximately once or twice a year. However, at this storage level there would be no demurrage costs as a result of the storage capacity. If the storage capacity only is allowed to be, for example, 6.4 Mton, less investment costs have to be made. However, at this storage level, for some ships demurrage costs have to be paid each year. Thus, the balance between the investment costs and the demurrage costs determines the required storage capacity.

Besides, the maximum storage capacity, other parameter values can be adjusted in the simulation model as well. Each change in parameters creates a different configuration of the simulation model. Subsequently, for each configuration, the required storage capacity that corresponds with the highest NPV must be determined. Therefore, an algorithm will be used that optimizes the NPV for each configuration. This optimization algorithm determines the storage capacity at which the highest NVP occurs. Details of the optimization algorithm will be provided in Section 3.3.

4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5 4 4.5 5 5.5 6 6.5

Simulation Time [years]

St ora ge Le v el [M ton ]

Page | 16

3.2 Simulation Model

In this paragraph, the details of the simulation model will be discussed. First, a short summary of the simulation model will be provided in Section 3.2.1. Subsequently, the input parameters and the base configuration of the model will be discussed in Section 3.2.2. Then, Section 3.2.3 provides a process description language of the processes of the terminal. Of course the simulation model must be verified as well. Methods for the simulation model verification are provided in Section 3.2.4. Here, also the specifications of the output screen of the simulation model are explained. The validation of the simulation model will be conducted in the results-section. Finally, in Section 3.2.5, the precision and accuracy of the simulation model will be tested.

3.2.1

Short Summary of the Simulation Model

In the PROPER model of Figure 2.6, the main processes can be recognized: the unloading of the shipload to the quay, transportation of the bulk to the stockyard, the stacking of the bulk on the stockyard, and the reclaiming of the bulk from the stockyard. The elements that are recognized in Figure 2.6 are the ships, the shipload that results in a stockpile and the stockpile itself. However, these elements and processes are not enough to construct the full simulation model.

In the model, three ship types are considered: large capesize, medium panamax and small handysize ships. These types are considered to be the most occurring at import dry bulk terminals. The inter-arrival time has been determined by the annual throughput and the average ship capacity. The ship is only able to berth if the quay is not fully occupied. Otherwise, the ship will have to wait outside the port. Ships berth in the same order as they were generated. Once berthed, the ship will be unloaded once the following two constraints are met:

- The load of the ship fits on the stockyard of the terminal. - There are cranes available to unload the ship.

After unloading, the load is instantly stacked on the stockyard by a stacker-reclaimer. If there is no stacker-reclaimer available, the load waits on the quay until a stacker-reclaimer is available. In contrast to the PROPER model in Figure 2.6, no system of belt conveyors is taken into account for the transportation of the load. This is because belt conveyors work continuously, and if the capacity of the conveyors belts has been made large enough, it has no effect on the size of the storage capacity.

At terminals, the dimensions of stockpiles have to fit on the stockyard. Therefore, it may happen that one shipload should be divided into several stockpiles that fit on the stockyard. To capture this characteristic, the user of the model is able to specify the maximum mass for a stockpile in the stockyard. As a result, a shipload that exceeds this maximum mass is divided into multiple stockpiles of the same size. Note that a shipload is never stacked on an existing stockpile: each shipload results in at least one new stockpile (identity-preserved).

Page | 17 For the calculation of the storage area, it is assumed that the stockyard consists of just one large lane of stockpiles. The piles on the lane are separated by two meters of empty space. More information about stockpiles can be found in Appendix D.

Within a certain amount of time, the ‘maximum storage time’, the whole stockpile is removed from the stockyard. The stockpile can be removed as a whole at once after this maximum storage time, or the stockpile can be removed in several batches. As the required storage capacity depends on the average storage time of bulk materials on the stockyard, smaller batch sizes decrease the required storage capacity.

The stockpile is removed by a reclaimer. The stockpile can only be removed when a stacker-reclaimer is available. After reclaiming, the load is considered to be out of the terminal, and thus outside of the system boundary. The outgoing modalities have been left out of the scope.

Assumption 1

Once the stockpiles are reclaimed, the model does not take into account how the bulk materials leave the terminal. Bulk materials leave the terminal by train, vessel and/or truck. The capacities of these modalities are much smaller than the capacities of ships at the input of the terminal and is therefore assumed to be continuous. While the arrival of two large vessels at the same time results in an enormous peak of the storage level, the arrival of two trucks at the same time will not have a large effect on the storage level.

Assumption 2

Stacker-reclaimers are used in the simulation model. This means that the same machines are used to stack the bulk materials on a stockpile and to reclaim them. There are also terminals that use separate stackers and separate reclaimers, but this has not been taken into account. Furthermore, it is assumed that the stacking capacity equals the reclaiming capacity. In practice, the stacking capacity is larger than the reclaiming capacity. However, the objective of this study is to determine the stockyard capacity and it is assumed that this is not affected by the stacking-reclaiming ratio.

