Ilgin Diss

Concept Development to Control Non-value Added Logistical Costs
in a Primary Aluminium Casthouse

by Interfacing Linear Optimization and Simulation
Von der Fakultät für Ingenieurwissenschaften,

Abteilung Maschinenbau und Verfahrenstechnik
der
Universität Duisburg-Essen
zur Erlangung des akademischen Grades
eines
Doktors der Ingenieurwissenschaften
Dr.-Ing.
genehmigte Dissertation
von
Gökay Ilgin
aus
Kirklareli, der Türkei
Gutachter: Univ.-Prof. Dr.-Ing. Bernd Noche

Univ.-Prof. Dr.-Ing. Thomas Steinhäuser
Tag der mündlichen Prüfung: 06.09.2013

Abstract
After the financial crisis in 2008, demand reduction especially from the automotive in-
dustry and changes in CO2 tax regulations which increased the energy prices the alu-
minium industry forced to review and reduce its operational expenditure. High energy
consumption in the production of primary aluminium dedicates most of the efforts on
technological development onto the electrolysis unit. However, other units of a smelter
also have the potential to improve their operational efficiency. In this thesis, the focus is
on the casthouse unit of the smelter. The aim of this research is to quantify and reduce
the non-value added logistical costs in the aluminium industry’s supply chain. This re-
search attempts to simulate the internal supply chain of a primary aluminium casthouse
and identify the wastes by implementing a lean thinking approach. After highlighting
the possible improvements, optimization models attempt to reduce these wastes which
create non-value added costs to the system. This concept is further developed by inter-
facing the simulation model with the optimization model to validate the improvements.
The success of the concept is tested by measuring the reduction in redundant logistical
costs of a case study founded on the real casthouse specifications. Scenarios are defined
to analyze the casthouse supply chain under different perspectives. The potential gain of
the new concept is verified by applying it to these scenarios. In conclusion, the results
analysis of the scenarios indicates the success of the main objective of this research; to
develop a new concept that controls the non-value added logistical costs in the primary
aluminium casthouse supply chain.
Keyword: Primary aluminium casthouse, supply chain analysis, lean thinking, logistics
simulation, linear optimization
II
This thesis is dedicated to
My Daughter
&
My Wife
III
Acknowledgements
Acknowledgements
I would never have been able to finish my dissertation without the guidance and pa-
tience of my advisor, my colleagues, my friends and my family.
Firstly, I would like to express my deepest gratitude to my advisor, Prof. Dr.-Ing. Bernd
Noche, for the excellent guidance and valuable discussions. I would like to express my
great appreciation to the Technology Department of Hydro Aluminium Deutschland
GmbH, where I have been working for more than 5 years. During this period, I learned
a lot about the aluminium industry and also strengthened my personal and professional
skills.
I would especially like to thank my colleagues Dr. Martin Segatz, Dr. Ingo Eick, Gregor
Bellinghausen, Stefan Jedeck and my friends Dr. Demet Cetiner and Dr. Sakine Batun
for their support and time that they spent proof reading. I would like also to thank
Simone Lehr, she always did her best to help me when I asked for support. Special
thanks should be given to Dr. Christian Droste, for his useful and constructive recom-
mendations and invaluable support on this research. I would like to express my great
appreciation to Dr. Anton Winkelmann, who worked very closely with me and helped to
puzzle over the problems that I was faced with.
I wish to thank my sister, my mother and my father for their support and encouragement
throughout my study.
Finally and most importantly, I would like to express my great thanks to my daughter,
Nisan, my princess. After studying long hours, she made me feel alive by wearing a
smile, or giving meaning to the most valuable word, “Baba”. And most of all to my
wife, Emine, my love, she never gave up standing by me through my bad and good
times. Her support and encouragement was in the end what made this dissertation pos-
sible.
Thank you.
IV
Table of Contents
Table of Contents
Abstract ......................................................................................................................................... I
Acknowledgements.................................................................................................................... III
Table of Contents ...................................................................................................................... IV
List of Figures ...........................................................................................................................VII
List of Tables ............................................................................................................................. IX
1 Introduction......................................................................................................................1
1.1 Introduction to Fundamental Concepts of the Study .................................................2
1.1.1 Supply Chain in General .......................................................................................2
1.1.2 Aluminium Smelter Casthouse ..............................................................................4
1.1.3 Discrete Event Simulation .....................................................................................6
1.1.4 Linear Optimization ..............................................................................................8
1.2 Research Motivation ...................................................................................................9
1.3 Research Objectives .................................................................................................10
1.4 Outline of the Thesis.................................................................................................10
2 Literature Review ..........................................................................................................13

2.1 Introduction ..............................................................................................................13
2.2 Literature about Supply Chain Simulation ...............................................................13
2.3 Literature about Supply Chain Optimization............................................................16
2.3.1 Lot Sizing Problem Approach .............................................................................18
2.3.2 Lean Thinking Approach.....................................................................................18
2.4 Literature about the Interfacing Simulation and Optimization Approaches .............20
3 Casthouses in Primary Aluminium Smelter ................................................................24

3.1 Introduction to Aluminium Production ....................................................................24
3.2 Primary Aluminium Production ...............................................................................28
3.3 Primary Aluminium Casthouse Supply Chain..........................................................30
3.3.1 Material and Information Flow in the Casthouse Supply Chain .........................31
3.3.2 Activities and Their Interaction in the Casthouse Supply Chain .........................35
3.3.3 Casting Process and Possible Product Range ......................................................38
V
Table of Contents
4 Simulation Model Development for Aluminium Casthouse ......................................41

4.1 Introduction and Literature Review ..........................................................................41
4.2 Boundary of the Simulation Model ..........................................................................42
4.3 Concept of the Simulation Model .............................................................................43
4.4 Main Components of the Simulation Model ............................................................46
4.4.1 Primary Hot Metal Distribution...........................................................................46
4.4.2 Secondary Hot Metal Transport ..........................................................................50
4.4.3 Scrap and Cold Metal Transport..........................................................................51
4.4.4 Furnace Operations..............................................................................................52
4.4.5 Down Stream Processes after Casting .................................................................55
4.5 Logistical Control Strategies in the Simulation Model ............................................55
4.5.1 Possible Scenarios for Primary Crucible Allocation ...........................................55
4.5.2 Traffic Management in the Simulation Model ....................................................57
4.5.3 Furnace Management ..........................................................................................64
4.5.4 Storage Management Concepts ...........................................................................66
4.5.5 Scrap & Cold Metal Filling Strategies ................................................................69
5 Lean Thinking Approach in the Primary Aluminium Casthouse Supply Chain.....71

5.2 Waste Analysis in the Casthouse from a Logistical Point of View ..........................72
5.2.1 Overproduction in the Casthouse Supply Chain..................................................73
5.2.2 Waiting Time in the Casthouse Supply Chain.....................................................74
5.2.3 Unnecessary Transportation in the Casthouse Supply Chain ..............................77
5.2.4 Inefficient Processing in the Casthouse Supply Chain ........................................78
5.2.5 Inventory of the Casthouse Supply Chain ...........................................................78
5.2.6 Unnecessary Movement in the Casthouse Supply Chain ....................................80
5.2.7 Defective Products in the Casthouse Supply Chain ............................................80
5.3 Summary of Waste Analysis ....................................................................................81
6 Optimization Model Development ...............................................................................82

6.2 Model Structure ........................................................................................................84
6.3 Short-term Production Planner .................................................................................86
6.3.1 Aim of the Model ................................................................................................86
6.3.2 Objective Function ..............................................................................................87
6.3.3 Equations of Costs in the Objective Function of the Short-term Production
Planner ................................................................................................................90
6.3.4 Equations of the Conversion Coefficients ...........................................................92
6.4 Production Scheduler................................................................................................93
6.4.1 Aim of the Model ................................................................................................93
VI
Table of Contents
6.4.2 Objective Function ..............................................................................................94

6.4.3 Equations of Costs in the Objective Function of the Production Scheduler ........98
6.4.4 Equation of the Conversion Coefficient ............................................................100
6.5 Interfacing with the Simulation Model ...................................................................100
6.5.1 Methodology of the Interfacing .........................................................................101
6.5.2 Architecture of the Interface ..............................................................................103
7 Description of the Case Study .....................................................................................106

7.1 Casthouse System Characteristics ..........................................................................106
7.1.1 Layout Description ............................................................................................108
7.1.2 Storage Specifications .......................................................................................109
7.1.3 Vehicle Fleet......................................................................................................111
7.1.4 Casting Furnace Specifications .........................................................................113
7.2 Parameter Set for the Simulation Model ................................................................114
7.2.1 Primary Hot Metal Flow Specifications ............................................................115
7.2.2 Material Types and Storage Descriptions..........................................................116
7.2.3 Simulation Model Items in Automod ................................................................118
7.3 Parameter Set for the Optimization Models ...........................................................119
7.3.1 Parameters for the Short-term Production Planner ............................................119
7.3.2 Parameters for the Production Scheduler ..........................................................120
7.3.3 Parameters for the Coefficients of Optimization Models ..................................122
8 Scenarios & Evaluation ...............................................................................................123

8.1 Scenario I: Fixed Production Amount ....................................................................123
8.1.1 Analysis of Scenario I with and without Optimization Models ........................124
8.1.2 Detailed Analysis of Scenario I with and without the Short-term Production
Planner ..............................................................................................................129
8.1.3 Detailed Analysis of Scenario I with and without the Production Scheduler ....130
8.2 Scenario II: Variable Production Amount ..............................................................136
8.2.1 Definition of Scenario II....................................................................................137
8.2.2 Analysis of Scenario II with and without the Production Scheduler Tool ........137
9 Research Summary & Future Work ..........................................................................140

9.1 Research Summary .................................................................................................140
9.2 Recommendations for Future Work .......................................................................142
References .................................................................................................................................144
VII
List of Figures
List of Figures
Figure 1 Different supply chain definitions and their boundaries [MDKMNSZ01] ....................................... 3
Figure 2 Primary aluminium production capacity in the last 10 years [HKK07] ........................................... 5
Figure 3 Aluminium production and life cycle [IAI03] .................................................................................. 6
Figure 4 Automod simulation software ..................................................................................................... 16
Figure 5 Literature review on supply chain optimization (extracted from [Bea98]) .................................. 17
Figure 6 Architecture and the main process flow in the verified simulation platform [KPM04] ................ 22
Figure 7 Coordination between simulation and optimization [GKL99] ...................................................... 22
Figure 8 OpenSolver optimization engine .................................................................................................. 23
Figure 9 Global production of primary aluminium until 2010 [EAA12] ...................................................... 25
Figure 10 Steps of aluminium production from bauxite to aluminium [Hoe04] ........................................ 25
Figure 11 Global production of bauxite in 2000 (%) [Fri04] ....................................................................... 26
Figure 12 Global production of aluminium oxide in 2003 (Mt) [Hoe04] .................................................... 26
Figure 13 Location of aluminium smelters [IAI03] ..................................................................................... 27
Figure 14 Schematic drawing of an alumina reduction cell [GK93] ........................................................... 28
Figure 15 Material flow and sub facility boundaries in primary aluminium supply chain [HKK07] ........... 29
Figure 16 Internal supply chain of primary aluminum casthouse .............................................................. 30
Figure 17 Material and information flow in casthouse internal supply chain ........................................... 32
Figure 18 Activities and their interactions in casthouse supply chain........................................................ 36
Figure 19 End products of primary aluminium casthouse [Kva11] ............................................................ 40
Figure 20 Casthouse simulation model structure....................................................................................... 43
Figure 21 Distribution of the crucible transport in one shift ...................................................................... 50
Figure 22 Category and type classification example.................................................................................. 52
Figure 23 Process cycle of casting furnace in the simulation model .......................................................... 53
Figure 24 The new navigation tool implementation to guide vehicles ...................................................... 58
Figure 25 Possible two layers node connection ......................................................................................... 59
Figure 26 An example to show the selection logic ..................................................................................... 60
Figure 27 Information flow in the decision logic behind the navigation tool............................................. 62
Figure 28 Layers of the zone management approach in front of the casting furnace ............................... 63
Figure 29 Control logic of the furnace management after the holding process ........................................ 66
Figure 30 Strategy behind the storage management concept .................................................................. 67
Figure 31 Furnace filling strategy (left) without and (right) with door buffering ...................................... 70
Figure 32 Furnace filling strategy with a box ............................................................................................. 70
Figure 33 Example utilization graph of a casting furnace in percentage ................................................... 76
Figure 34 An example utilization graph of a transportation unit in percentage ....................................... 77
Figure 35 Optimization platform in MS Excel ............................................................................................ 84
Figure 36 Snapshot of the first optimization tool ...................................................................................... 87
VIII
List of Figures
Figure 37 Snapshot of the second optimization tool.................................................................................. 93

Figure 38 Another snapshot of the second optimization tool .................................................................... 94
Figure 39 Data transfer between simulation and optimization models .................................................. 101
Figure 40 Methodology of the interface between the simulation and optimization models ................... 103
Figure 41 Layers of application fields in the interface structure .............................................................. 104
Figure 42 Layout of the casthouse defined for the case study ................................................................. 109
Figure 43 Areas according to transported material ................................................................................. 113
Figure 44 Simulation model items described in Automod........................................................................ 118
Figure 45 The duration of processes and their allocation for each furnace group in a shift .................... 122
Figure 46 Stepwise visualization of the analysis of Scenario I ................................................................. 125
Figure 47 Non-value added logistical cost supplied from two days simulation run ................................. 127
Figure 48 Analysis of cost items in the objective function of tool 1 (the short-term production planner)130
Figure 49 Analysis of cost items in the objective function of tool 2 (the production scheduler) .............. 131
Figure 50 Process cycle of furnaces as output of the simulation model without optimization models
(reference case)........................................................................................................................................ 133
Figure 51 Process cycle of furnaces as output of the simulation model after interfacing with optimization
model 2 (Step 2.4) .................................................................................................................................... 133
Figure 52 Hot metal distribution in the last three hours of shifts supplied from the simulation model
without optimization models (reference case) ........................................................................................ 134
Figure 53 Hot metal distribution in the last three hours of shifts supplied from the simulation model after
interfacing with optimization model 2 (Step 2.4) .................................................................................... 134
Figure 54 Batch lead times of furnaces supplied from the simulation model without optimization models
(reference case)........................................................................................................................................ 135
Figure 55 Batch lead times of furnaces supplied from the simulation model after interfacing with
optimization model 2 (step 2.4) ............................................................................................................... 135
Figure 56 Scenario comparison with and without optimization model 2 (the production scheduler) .... 139
IX
List of Tables
List of Tables
Table 1 The major objectives of supply chain studies [BADW04] ................................................................ 4

Table 2 Some commercial software packages [Fu01] ................................................................................ 21
Table 3 A comparative evaluation of commonly used metals [Jai03] ........................................................ 28
Table 4 An example of a furnace status tracing in the simulation model .................................................. 75
Table 5 Categorization of wastes depending on improvement possibility in the scope of the optimization
model ......................................................................................................................................................... 81
Table 6 Production units in the casthouse ............................................................................................... 107
Table 7 Annual and daily required hot metal and production capacity of the casthouse ........................ 108
Table 8 Vehicle fleet of the casthouse ..................................................................................................... 112
Table 9 Some specifications of the casting furnaces ................................................................................ 114
Table 10 Electrolysis specifications .......................................................................................................... 115
Table 11 Hot metal flow characteristics .................................................................................................. 116
Table 12 Material types and their properties .......................................................................................... 117
Table 13 Storage descriptions and their properties ................................................................................. 117
Table 14 Pre-assumed process durations of the furnace groups ............................................................. 121
Table 15 Required parameters for coefficients of optimization models .................................................. 122
Table 16 Optimization and simulation model results with and without optimization tools .................... 126
Table 17 Detailed cost analysis with optimization model 1 (the short-term production planner)........... 129
Table 18 Detailed cost analysis with optimization model 2 (the production scheduler) .......................... 131
Table 19 The mean value, standard deviation and RSD values for both steps ........................................ 136
Table 20 Results for scenario I with and without optimization model 2 (the production scheduler) ....... 138
Table 21 Results for scenario II with and without optimization model 2 (the production scheduler) ...... 138
1 Introduction
Before the financial crisis in 2008, the ultimate goal of industry was to increase the
sales and also the production amount. Nowadays, the direction of the storm changes to
reduce the operational expenditure to find a place in the shrinkage of the market share
with lower sales price.
The aluminium industry has also been influenced from this unstable economic situation.
Tremendous decrease of demand in the automotive industry increased the stock levels
of aluminium in the last four years. And also the gap between high supply and low de-
mand reduced the sales price of aluminium.
In addition to effects of the financial crisis, CO2 tax regulations due to its environmental
impact increased the energy prices day by day. Besides that, inevitable growth of alu-
minium production in China creates big challenges in the aluminium industry, especial-
ly in Europe.
The technological developments in aluminium industry focus on how to reduce the pro-
duction costs. Energy consumption forms the main part of these costs due to high ener-
gy prices. Therefore, the production process is tried to be optimized so that it consumes
less energy. However it is also recognized that logistical activities in the facilities have a
potential for improvement.
The main focus of this research is the development of a concept that controls the logisti-
cal activities creating extra unpredictable costs in a part of the primary aluminium sup-
ply chain. In a smelter concept, the electrolysis first comes into the mind because the
actual production process takes place in this sub facility. However, the selected unit of
smelter for this research is the casthouse area due to its potential for improvement in
logistical perspective and its direct contact to the external customer which brings more
challenge for the investigation.
2
Introduction
In this chapter, at the first, the fundamental concepts of the thesis are discussed, then,
the research motivation and the objective of the research are described. Finally, the out-
line of the thesis is presented.
1.1 Introduction to Fundamental Concepts of the Thesis

The fundamental concepts of the thesis contain
 supply chain as a focus area,
 primary aluminium casthouse as a core concept,
 discrete event simulation as an analyzing tool and
 linear optimization as an improvement method.
Definitions and introduction to these basic concepts are given in this subchapter.
1.1.1 Supply Chain in General

Business interest has been aroused toward supply chain concept since the 1980’s after
recognizing its benefits on the company performance according to Braziotis [Bra10].
The term was ill-defined before 1990’s. Academic researchers and the companies have
approached to supply chain delicately since 20 years. It has become a crucial concept
for industry to survive in the competitive environment [ARA10].
Beamon [Bea98] defined supply chain as “an integrated process wherein a number of
various business entities work together in an effort to acquire raw materials, convert
these raw materials into specified final products, and deliver these final products to
retailers”.
Mentzer et.al. [MDKMNSZ01] enlarged the definition by adding service, finance and
information in addition to product flow. They defined supply chain as “a set of three or
more entities (organizations or individuals) directly involved in the upstream and down-
3
Introduction
stream flows of products, services, finances, and/or information from a source to a cus-
tomer.”
The term “supply chain” has still various definitions depending on its boundary. Figure
1 shows different supply chain definitions and their boundaries. Direct supply chain
contains organization with its direct customers and suppliers. In this thesis, the supply
chain concept refers to material and information flow of the center facility. Both of the
flows start from the direct supplier and end with the direct customer. This kind of sup-
ply chain is also called internal supply chain by Harland [Har99].
Extended version of the supply chain enlarges the boundary with supplier’s supplier and
customer’s customer. Financial provider supplying financial support and evaluating the
possible financial risks, market search firm providing detailed analysis about ultimate
customers and third party logistics (3PL) handling the logistical activities between or-
ganization and its customer are included in the ultimate supply chain boundary by
Mentzer et.al. [MDKMNSZ01].
Figure 1 Different supply chain definitions and their boundaries [MDKMNSZ01]
According to Buxmann et. al [BADW04] the major objectives of supply chain studies
are listed in Table 1. Most of these objectives, such as reduction in lead time, in inven-
tory, transportation costs and improvement in service levels, create the focus area of
supply chain optimization concept of the research.
4
Introduction
Major Objectives Authors and Year of the Study