Assumption 3

The system of belt conveyors to transport the bulk from the quay to the stockyard is not taken into account. The model assumes that the bulk can always be transported instantly to the stockyard and that the transportation time compared to the total unloading time can be neglected.

Assumption 4

The net effective capacity has been used for all equipment (Ligteringen, 2006). In order to explain what this net effective capacity is, the quay cranes will be used as an example. The capacity of quay cranes is not constant for the unloading of a ship. In the beginning of the unloading of a ship, full

Page | 18 buckets of products can be obtained. This is called the free digging stage. However, after about 50% of the load, the capacity decreases as it takes longer to perform a grab cycle. This intermediate stage continues for about 35% of the load. The last 15% of the load is harder to collect and takes much longer: the trimming stage. Also, quay cranes have to switch between holds on the ships and holds have to be cleaned by wheel loaders during the process. Furthermore, between two ships, quay cranes have to move again. All these factors decrease the effective capacity of the crane and similarly for the stacker-reclaimer.

As in the model, the focus is not on individual ships, but about the performance of a terminal throughout many years. Therefore, the average effective capacity has been used. For the quay cranes this capacity is defined as the average hourly tonnage during the unloading of the entire load of a ship. The necessary interruptions for trimming, cleaning up, moving between holds, and shifting the cranes from one ship to another have been taken into account. For the stacker-reclaimers this capacity is defined as the average hourly tonnage during the stacking or reclaiming of an entire stockpile. All necessary interruptions are taken into account.

3.2.2

Input Parameters and Base Configuration

The idea of the simulation model is that the user can easily change the specifications of the terminal. This is done by changing the values of the input parameters, which results in a different configuration. Each configuration results in a different NPV. In the aforementioned text, it was explained that an optimization algorithm will be used to optimize the NPV with respect to the required storage capacity. So, a change in the value of the storage capacity in the simulation model of the terminal causes a change in the NPV and will finally result in a different required storage capacity. Of course, the values of all other parameters must be kept equal.

In the simulation model, the maximum mass of an output batch and the annual throughput can be adjusted in the main simulation screen. All other parameters have their own ‘options screen’ in the simulation model. Four types of these input parameters can be distinguished:

1. Terminal finances (Section 3.2.2.1) 2. Shipload storage (Section 3.2.2.2) 3. Arrival pattern of ships (Section 3.2.2.3)

4. Design parameters and number of equipment (Section 3.2.2.4)

When one of these options is changed, the user has made a new configuration. Each configuration results in a different NPV and consequently, with the use of the optimization algorithm, in a different required storage capacity.

Page | 19 The characteristics of the four types of input parameters will now be discussed in more detail. Also, explanations for the values set as standard are given. The combination of these standard values is called the base configuration of the simulation model. In most cases, values are chosen in the base configuration that correspond to the characteristics of EMO. This is because EMO is assumed to be a proper functioning large import dry bulk terminal that handles both coal and iron ore, and can also berth all types of ships.

3.2.2.1 Terminal Finances

Mainly, dry bulk terminals earn their money by the handling and storage of bulk materials. In the simulation model, the concept of NPV is introduced in order to see how the terminal performs for different configurations: so, the main output of the simulation model is the NPV. Here, NPV is calculated as the handling and storage revenues minus the demurrage and investment costs. The NPV is used as input for the optimization algorithm in Section 3.3 to determine the required storage capacity. The calculations for the NPV are based on Berk and DeMarzo (2007). Figure 3.2 shows a representation of the factors that determine the value of the NPV.

NPV

Revenues Costs

Handling revenues

Storage

revenues Investment costs Demurrage costs

Total hours of vessels in demurrage [hours] Average demurrage rate [$/hour] Investment stockyard Investment stacker-reclaimers Investment quay cranes

+

x

-+

+

Total tonnage of bulk [ton] Revenues per ton [$/ton]

x

Total storage _{time [days]} storage rate Average [$/day]

x

Figure 3.2. Representation of the construction of the NPV.

Note that the NPV is very simplified: other expenses like overhead costs and marketing are not taken into account. However, the NPV still has meaning: it measures the balances for the costs of equipment and the revenues of the terminal. Therefore, different configurations can be compared to see which configuration performs better: the configuration with the higher NPV.

For example, when a terminal turns out to have a NPV of $5,000,000 in the simulation model, the sum of the revenues is $5,000,000 higher than the total investments and demurrage costs. However, if the terminal has a NPV of $7,500,000 in another configuration (for example, due to a change in the value of the maximum storage capacity), the user knows that the second configuration performs better. So, the NPV of the terminal in the model is used to compare different configurations.