Heizer & Render (1999),
Morton (1999),
Reduction in production costs Shim & Siegel (1999),
Waller (1999),
Stevenson (2002)
Reduction in lead time Markland et al . (1998)
Baganha & Cohen (1998),
Ho (1999),
Waller (1999),
Reduction in inventory
Helms et al . (2000),
Simchi-Levi et al . (2000),
Toomey (2000)
Ballou (1999),
Reduction in transportation costs
Toomey (2000)
Reduction in purchase costs
Monczka et al . (2001)
Improvement in supplier evaluation Christopher & Jüttner (2000),
and selection Monczka et al . (2001)
Improvement in service levels Knolmayer et al . (2002),
Tan (2002)
Improvement in cooperation Lee et al . (1997)
Table 1 The major objectives of supply chain studies [BADW04]
1.1.2 Aluminium Smelter Casthouse

Aluminium cannot be found as an element in nature due to its chemical characteristics.
Its raw material alumina is refined from bauxite which is extracted from the earth. Al-
uminium has been produced since the beginning of 1900’s in industrial range. However,
the total demand of the industry is increasing continuously. Figure 2 shows the primary
aluminum production capacity per country and the capacity increase in the last 10 years
which is around “15 Mt/a”.
5
Introduction
Figure 2 Primary aluminium production capacity in the last 10 years [HKK07]
Aluminium is mostly used in industrial fields such as construction, transportation, pack-

aging etc. The increase in aluminium demand is driven by its advantageous properties
some of which are listed as:
• Strong and light
• Highly corrosion resistant
• Good conductive
• Easy to form and process
• Recyclable
Aluminium production can be categorized into primary and secondary smelting. Sec-
ondary aluminium production is also called recycling. The steps in aluminium produc-
tion and its life cycle are shown on Figure 3. In this thesis, the focus is on primary
smelting part in which aluminium is produced by processing alumina.
The facility called smelter is normally composed of three sub facilities:
• Carbon plant; supplies anodes to the electrolysis unit
• Electrolysis (pot room); contains reduction cells where the primary aluminium is
produced
6
Introduction
• Casthouse; casts the hot metal transported from the electrolysis as ingots for fur-
ther processing.
Figure 3 Aluminium production and life cycle [IAI03]
The casthouse facility contains casting furnaces which enable mixing the metal with
various alloys to meet the customer specifications. Casting furnaces may also melt the
aluminium scrap depending on the quality requirements. The casthouse stands at the last
chain of the primary aluminium production. It supplies primary hot metal from electrol-
ysis, alloy and cold metal (pure aluminium ingots and scrap metal) from the suppliers.
Possible customers of the casthouse are rolling or extrusion plants which process the
aluminium ingots or slabs for further manufacturing industries.
1.1.3 Discrete Event Simulation

Simulation is imitating a real-world process or system by analyzing its behavior asking
“what-if” questions [Ban04]. In this research, the system represents a group of entities
such as vehicles, resources, materials etc. The dynamic simulation model concept is
split into two sub groups which are called continuous flow simulation and discrete event
simulation.
7
Introduction
A discrete event simulation differs from continuous flow simulation by sequencing the
events according to their occurrence time which requires having discrete simulated time
points. According to Schriber et. al. [SB10] these event happening times may overlap at
the same time, and then these “simultaneous” movements of traffic are simulated by
manipulating this train of events serially at that instant. Continuous flow simulation in
that context represents the cases mapping the flow of liquid materials. It is possible to
approach continuous flow problems with discrete event concept which brings some lim-
itations described by Damiron et. al. [DN08] as:
• Execution speed of the model slows down due to the many events created to
map the level of flow discretely
• Model becomes more complicated than necessary due to the definition of many
events
• The accuracy of the system never reaches 100% due to the prediction of contin-
uous flow change over time
Mason et. al. [MRFK03] discussed the possibility of information flow mapping next to
the flow of goods in the models, which increases the range of discrete event simulation
usage. Babulak et. al. [BW10] (quoted from Wang, Sun and Nooh (1995)) listed the
application fields of discrete event simulation as:
• New manufacturing processes, at the design and evaluation phase
• Existing processes, to improve the performance
• Operational area, to establish optimum policies
• Production field, to support planning and scheduling
In this research, the discrete event simulation method is selected to analyze the cast-
house supply chain.
8
Introduction
1.1.4 Linear Optimization

Linear optimization is a mathematical method which specifies the optimum solution of
the problem that is formulated in linear equation. The aim of the linear optimization is
to find the maximum or the minimum value of the equation in the defined feasible re-
gion. The elements in the linear optimization are defined by Chinneck [Chi01] as:
• Variables: represent the items that can be adjusted to find the optimum solution
of the linear problem
• Objective function: sets the goal of problem by combining the variables. Its
mathematical equation must be in linear form.
• Constraints: draw the limits of the feasible region. These mathematical expres-
sions have to be linear.
• Variable bounds: represents the range of values of the variables
Standard representation of linear optimization is [Fer12]:
max c1 x1  c 2 x 2  ...  c n x n
(1.1)
subject to a 11x1  a 12 x 2  ...  a 1n x n  b1
...
a m1 x1  a m2 x 2  ...  a mn x n  b m
(1.2)
and x1  0,..., xn  0
(1.3)
The objective function is represented as maximization problem in Equation (1.1). This

is the primal definition of the problem. Each linear programing problem has duality. In
this case the dual problem is expressed as minimization of Equation (1.1) by changing
9
Introduction
the coefficients of the constraints, the inequalities of which are shown in Equation (1.2).
The variables of the objective function and their bounds are formulated in Equation
(1.3).
Several methods are developed to solve the linear programming problem. The algo-
rithms developed for linear programming (such as the simplex algorithm, the ellipsoid
algorithm, interior-point method etc.) are defined, compared and examined in detail in
the literature [DD09], [Tod01], [HMSW53] and [Meg91]. In this research, the integer
linear method is selected as a solution algorithm for the linear programming in which all
of the variables must have integer values.
1.2 Research Motivation

During more than four years of professional experience in one of the world’s three lead-
ing integrated aluminium companies, it is recognized that the technological focus in the
aluminium industry is on the electrolysis unit. There are several valid reasons for that,
such as:
• Electrolysis has still potential in its production division for improvement
• Energy consumption in electrolysis unit forms ~ 25 % of the production cost
• Complex operation techniques
The author has been working as a project engineer in the technology department and is
responsible for development of simulation models. These studies enable him to analyze
each operational field in aluminium smelter. Possible improvement in the casthouse
from the logistical point of view aroused his interest.
As mentioned, the trend in aluminium industry is to reduce the production costs rather
than increase the production amount due to actual market conditions. Therefore, the
focus of the research should have been to reduce the redundant logistical costs in alu-
minium casthouse facility. Lack of literature on this concept and the potential on im-
provement of logistical activities in casthouse have generated strong motivation.
10
Introduction
1.3 Research Objectives

The main objective of this research is to develop a new concept that controls the non-
value added logistical costs in primary aluminium casthouse supply chain. The follow-
ing objectives have to be accomplished in this research to reach this goal:
• Identification of material and information flow in the primary aluminium cast-

house internal supply chain.
• Drawing of the supply chain boundary by considering the close interaction with
the possible suppliers and customers.
• Development of flexible discrete event simulation model which can be imple-

mented to most casthouses easily by modifying some parameters.
• Analysis of casthouse supply chain with the simulation model and identification
of non-value added logistical costs.
• Discovering the reasons of the non-value added logistical costs by approaching

to the concept with the lean thinking method.
• Development of linear optimization model reducing these redundant logistical

costs by planning the production and scheduling the processes in the casthouse.
• Interfacing the simulation model with the optimization model to increase the ac-
curacy of the optimization part.
• Verifying the gain from the new concept by implementing it to the case study
founded on the real casthouse specifications
1.4 Outline of the Thesis

The outline of the thesis follows a similar path like the sequence of steps in a simulation
model programming. Firstly, previous studies are examined to define the problem in a
better way, secondly the data about the studied system is collected, and then the pro-
11
Introduction
gramming phase is executed. As a next step, a case study to test the models is prepared
for verification and then the evaluation phase takes place. Finally, the conclusion of the
results is discussed and the further follow-ups are defined.
Chapter 1 starts with giving a short description of the fundamental concepts of the the-
sis. The motivation for the concept is described and the objective of the thesis is set. At
the end of the chapter the roadmap of the thesis is drawn and the content of each chapter
is shortly explained.
Chapter 2 contains the literature survey on the main concepts of the thesis. The research
topics are supply chain simulation, optimization and possible interface between these
two models. Lean thinking approach and lot sizing problem definition can also be found
in this chapter.
Chapter 3 provides extensive knowledge about the primary aluminium production. The
smelter concept and its units are introduced. This chapter also contains the description
of a primary casthouse and its internal supply chain which is the core concept of the
research.
Chapter 4 explains the simulation model mapping of the material flow in primary alu-
minium casthouse. The strategies behind the control logic and the main parts of the
simulation model are elaborated. The programming approach enabling to handle the
problem in the simulation environment is explained in detail.
Chapter 5 highlights the logistical wastes of the casthouse supply chain by evaluating
the results gathered from the simulation model. The lean thinking approach application
to the primary aluminium casthouse and its possible outcomes are explained in this
chapter.
Chapter 6 contains the optimization part of the thesis. The optimization models created
to reduce the non-value added costs in the casthouse internal supply chain are presented.
The interface between the simulation model and the optimization tools are also ex-
plained in this chapter.
12
Introduction
Chapter 7 expresses the case study created to verify and test the models and their inter-
facing. All of the required details of the case are highlighted and also assumptions for
the models are listed.
Chapter 8 evaluates the results of both the simulation and the optimization models. Sce-
narios analyzing the casthouse supply chain in different perspectives are described and
the results obtained for each scenario are compared. Possible conclusion is drawn for
each case that is created to reduce the non-value added costs occurring due to logistical
activities in the casthouse.
Chapter 9 concludes the study by reminding the problem and presenting the outcome of
the evaluations. Additionally, recommendations for further development alternatives are
also listed in this chapter.
13
Literature Review
2 Literature Review
In this chapter, literature about main concepts of this thesis is reviewed. This includes
simulation studies done for different material flow principles, supply chain optimiza-
tion, possible approaches for the improvement phase and interfacing methods of simula-
tion and optimization models.
2.1 Introduction
A plenty of literature exist where simulation and optimization concepts are integrated
into supply chain investigation. Nowadays, simulation tools provide opportunities to
make parameter variations during the simulation run. These investigations are mostly
based on sensitivity analysis of the selected variable or variables of the system. Existing
approaches combining simulation and optimization are reviewed in the thesis.
Environmental aspects and challenges in technological development force the metal

industry to optimize its supply chain to stay competitive. Therefore, as an application
field not only aluminium but also other metal production industries (like steel foundry)
are investigated during this review. Casthouse has a common concern in the metal pro-
duction because the casting process and the furnace show similarities in different indus-
trial fields. A specific literature review of simulation and optimization applications to
casthouse field is discussed at the beginning of their corresponding chapters (Chapter 4
and 6).
2.2 Literature about Supply Chain Simulation

Supply chain simulation is an analyzing method that experimentally shows the system
characteristics of supply chain. The possible objectives of supply chain simulation are
defined by Campuzano et. al. [CM11] as:
14
Literature Review
 Providing supply chain knowledge
 Giving ideas to improve the system
 Creating and verifying the alternative strategies
 Quantifying improvements of supply chain management
Ganeshan et. al. [GH95] integrated comprehensive supply chain in the application fields
of simulation by paying attention to strategic and operational elements. Within the con-
text of analyzing complex supply chains via simulation approach, the following ad-
vantages were summarized by Tompkins et.al [TS98]:
 Simulation helps to design and analyze the complex internal interactions of

specified or existing systems.
 Before implementing the material handling equipment on the actual system, the
effects of the changes can then be tested via simulation.
 By creating different scenarios, simulation is used to answer the “what if” ques-
tions in analysis and design.
 There are risks to make experiments on the real system; thus the simulation
model can help to prepare decision rules and policies about how to operate the
real system.
 The time frame can be changed in the simulation approach, by compressing or

expanding.
 The bottleneck of the production facility can be seen identified by the help of the
simulation model.
 The statistical analysis of introducing new machines to the system or changing

the current production line can be obtained.
15
Literature Review
 With some simulation software the 3D view of the model environment can be
presented.
After recognizing the benefits of simulation in supply chain analysis, a lot of research
and study were done. Banks et. al. [BBJLM02] discussed the opportunities of simula-
tion in supply chain analysis and the possible application fields in detail. Zeigler et. al.
[ZKB99] described the benefit of supply chain simulation as helping companies to de-
termine which strategies will provide the most flexible and profitable operating envi-
ronment. Capability of capturing uncertainty and complexity [JWCEL01] enables to
analyze any kind of system with simulation. Kovacs et. al. [KVKMP03] listed the pos-
sible types of uncertainties mapped in simulation environment as:
 Start time uncertainty
 Downtimes
 Processing time
 Re-work and adjustment
 Quality uncertainty
Kleijnen [Kle05] described the characteristics of the simulation models as; quantitative,
mathematical and dynamic computer models having at least one equation with one vari-
able.
Commercial simulation tools with different features and application areas have been
released in recent years. They have been used widely in the supply chain research.
Some examples of these software and studies done by using them are; SimFlex
[WG03], Arena [SB99], e-SCOR [BM00], ProModel [Ben97], LOGSIM [Hie98], Wit-
ness [WHP98], SISCO [CHH06]. In this thesis the simulation tool Automod, the de-
tailed information of which is shown in Figure 4, is used to analyze the internal supply
chain of aluminium casthouse. The type of the simulation model in this study can be
considered as a production and distribution based simulation approach as one of the
simulation approaches classified by Tumay [Tum96].
16
Literature Review
Figure 4 Automod simulation software
2.3 Literature about Supply Chain Optimization

Optimal supply chain becomes challenging and demanding for nearly each industry
which is under pressure to stay competitive in the global markets. In optimization task,
the aim is to find the best solution and decision among several alternatives. Voss et.al.
[VW03] analyzed the supply chain optimization in detail and described possible meth-
ods to overcome the planning problems occurred in the production phase.
Beamon [Bea98] made a literature survey about the tools and their criteria to evaluate
and improve the supply chain, which can be seen on Figure 5. Lakhal et. al. [LMKO01]
described a supply chain strategy which maximizes the value added in internal activi-
ties.
17
Literature Review
Model Performance Decision

Types Measures Variables
Production/Distribution
Deterministic Analytical
Stochastic Analytical
Inventory Levels
Plant- Product
Activitiy Time
Assignment
Simulation
Scheduing
Economic
Flexibility
Cost
Author(s)
Altiok and Ranjan (1995) X X X
Amtzen, et. al. (1995) X X X X X X
Camm, et. al. (1997) X X
Cohen and Lee (1988) X X X
Cohen and Lee (1989) X X X X
Cohen and Moon (1990) X X X
Ishii, et. al. (1988) X X X X
Lee and Billington (1993) X X
Lee and Feitzinger (1995) X X
Lee, et. al. (1993) X X X
Lee, et. al. (1997) X X X
Pyke ynd Cohen (1993) X X X
Pyke ynd Cohen (1994) X X X
Newhart, et. al. (1993) X X X
Svoronos and Zipkin (1991) X X X
Towill (1991) X X X
Towill, et. al. (1992) X X X
Towill and Del Vecchio (1994) X X
Tzafestas and Kapsiotis (1994) X X X
Voudouris (1996) X X X X
Williams (1981) X X X
Williams (1983) X X X
Wikner, et. al. (1991) X X X
Figure 5 Literature review on supply chain optimization (extracted from [Bea98])
Ten Hompel et.al. [HS07] classified the optimization problem as;
 Maximization problem; the best result with the defined expenditure
 Minimization problem; the defined result with the minimum expenditure
 Dual problem; the optimal result with the possible expenditure

18
Literature Review
In this study the aim of optimization is to create an optimal schedule for an aluminium
casthouse by minimizing the non-value added logistical costs. Planning is handled as a
lot sizing problem while distributing customer orders to the batches. This dispatching
process is done by combining the methods of order- and resource based dispatching
which is classified by Sauer [Sau02]. Additionally the lean thinking approach (dis-
cussed in Chapter 2.3.2) is used to eliminate or to minimize the non-value added logisti-
cal costs in the internal aluminium casthouse supply chain by scheduling the operations
to the casting furnaces.
2.3.1 Lot Sizing Problem Approach

A lot-sizing problem was formulated by Wagner et.al. [WW58] as a dynamic single
item economic problem determining minimum cost production to satisfy the demand
[GAN06]. Summerauer [Sum08] identified the costs as inventory, setup and production
costs which have direct impact on a company’s efficiency and competitiveness in the
market. Peres et.al. [PBNN02] described the lot-sizing problem as a planning problem
which aims to determine the production time and the amount. The lot-sizing problem
has aroused the interest for the production planning. Drexl et.al. [DK97] made a com-
prehensive literature survey about the lot-sizing and also scheduling problem.
Potts et.al. [PW92] referred the benefit of lot–sizing problem on splitting the lot of iden-
tical items into sub-lots. Duda et.al. [DO05] applied the lot-sizing problem to produc-
tion planning of steel foundry and Karmarkar et.al. [KKKF85] analyzed the analytical
solution approach to the lot-sizing problem with simulation model.
In this research, the lot-sizing problem stays as the backbone of the planning. Its ap-
proach on cost reduction and batch creation from customer orders plays an important
role at the optimization part of this study.
2.3.2 Lean Thinking Approach

Lean thinking approach can be seen as a main part of the Toyota Production System
which firstly appeared in 1950’s. To understand the complete Toyota production ap-
19
Literature Review
proach, literature [Ohn88] and [Shi89] are recommended. The Industrial Technology
Center [ITC04] described lean thinking as a method aiming to improve productivity,
efficiency and quality of the system.
In the lean thinking principle, activities are separated into two categories; “value-added”
and “non-value added”. These terms are introduced, well defined and distinguished
from each other in lean thinking approach which is highlighted by Womack et.al.
[WJR91]. Koning et. al. [KVHBD06] identified the distinction between value added
and non-value added activities as value added activities contributing to what the cus-
tomer requests from a product or service. All other processes are defined as non-value
added activities.
The main focus of lean thinking is “to eliminate the waste” from the system. Ohno
[Ohn88] classified the waste (or Japanese word for waste “muda”) in the production
system as:
 Waste of overproduction
 Waste of time on hand (waiting)
 Waste in transportation
 Waste of processing itself
 Waste of stock on hand (inventory)
 Waste of movement
 Waste of making defective products
This approach is implemented to different sectors (like health care [UAKEN07]) and
concepts (like supply chain management). Lean thinking in supply chain management
described in detail by Wisner [Wis11] and Rother et.al. [RS03] provides an understand-
ing of the benefits of mapping the material and information flow to reduce wastes in the
system.
20
Literature Review
Value stream mapping, process mapping, spaghetti diagrams, and simulation are well
known tools that analyze the flow and help to implement the lean approach to the facili-
ties. Among these tools, according to Detty et.al. [DY00], simulation is capable of
showing the benefits of lean manufacturing including:
 Stock levels
 Transportation requirements
 The utility of production scheduling system
 Detailed delivery analysis from suppliers
 Detailed shipment analysis to customers
In this thesis, reduction of non-value added logistical costs is the goal of the optimiza-
tion part.
2.4 Literature about the Interfacing Simulation and