Page | 20 The user of the simulation model can change the specifications of the model in the ‘finances window’. The options of this window are shown in Table 3.1. Note that the base values for these options are filled in, but can easily be adjusted by the user.

Table 3.1. The options corresponding to the finances of the terminal. The shown values correspond to the base configuration and can be adjusted by the user.

Parameter Value Parameter Value

Investment cranes 3000 [$/ton/hour] Demurrage capesize 13,000 [$/day] Investment stacker-reclaimers 3000 [$/ton/hour] Demurrage panama 10,500 [$/day]

Investment stockyard 45 [$/ton] Demurrage handysize 10,000 [$/day]

Time low storage rate 60 [days] Discharge rate capesize 30,000 [ton/day] Low storage rate 0.02 [$/ton/day] Discharge rate panama 20,000 [ton/day] High storage rate 0.03 [$/ton/day] Discharge rate handysize 15,000 [ton/day] Handling revenues unloading 0.1 [$/ton] Depreciation of investment 30 [years] Handling revenues stacking/reclaiming 0.05 [$/ton] IRR 0 [-]

Each of the options in Table 3.1 will be thoroughly explained in the corresponding text of the following types of finances that can be distinguished in Table 3.1:

a) Investment costs b) Revenues

c) Demurrage costs

These types of finances, the options and their corresponding base values will now be discussed in further detail. Subsequently, in d), the formulas to calculate the NPV will be explained.

a) Terminal Finances: Investment costs

In this study, the investment costs can be divided into three types:

- The investments for the capacity of the cranes that unload the ships - The investments of the stacker-reclaimers

- The investments costs of the stockyard

These three investments combined determine all investments of the terminal:

C

_INV. The more investments are spent on these types, the more capacity each of the types has and the easier the simulation model will be able to deal with the peaks due to the effects of the stochastic properties. However, these investment costs are very hard to estimate.

In a study conducted by Strien (2010), the investment costs of a bucketwheel stacker-reclaimer with a capacity of approximately 3000 ton/hour are estimated to be around $10,000,000. As this is an approximation, the investment costs of a stacker-reclaimer in the model are set to $3,000 per ton per

a) Investment Costs

b) Revenues

c) Demurrage Costs

Page | 21 hour. The investment costs of the capacity of the cranes are set equal to the investments costs of the capacity of the stacker-reclaimers. Note that these numbers are estimates and can be adjusted by the user in the simulation model.

The investment costs of the stockyard are determined by the amount of bulk that can be stored per hectare and the investment price per hectare. Most of the time, import terminals are located in extremely expensive areas. For the base configuration, the price of the Maasvlakte 2 in the port of Rotterdam is used. Assuming prices at Maasvlakte 2 equal prices of the port of Amsterdam, the costs per m2 _{is approximately €300 or $375 (source: Grondprijzenbrief Amsterdam, 2012). With the use of} the data in T.A. van Vianen et al (2010) the average mass per hectare of import dry bulk terminals that handle both coal and iron ore is determined to be around 80,000 ton. Note that this mass represents a mix of dry bulk materials. Subsequently, the investment costs for the storage result in

( )

per ton and are rounded to $45 per ton in the base configuration.

b) Terminal Finances: Revenues

In the simulation model, the revenues of the terminal are determined by the handling revenues of the ship unloading, by the handling revenues of the stacking and reclaiming of the bulk and by the money received for storage. The sum of all revenues is denoted by

REV

. These earnings will now be

discussed in further detail.

- Ship unloading. Terminals receive money for each ton of bulk they unload from the vessels. The amount of money does not depend on the ship type. Assuming a handling rate per ton of $2 and a 5% profit, the terminal makes $0.10 profit for each ton of bulk they unload with the cranes. Note that this number equals the operational costs of the cranes per ton bulk minus the money customers pay per ton bulk. The revenues for the ship unloading are denoted by ‘handling revenues unloading’ in the cost options window.

- Stacking and reclaiming. Terminals earn revenues for each ton of bulk they stack and reclaim on the stockyard. Assuming a handling rate per ton of $1 and a 5% profit, the revenues per ton for the combination of stacking and reclaiming is set to $0.05 in the base configuration. Again, this number equals the operational costs of the stacker-reclaimer for a ton of bulk materials minus the handling price customers pay per ton bulk materials. The revenues for the stacking and reclaiming are denoted by the ‘handling revenues reclaiming’ in the cost options window.

- Storage. Terminals receive money for each day bulk is stored on the stockyard. There are terminals where the prices are raised after a certain amount of time. This is because terminals want to be able to guarantee storage capacity to their customers. In the model, the price per ton per day before and after these days can be specified. In the base configuration, prices increase