Optimization Approaches
A simulation model is imitating the real system by using defined logic without making
any attempt to find the best option or choice. It provides the results according to the
predefined strategies and control logic. On the other hand, optimization identifies the
best decisions without paying attention to the dynamic interaction between the circum-
stances. A new approach in the industry is to combine these two beneficial tools to get
one package. Fu [Fu01] listed some commercial simulation and optimization packages
which are shown on Table 2. Also according to Fu [Fu02], these tools are capable of
searching the optimum parameters selected by user with respect to the performance in-
dicators. They use mathematical programming to find the best combination but the user
does not have any control on this part of the software. It can be concluded that these
tools perform sensitivity analysis on the selected variables in the applied field.
21
Literature Review
Table 2 Some commercial software packages [Fu01]
Simulation and scheduling system have also studied to be interfaced with each other. A
new wording has appeared in recent days, the “simulation-based scheduling” which has
these two approaches in focus. Miller et.al. [MP00] defined the term simulation-based
scheduling and the selection rules behind it. They classified the selection rules as:
 Operation selection rule: Applicable when there is an available resource for sev-
eral operations at an instant time
 Resource selection rule: Applicable when there is an operation which can be

processed by several available resources at an instant time
However this approach deals with dispatching of tasks to resources. The problem of
production planning and scheduling contains several constraints and parameters, so it is
not only an allocation problem. Ganapathy et. al. [GNS03] described another approach
to combine production planning and scheduling with simulation. This approach is to
create an emulation platform with interfacing decision support system and simulation
model. Kadar et.al. [KPM04] sketched the process flow in such a platform which is
shown on Figure 6. Gupta et.al. [GS02] compared the typical simulation and simulation
with production planning control studies. In this comparison they focused on scope,
model, experiment and output of both platforms.
22
Literature Review
Figure 6 Architecture and the main process flow in the verified simulation platform [KPM04]
Lendermann et. al.[LPM01] focused on interfacing the simulation model and optimiza-
tion tool which has an objective of cost minimization. However like other recently dis-
cussed studies the information flow occurred in one direction from optimization to sim-
ulation. Glover et.al. [GKL99] defined another approach for the combination of simula-
tion and optimization models. They created two directions of information flow in be-
tween two models (like [TA03]). The sketch of this information flow is shown on Fig-
ure 7.
Figure 7 Coordination between simulation and optimization [GKL99]
The objective in the interfacing phase of this study is to create a platform and method-
ology to share the information which is one’s output and becomes other’s input. The
software used for the optimization part is called “OpenSolver” shown on Figure 8 which
contains macros written in MS Excel environment. These macros enable to eliminate
23
Literature Review
the limitation of MS Excel Solver platform on the numbers of variable and constraints.
The simulation software Automod has a feature to read and to write the data from the
external sources like MS Excel or “txt” file. With the help of this characteristic the
communication bridge between simulation and optimization can be easily built. There-
fore, it is not needed to have another tool to perform the data flow.
Figure 8 OpenSolver optimization engine

24
Casthouses in Primary Aluminium Smelter
3 Casthouses in Primary Aluminium Smelter
In this chapter, aluminium production is introduced and general information about the
steps from bauxite to aluminium and their global production is explained. Additionally,
production in primary smelters and their sub facilities are described. Finally a deep
analysis is done on primary aluminium casthouse supply chain by considering material
and information flow, activities and their interaction, casting process and product range.
3.1 Introduction to Aluminium Production

In aluminium life cycle (shown on Figure 3 in Chapter 1) there are two important pro-
cessing steps which are called primary aluminium and secondary aluminium production.
Primary aluminium production is the reduction of aluminium oxide to aluminium. Sec-
ondary aluminium production is re-melting of scrap aluminium recycled after usage.
After the development of Hall-Héroult process [GK93] which is an electrochemical

process used to produce primary aluminium from alumina, production of aluminium
increased continuously. Figure 9 shows the global production of primary aluminium
until 2010. The effect of financial crisis in 2008 on production and the growth of prima-
ry aluminium industry in China can be observed. The percentage of primary aluminium
production in Europe compared to other continents decreased in the last four years after
the financial crisis.
25
Figure 9 Global production of primary aluminium until 2010 [EAA12]
Aluminium cannot be found on the earth as an element. The ore bauxite is processed to
have aluminium oxide (also named as alumina) which is the raw material of primary
aluminium production. The Bayer process, cycle of which consists of processes such as
digestion, clarification, precipitation, evaporation and calcination, is used to purify
bauxite to aluminium oxide [GW88]. Figure 10 shows the intermediate material forms
in the steps of aluminium production from bauxite. For production of approximately one
kg aluminium, two kg alumina is needed. For two kg of alumina four kg bauxite has to
be used.
Figure 10 Steps of aluminium production from bauxite to aluminium [Hoe04]

26
Bauxite contains not only aluminum (Al2O3) which has around 50 % of the mass but
also ferric oxide (Fe2O3), titanium dioxide (TiO2), silicon dioxide (SiO2) and other im-
purities. Figure 11 shows the global production of bauxite in 2000 in “%”split to coun-
tries. Figure 12 shows the global production of aluminium oxide by the Bayer process in
2003 (in unit Mt) split to continents.
Figure 11 Global production of bauxite in 2000 (%) [Fri04]
Figure 12 Global production of aluminium oxide in 2003 (Mt) [Hoe04]

27
Energy consumption expenditure in primary aluminium production cost is around 25 %

which forces aluminium smelters to be located in the fields serving cheaper energy. The
other important criterion in selection of smelter location is the closeness to customers
and aluminum oxide suppliers. These criteria explain the inevitable growth of alumini-
um production in China. Figure 13 shows the location of aluminium smelters in the
world.
Figure 13 Location of aluminium smelters [IAI03]
The increase in global aluminium demand and production occurs due to its material
properties (such as light weight, cast ability, machinability, surface finish etc.) com-
pared to other metals. Table 3 shows the comparative evaluation of commonly used
metals in industry. Although aluminium has higher cost compared to the others, its
properties such as light weight, high resistance to corrosion, easy to form and recycla-
bility increase its use percentage by the industry. The main industries which have high
demand to aluminium are automotive, construction, electronics, packaging etc.
28
Aluminium Steel Grey cast iron Brass

Availability Good Good Good Good
Castability Less difficult Difficult Good Good
Machinability Very good Good Good Very good
Surface finish Very good Good Good Very good
Lending to modification Good Good Good Very good
Weight Very light Very heavy Very heavy Heavy
Brittleness Low Low High Low
Tendency to oxidation No Yes Yes No
Requiring machining Not much Less Yes Not much
Cost Medium Low Low High
Table 3 A comparative evaluation of commonly used metals [Jai03]
3.2 Primary Aluminium Production

The production of primary aluminium is based on Hall-Héroult process which was in-
vented in 1886. The basic of this process is reduction of alumina in a molten salt bath.
The process takes place in a cell which is shown on Figure 14. The cell (also called pot)
is formed by a cathode block made of carbon, carbon anodes and electrolytic bath where
alumina is dissolved. Electric current flows from anode to cathode heats up the bath to a
temperature around 950 °C, well above the melting point of the aluminium and electro-
lyte.
Figure 14 Schematic drawing of an alumina reduction cell [GK93]

29
The electrolysis unit in an aluminium smelter facility contains many cells where the
aluminium is produced. A modern smelter having production capacity around 500000
“tons/year” contains approximately 700 cells connected in series.
Smelter may have two other main units except electrolysis. The first one is the carbon
plant where the anodes are produced from coke and pitch. And the second one is cast-
house where the liquid aluminium is casted as an end product to be delivered to custom-
er. There are some other auxiliary plants such as fume treatment plant, bath treatment
plant etc. Figure 15 shows the material flow and the boundaries of the sub-facilities in
an aluminium smelter.
Figure 15 Material flow and sub facility boundaries in primary aluminium supply chain [HKK07]
The material flow from carbon plant to electrolysis contains the fresh anodes, and in
reverse direction transportation of used anodes. As raw material coke and pitch enter to
the carbon plant supply chain. Carbon plant contains paste plant where the coke and
pitch form the green anode, baking furnace where the anodes are baked and rodding
shop where the anodes are rodded and ready to be used in electrolysis.
30
The electrolysis is at the center of a smelter in the production point of view and from the
material flow perspective. It has direct contact to the carbon plant and casthouse units.
Fresh anodes and alumina are the main incoming materials and hot liquid metal is the
outgoing material to casthouse unit. Alumina is transported into electrolysis continuous-
ly but on the other hand, anode and hot metal transportation form the main part of the
logistical activities in electrolysis.
3.3 Primary Aluminium Casthouse Supply Chain

The casthouse is the last unit in the process flow of an aluminium smelter and has a di-
rect interface with electrolysis by receiving hot metal from this unit. However, there is
not any material flow between carbon plant and casthouse. Figure 16 shows the internal
supply chain of a primary aluminum casthouse. The boundary of this supply chain starts
with delivery of hot and cold metal and ends with shipment of end products to custom-
er.
Figure 16 Internal supply chain of primary aluminum casthouse

31
The main resources in a casthouse are the casting units including the casting furnaces.
These furnaces are filled with hot and cold metal to melt and mix them by considering
the quality of the metal. It supplies this metal to casting line to cast the end product as
requested by customer.
Hot metal filled in the casting furnaces may have two sources, one is the electrolysis
unit where aluminium is produced from alumina and the other is the remelting furnaces
where the recycled scrap material is melted. In most cases the remelting furnaces belong
to the casthouse layout. In this case, high amount of scrap material supply for re-melters
is included in the casthouse supply chain.
Cold metal as an incoming material contains aluminium ingots like bundle, T-bar, sow
etc. Alloys like silicon, mangan, magnesium etc. follow the same route in the supply
chain like cold metal as an incoming material. In addition to cold metal and alloys,
scrap materials can also be added to inbound materials of casthouse supply chain. Scrap
materials are required not only by re-melters but also by casting furnaces. They can be
added to the batch of casting furnaces depending on the customer quality specifications
and the melting capacity of the furnace. The amount of scrap melted in casting furnaces
are very low compared to the amount melted in re-melters.
The layout of furnaces and casting lines has big influence on the logistical activities in
casthouse. It has direct impact on the concepts like storage strategy, traffic, safety, park-
ing location of vehicles etc. Berlioux et. al. [BBB11] focused on optimizing layout of
aluminium casthouse by arranging the configuration of casting furnaces per casting line.
The casthouse production and the furnace batch planning also depend on the layout con-
figuration. Batch start time and distribution of orders to the furnaces are the main im-
pacts of layout on planning activities.
3.3.1 Material and Information Flow in the Casthouse Supply Chain

A casthouse unit differs from other smelter sub-facilities due to its direct contact to cus-
tomer and also its relations with internal and external metal suppliers. In addition to
that, the hot metal flow from electrolysis forces the production in the casthouse to fol-
32
low the planned schedule. The challenge is that it is impossible to buffer hot metal
without significant heat loss. These circumstances conduce to map information flow,
material flow and their correlation in detail. Figure 17 shows the material and infor-
mation flow in casthouse internal supply chain.
The information flow in Figure 17 may differ from casthouse to casthouse depending on
the characteristics of the casthouse supply chain. One of the levels from production plan
to batch plan may disappear or steps of data flow may differ. However, the backbone of
the flow including time frame, approach and transferred data do not vary so much.
Figure 17 Material and information flow in casthouse internal supply chain
The arrows with dashed lines show the information flow and solid lines show the mate-
rial flow between the units of casthouse internal supply chain in Figure 17. The time
frame of the figure starts with three months before production and ends with the next
three days after production. The production plan has two input sources. The first source
is the demand forecast according to analysis of historical data and the current situation
of the market. The second one is the customer orders that have already been received.
33
The orders generated according to demand forecast are replaced with the received cus-
tomer orders.
The production plan contains information such as customer identification number,

product type, product specifications (such as dimension, alloying etc.), production
amount, assigned casting furnace, assigned casting line, packaging mode, SKU (stock
keeping unit) specifications, delivery location and latest shipping date. The production
plan also supplies data for external suppliers by preparing supply orders according to
the customer order specifications. This planning operation may be updated up to two
days before the production when the production schedule is prepared.
External suppliers hand in delivery plans which provide information about the delivery
of incoming material to the production schedule. The delivery plan contains data such
as supplier definition, material type, material specifications, delivery amount, SKU of
delivery, date of delivery and transportation type etc. The material flow from the exter-
nal suppliers takes place latest one day earlier than the production date. The reason be-
hind this limitation is that some extra processes (such as drying, quality check etc.) may
be applied to incoming material before filling into a furnace.
The last data is received from tapping schedule to create the production schedule of the
casthouse. The operational management of the casthouse and the electrolysis units has
impact on the tapping schedule. Larrivée [Lar09] listed the factors that have to be taken
into account during preparation of tapping schedule as:
 Amount of hot metal required by casthouse
 Operation cycle of electrolysis
 Quantity of hot metal that can be obtained from one pot
 Chemical analysis of the hot metal
 Tapping process and operation characteristics (such as, tapped pot per transpor-
tation, direction of tapping in pot line etc.)
34
 Transportation and resource specifications (such as capacity of crucible, distance

between electrolysis and casthouse etc.)
The production schedule organizes the operations in the casthouse. It is valid for the
whole casthouse and contains the information about start time of each process and oper-
ation, their durations, assigned casting furnace and line of each batch, amount of re-
quired material and its location, next steps that product follows after production etc. The
production schedule has to be updated periodically (like shift wise) to highlight the cur-
rent situation and possible modifications on the tapping schedule and process organiza-
tion in the casthouse.
The production schedules distribute the information to batch (charging) plans. Each
furnace has its own batch plan which contains the data for the production in the next
one or two days according to the casthouse organization. The batch plan also matches
the batches with customer identification obtained from production plans. The batch plan
controls the production in casting furnace by providing information about the batch start
time, duration of production, batch composition, required hot metal quality, end product
properties and cold metal specifications (such as storage location, amount, allocated
transportation unit etc.). The information flow in the reverse direction occurs after pro-
duction to verify the customer order before shipment.
The material flow contains cold and hot metal as inbound materials. Cold metal has
intermediate storages between supplier and casting furnace (final destination). On the
other hand, hot metal is filled into a furnace directly after it is tapped from the electroly-
sis. The equipment used to transport hot metal from electrolysis to casthouse is called
crucible. It is designed to keep heat loss at minimum during the transportation.
The molten metal in the casting furnace is transferred into the casting unit. The next
production steps after casting process differ according to the product type. In most cas-
es, approximately two or three days are required to ship the material from the casthouse
to customer after it is stored in the end product storage. The duration of preparation for
shipment depends on the shipment type, international or domestic customer and re-
quired paperwork etc.
35
3.3.2 Activities and Their Interaction in the Casthouse Supply Chain

The process flow in casting furnaces drives the activities in the casthouse supply chain.
A casthouse operation management system triggers the material flow, allocates the re-
sources, controls the traffic and organizes the processes in its supply chain according to
batch plan of casting furnaces. Figure 18 shows the activities and their interactions in
the casthouse internal supply chain. As it can also be seen on the figure, the casting fur-
nace stays in the center of the process flow.
The process flow of casting furnace may vary depending on:
 the furnace specifications (such as inaptitude of furnace to stirring process)
 end product type and properties (such as pure ingot does not require any alloying
process)
 operational strategy (such as sampling or one of the holding processes may dis-
appear in the sequence).
36
Figure 18 Activities and their interactions in casthouse supply chain
The process flow in casting furnaces may start with a preparation phase depending on
the relation between the last and the next batches. If the alloy type differs and requires
any wash-up process, cleaning will be done in this step by removing all of the metal
from furnace and adding some alloys. Even though there is not any production, furnaces
cannot be turned off. During this idle period, a furnace still contains a little amount of
metal which is called sump.
The cold metal addition process normally takes place after the preparation phase, but in
some cases when the temperature is higher than planned, cold metal is added to the fur-
nace in the next steps of process flow. It cools down the furnace and regulates the tem-
37
perature. In this concept, cold metal represents pure aluminium ingots supplied from
external sources. Depending on the furnace melting rate and customer order specifica-
tions, a little amount of scrap can also be added to the furnace.
Cold metal and scrap follow the same process flow before being filled into furnace.
They are unloaded from external transportation unit (trucks or boats) by forklifts or
shovel depending on their transportation type. A quality check is done to identify their
chemical contents. They have to be dried to remove water particles before being melted
in a furnace. While furnace melts the added cold metal, the stirring process helps to mix
the bath of metal to decrease the dissolution duration [BC02]. A special vehicle and
equipment are used to stir the furnace.
After melting the cold metal, hot metal can be added to furnace. The source of the hot
metal can be the electrolysis unit or re-melting furnaces depending on the casthouse
system. Primary hot metal flow is triggered by tapping schedule. It organizes tapping
time, identification of pots to be tapped and allocation of tapping vehicles. After the
tapping operation, the fluxing operation takes place depending on the smelter specifica-
tion. The reason of the fluxing process is to purify the metal.
The routing of tapping vehicles between electrolysis, fluxing station and casting furnace
differs from smelter to smelter. The whole transportation can be done by the same tap-
ping vehicle or it can be split in two loops at the fluxing station. Tapping vehicles do the
tapping operation and transport the crucible to the fluxing station and another crucible
carrying truck handles the fluxed crucible and transports it to casthouse. This operation-
al strategy depends on the vehicle fleet and specifications of hot metal flow manage-
ment in the smelters.
Another possible source for the hot metal is a re-melting furnace which supplies sec-
ondary hot metal to casting furnaces. The process flow in a re-melting furnace has simi-
larities to the process flow in casting furnaces. However, it does not contain hot metal
filling, alloying and casting operations.
The holding process in the flow represents the heating process to keep the metal at de-
sired temperature for the casting process. During the holding process alloying metals
38
can be added to the bath according to the customer specifications. Kaufman et. al.
[KR04] listed seven basic families for the alloys in aluminium casthouses. They are;
aluminium-copper, aluminium-silicon-copper, aluminium-silicon, aluminium-silicon-
magnesium, aluminium-magnesium, aluminium-zinc-magnesium and aluminium-tin.
Sampling operation in the process flow, checks the chemical composition of metal if it
satisfies the requirements or not. If it is not in the defined specification, other alloying
processes will take place until the customer requirements are reached.
The skimming process is used to remove the dross over the surface of the metal in the
furnace. Dross is the residual material occurred during melting process [CWD09]. It is
not a waste product, after a recycling process it can be used as a scrap material in sec-
ondary aluminium production. For the skimming process, a skimming beam is used to
remove dross over the surface. In the industry special furnace tending vehicles are de-
signed to simplify the skimming process. After skimming process, the gathered dross
product is kept in cooling boxes. Transportation of these boxes is done by forklifts in
the casthouse. After cooling down, this residual material is shipped to other facilities to
be processed for recycling.
The metal in the furnace is held until the casting unit is available for the casting process.
Normally a casting unit is linked to more than one casting furnace. However, the mix-
ture of the metal from different furnaces is not possible due to their chemical composi-
tion. After emptying the furnace to the casting unit, it will be ready for the next batch
cycle.
3.3.3 Casting Process and Possible Product Range

European aluminium association [EAA12] separates the casting techniques used in al-
uminium industry into two groups according to the final product. Primary and second-
ary aluminium casthouses which produce semi-finished products use the ingot casting
technique and the other foundries cast aluminium with mould casting technique. How-
ever some casthouses in primary aluminium smelters also use the mould technique to
cast semi-finished products such as T-bars. Literature contains detailed studies (Some of
39
them are [BB99], [KU07], [TM03], [Bee01]) about casting process and different casting
techniques.
The quality check of the final goods is done after the casting process. Failed products
are sent back to casthouse production area to be used as cold metal for the next batches.
Schmitz [Sch07] called these products “new scrap” which are received from own pro-
duction of the casthouse. In this study, this material is called production scrap to prevent
confusion with external scrap which has high contamination.
After casting and quality check, some further processes (such as homogenization, stack-
ing etc.) take place. These processes are often fully automated in today’s aluminium
industry. So after packaging, the material is transported to the storage area to be shipped
to customer. According to the transportation type some extra handling are done in the
end product storage operations (like container filling).
End products of primary aluminium casthouses are semi-finished products which are
used as raw material in foundries, in rolling mills or in extrusion facility. The specifica-
tions such as alloy content, dimension, level of impurities, surface quality, cast ability
and conductivity form the possible product range of primary casthouse. Figure 19 shows
some end product types produced by Hydro Aluminium. In addition to the types shown
on the figure, T-bars, standard ingot and sows are also produced in the smelters as an
end product.
40
Figure 19 End products of primary aluminium casthouse [Kva11]
The usage fields of end products differ according to their types. For example, wire rods
due to its conductivity are used in cable production area. Extrusion ingots (also called
billets) are the raw material for the extrusion facilities which have automotive, construc-
tion and transportation industries as a client. On the other hand, sheet ingots are used in
rolling industry to serve the customers from the fields packaging, construction and
transportation. Foundry alloys are re-casted in foundries to be used in manufacturing of
wheel rims, suspension parts and engine cradles in automotive industry.
41
Simulation Model Development for Aluminium Casthouse
4 Simulation Model Development for

Aluminium Casthouse
The focus of the simulation part of the thesis is to map the whole material flow in alu-
minium casthouse, and the aim of the model is to create a universal platform which is
applicable to any casthouse. As mentioned, Automod is used as the simulation tool for
this study. In this chapter, after giving brief information about the literature review of
simulation usage in aluminium industry, the model concept and logistical control ap-
proaches behind the model will be described in detail. The strategies behind the control
of the model increase flexibility of the model implementation.
4.1 Introduction and Literature Review

Logistics simulation studies performed in metal industry have become challenging in
the recent years. The reason is that attitude toward simulation models are changing. Ac-
cording to Guo [Guo03], in the past expectations were “quick and dirty” simulation
models with “Know-how” approach, but today it is expected to model the system as
accurate as possible to find the answer to the question “Why”. Additional to that, due to
safety reasons in the heavy metal industry, detailed analysis of the human behaviors is
also intended to be in the main focus of logistical studies done for the metal industry.
Several simulation studies have been done to analyze the material and information flow
in the aluminium production. The focuses of simulation studies were to reduce the fixed
costs and eliminate the risks from the safety perspective. As mentioned, in aluminium
industry the electrolysis part of the smelter has the highest priority due to high energy
cost. Casthouse arouses the interest in logistical analysis due to the hot metal transport.
Eick et al. [EVB01] and Meijer [Mei10] focused mainly on material flow of the elec-
trolysis and made some investigations only for the pot room part of the smelter. Harton
[Har10] simulated the hot metal flow between electrolysis and casthouse. Baxter et al.
42
[BBTM09] built a simulation model focusing on the anode production plant area which
has not a correlation with the hot metal transport. Tikasz et al. [TBPM10] and Pires et
al. [PBTM11] investigated the full smelter logistics with the objective to improve the
system from the safety perspective. Jaouen [Jao11] made a simulation study for the
downstream part of aluminium casthouse. This simulation study differs from others by
aiming to map the whole material flow taking place in the casthouse of an aluminium
smelter.
4.2 Boundary of the Simulation Model

The system boundary for the casthouse simulation model has to take into account the
close interaction with the electrolysis part of the smelter. Electrolysis has an ongoing
process cycle of continuous aluminium production. The liquid aluminium is tapped
batch-wise and the tapped aluminium is transported to the casthouse in crucibles. There-
fore, in the casthouse supply chain analysis, the hot metal transported from the electrol-
ysis should be considered in detail to get accurate results.
The tapping cycle may shift with short delays when a disturbance occurs but long de-
lays cannot be tolerated due to the system restriction (such as hot metal characteristics).
However the transported load in between the pot room and casthouse cannot be buffered
longer and also cannot be reserved as scrap. Winkelmann et al. [WEDS09] defined a
full smelter simulation approach to map the system in more accurate way and also to
eliminate the risks ( e.g. inaccuracy in hot metal flow) occurring due to some simplifica-
tions in the system interactions.
Having a flexible model, which can be applied to different casthouses, brings the chal-
lenge to consider many possible control approaches that can be used in the material
flow field. For that reason, various casthouses were studied to have wide eligible logis-
tical perspectives in the model. The model boundary of the simulation study contains
not only the primary casting furnaces but also re-melting furnaces. Extrusion billets,
foundry alloys, sheet ingot slabs and wiring rods are considered as the possible end
products of the casthouse.
43
4.3 Concept of the Simulation Model

Structure of the casthouse simulation model contains four layers which are shown on
Figure 20. These layers are classified according to their attributes and file format. Ar-
rows show the direction of information flow occurs in between these layers in the simu-
lation model structure. The first layer is input data definition which is done in MS Ex-
cel, the second and the third parts, control logic and model elements, are defined in Au-
tomod simulation software. The last part is determined in Automod and converted to
“txt” format.
Figure 20 Casthouse simulation model structure

44
Input Data
This part of the model concept contains planning data and the characteristics and com-
ponents of the system which are defined in MS Excel. The data explored in this section
in most cases is received from the other departments of the smelter like planning and
operational departments. All of the data defined in MS Excel environment inserted to
the simulation environment with the help of functions written in Automod and macros
defined in MS Excel. Information flow occurs from this layer to second and third layers
of the model structure. Items defined in this part of the concept are;
 Batch plan; is defined for each furnace in the system. It contains the information
like; identification and the recipe of batches, final product specifications, aver-
age durations (filling, melting, holding) etc.
 Delivery plan; is prepared for the incoming material. Information defined in de-
livery plan is described in the subchapter 4.5.4.
 Shipping plan; is prepared for the outgoing materials. It contains the information
like; customer identification, product specifications, date of delivery, packing
type, transportation type etc.
 Hot metal distribution plan; contains the information about electrolysis part of
the smelter, hot metal transportation specifications, average driving duration in
between casthouse and electrolysis etc.
 Furnace characteristics; contains the information about furnaces like; furnace

design specifications (identification, capacity, melting rates), responsible casting
line, selected strategies for metal filling etc.
 Vehicle definitions; contains the design data of each vehicle in the system
boundary, vehicle fleet, possible task allocation groups etc.
45
Control Logic
Control logic behind the simulation model is programmed in Automod simulation soft-
ware with the help of its functions. The tool enables to build the logic, not only with its
predefined functions but also user defined functions written in C or C++ programming
language. Control logic segment controls the all other parts of the simulation model.
The parts integrated into control logic concept are discussed in detail in the subchapter
4.5. In the development phase of the casthouse simulation model, some parts of the
software programming in Automod were done in cooperation with IdeCraft, Norway.
Model Elements
All of the items enabling to visualize the system are integrated in this layer. Automod
defined objects and their attributes are used to visualize the system. Some object orient-
ed simulation tools (such as Enterprise Dynamics, Simul8) attach its own control logic
to each element in the model. In Automod, only the characteristics of item are linked to
it, all of the control logic has to be defined separately. This has disadvantages such as
requirement for a skilled programmer. On the other hand, defining the logic separately
has many advantages like user defined strategies and needs, advanced logic control and
flexible models to be modified easily. Items in the model element segment are;
 Process layout; contains the queues, resources and loads defined to visualize the
real system. All of these Automod defined objects are placed in the model to
identify furnaces, storage locations, materials etc.
 Movement system ; contains the paths, vehicles and vehicle segments defined in
the model
 Static background; Automod enables to import some drawings at the background

of the model to improve the visualization. This may be a drawing or bird view
picture to describe the layout of the system in a better way.
46
Output Data
Result gained from the simulation model depends on the objective and the demand from
the user, but there are some standard statistics that are valid in the casthouse model for
example bottleneck analysis, resource utilization, traffic density safety and storage
analysis. In this thesis, the required output data from the simulation model is defined to
interface with the optimization model in detail in Chapter 6.
4.4 Main Components of the Simulation Model

The simulation model contains the components classified into sub-models with respect
to their properties. These sub-models are primary hot metal flow, secondary metal flow,
cold metal flow, furnace operations and material flow after the casting process. Each of
these components can be examined independently. If only the partial material flow
analysis is required then only the needed part of the model can be used for further inves-
tigations.
4.4.1 Primary Hot Metal Distribution

Hot metal flow from electrolysis constitutes the crucial part of the casthouse material
flow. Hot metal flow without detailed electrolysis concept makes the casthouse supply
chain so complex to design in the simulation environment. For that reason, the logic of
the flow trigger system in the model should have been designed in detail to allow map-
ping of various strategies. The solution applied at the simulation model is to create a
distribution of the crucible numbers according to time [Des110]. The distribution may
vary depending on the possible strategies applied in the real environment.
The starting point is to calculate the amount of metal produced in the electrolysis on the
shift based system. The equation (4.1) shows how the average amount of metal pro-
duced per shift is calculated in the model logic.
47

k
potline1
(NPpotline * TMTCpotline) * DD
AvgMPS 
SD * TC
(4.1)
where;
AvgMPS = Average metal produced per shift (ton)
NP = Number of pots for each pot line
TMTC= Total metal tapped from one pot per tapping cycle of the pot line (ton)
DD= Day duration (24 hours)
SD = Number of the shifts per day
TC = Tapping cycle duration of the pot line (hr)
k= Total number of pot lines in the system
Average metal produced per shift helps to calculate the average number of crucibles
transported per shift from the electrolysis to the casthouse. The equation (4.2) shows the
basic calculation of the average number of crucible transported per shift.
AvgMPS
AvgNCS 
AvgTMC
(4.2)
where;
AvgNCS = Average number of crucibles transported per shift
AvgMPS = Average metal produced per shift (ton)
AvgTMC = Average tapped metal per crucible per transport (ton)
After calculating the crucible number transported per shift, the tapping operation cycle
duration is needed to distribute the crucibles in time perspective. To calculate the aver-
48
age tapping operation and transportation duration, four main constraints should be taken
into account. First one is the transportation duration, the second is the tapping operation
duration, the third is the time spent at the fluxing station and the last one is the crucible
tilting process duration. The equation (4.3) shows the correlation of these constraints to
find the average tapping operation and transportation cycle duration.
AvgTCD  TDP * NPTC  AvgDRT  AvgFOD  AvgTPD
(4.3)
where;
AvgTCD = Average tapping operation and transportation cycle duration (min)
TDP = Tapping duration per pot (min)
NPTC = Number of pots per crucible
AvgDRT= Average driving duration of round trip outside of the casthouse (min)
AvgFOD= Average fluxing operation duration (min)
AvgTPD= Average tilting operation duration (min)
The hot metal from electrolysis transported in a shift is defined as hour based distribu-
tion. The maximum number of crucible transport per hour is determined. However it is
not allowed to park the crucible transporting vehicle at the electrolysis and casthouse
area due to the safety reasons. Therefore, at most of the smelters, tapping vehicles are
parking at the special parking locations or maintenance area. Due to that reason, during
the first hour of the shift the crucible transport may not reach the maximum number. In
the model logic, the crucible distribution in the first hour of the shift is calculated by
equation (4.4), while the maximum crucible transportation number is determined from
equation (4.5).
49
NTV * (60  AvgTCD)

FNCT 
AvgTCD
(4.4)
where;
FNCT= Number of crucible transport for the first hour of the shift
NTV= Number of tapping vehicles operating in the system
NTV * HD
MaxNCT 
AvgTCD
(4.5)
where;
MaxNCT= Maximum number of crucible transport that can be performed in one

hour
NTV= Number of tapping vehicles operating in the system
HD= Hour duration (min)
After the calculation of the hot metal transportation in the first hour of the shift and the
next hours, it has to be determined the amount left for the last transportation hour. Equa-
tion (4.6) is used to calculate the number of crucible transport for the last hour of the
planned hot metal transportation frame for one shift.
LNCT  AvgNCS  (FNCT  MaxNCT * (PTS  2))
(4.6)
50
where;
LNCT= Number of crucible transport for the last hour of the planned hot metal
transportation frame for one shift
AvgNCS = Average number of crucibles transported per shift
FNCT= Number of crucible transport for the first hour of the shift
MaxNCT= Maximum number of crucible transport that can be performed in one

hour
PTS= Planned hot metal transportation frame for one shift
After calculating the first, last and maximum number of crucible transport, the trigger
system can use the distribution approach which is shown in the Figure 21.
Figure 21 Distribution of the crucible transport in one shift
4.4.2 Secondary Hot Metal Transport

Secondary hot metal transport has to be taken into account for the casthouse supply
chain when there is a re-melting furnace in the system. Normally re-melting furnace or
furnaces are located inside the casthouse. Therefore, material flow to and from the re-
melting furnace has to be considered in the interim casthouse supply chain concept.
51
A re-melting furnace is used to melt external scrap, internal pure scrap and cold metal.
The operations and material flow of the re-melting furnace is similar to the casting fur-
nace except the primary hot metal filling and the final process. In the simulation model
concept, furnace management of the re-melting furnace and casting furnace are mapped
in the same way. For the final process, if the re-melting furnace has the batch-wise pro-
duction then a “Push” concept will be applied to transport the metal from the re-melting
furnace to the casting furnace. Alternatively, if the system contains a continuous re-
melting furnace then the transportation logic control will turn to a “Pull” system.
In both cases, due to the uncertainties in the system, small buffers should be considered.
To place the buffers between re-melting furnace and casting furnace mostly the empty
vehicles or crucible filling area are occupied. Casthouses with re-melting furnaces need
an optimum planning concept to distribute not only the primary metal from electrolysis
but also the secondary hot metal from the re-melting furnaces.
4.4.3 Scrap and Cold Metal Transport

Hot metal transport is the core part of the casthouse external supply chain but also the
internal material flow scrap and cold metal transport requires detailed consideration. It
is not possible to charge the cold metal directly coming from the supplier. Some pro-
cesses like sampling, drying and warming should be applied beforehand. For each pro-
cess, new storage location and also reshuffling possibilities in between the storages have
to be taken into consideration. In addition, surplus, pure scrap from the own production,
is also added to the cold metal inventory. From the logistical perspective, detailed in-
ventory and storage management should be determined for the cold metal material flow.
In the simulation model, a classification approach [Des210] is implemented to identify

the materials. Firstly, the material category, which is the upper level of the classifica-
tion, contains the piles such as ingot, scrap, pure metal, alloy etc. In the second level of
the classification, the material type is identified according to the transportation type or
properties of material such as top bottom pieces, T-bar, bale scrap etc. Figure 22 shows
one example for the category and type classification of additive metals. In the last part
of the classification has the quality level which is linked to the supplier of the metal.
52
Figure 22 Category and type classification example
In the model, this classification approach is also linked to the storage location. The pre-
defined storage locations contain the information about the classification data coming
from the material specifications. Therefore, it is enough to define the storage location in
the batch plan if the specifications are known for the simulation period. To define the
material type, it is sufficient to identify the location of the required material.
4.4.4 Furnace Operations

In the simulation model, all of the decisions taken in the internal casthouse supply chain
are based on casting furnaces and their operations. Processes handled in a casting fur-
nace are logically lined up and mapped as duration in the system. The uncertainties in
this duration mapping influence the decisions taken by the system. Therefore, some
processes may be repeated due to delays or the lack of material at current state.
Figure 23 shows the flow diagram of processes in the production cycle of a usual cast-
ing furnace. The simulation model follows this sequence of processes which can be
modified according to the system characteristics. Re-melting furnace also follows the
same process order except the hot metal filling process. The process durations of the re-
melting furnace are based on the furnace specifications such as casting furnaces, and the
final process of the re-melting furnace is also named as casting in the simulation model
despite the process is not the usual casting operation.
53
Figure 23 Process cycle of casting furnace in the simulation model
Preparation
Due to alloy usage, in some cases wash-up process is needed in between two batches. In
the model concept, the predefined correlation in between alloy types determines the
length of the process duration.
Cold Metal Addition
Before hot metal filling due to the cost factors and need for cooling the furnace, cold
metal is filled into the furnace. This process contains also pure scrap metal additions.
The duration of this process depends on handling and the transportation duration of the
metal from storage into the furnace.
Pre-Heating
After cold metal addition, the duration used to melt the metal is named as “Pre-Heating”
process. The duration of this process depends on furnace specifications. Melting rate for
cold metal may differ from the melting rate for scrap. Therefore, the specifications
should be well determined at the beginning of simulation.
Stirring
After cold metal addition, the stirring process increases the efficiency of the system by
mixing the bath which decreases the dissolution duration. Special equipment is used to
stir the metal, so in the model, equipment handling and travelling time are considered in
the stirring process duration.
54
Hot Metal Filling
This process starts with the claim for the first crucible and ends with the tilting of the
last crucible into the furnace. In this process not only the primary hot metal, if the sys-
tem contains, secondary hot metal flow is also integrated. The process depends on the
handling strategy of the crucible from vehicle to the furnace. Handling by crane, siphon-
ing by special equipment and tapping system on the vehicle are possible alternatives
that are considered in the model logic. Handling procedure has big influence on the pro-
cess duration. Additionally, equipment availability and resource idleness cause the sys-
tem being behind the schedule.
Holding, Alloying
The duration from hot metal filling until the casting process is named as “Holding” in
the simulation model. Alloying process is also integrated as duration in the model logic.
Aim of the holding process is to keep the temperature constant in the furnace. The
planned holding process duration is calculated statically at the beginning of the simula-
tion based on the furnace specifications, and it is kept constant during the simulation
run.
Skimming, Sampling
Skimming process contains the dross handling process and its transportation. Skimming
has an influence on the process duration due to special tool usage to skim the furnace.
The vehicle transportation and tool handling are mapped in detail in the model logic.
Also the dross box availability, handling and transportation may cause some delays in
the system. Sampling process occurs just after skimming and has constant duration in
the model. Only uncertainty may occur due to the skimming process. In the real life,
sampling process may cause some process repetitions (e.g. alloying), but in the model
effects of the chemical composition are ignored.
55
4.4.5 Down Stream Processes after Casting

The downstream part of casthouse supply chain contains the processes after casting and
also the handling operations until shipment of orders. Casthouse specific processes, for
example cooling, homogenization etc., and also general packing processes such as
strapping, cutting, stacking etc. are considered in the simulation model logic.
All of these processes are mapped as duration in the model. Possible activities and in-
teractions of the processes are not taken into consideration. In other words, the average
processing time is applied to the system from the first product of the batch entering the
sub-model until the last product of the batch leaving the system. The process lead time
planned for each process is kept constant during the simulation, so in this part of the
system disturbances may only occur due to the resource availability.
4.5 Logistical Control Strategies in the Simulation Model

The core point of logistics simulation models is the analysis of reality with the different
logistical control approaches. In some cases two or more of these approaches can be
mixed and simulated system may have its own control logic. Therefore, selecting and
implementing correct logistical strategy to the simulation model should be well struc-
tured at the beginning of the study. In this sub-chapter, some of the logistical control
strategies considered in the simulation model are described.
4.5.1 Possible Scenarios for Primary Crucible Allocation

Under subchapter 4.4.1 the primary hot metal trigger system, its distribution and the
logic behind it were discussed in detail. As mentioned in casthouse supply chain hot
metal flow plays an important role due to its characteristics. For the production schedul-
ing, crucible-carrying primary hot metal- and casting furnace allocation becomes very
crucial. This allocation strategy may differ from casthouse to casthouse. Therefore, the
logistical approach behind this allocation should be mapped with alternative scenarios
in the model to match with the reality.
56
In some smelter organizations, the tapping schedule of the electrolysis manages the hot
metal flow. In the others, allocated pots and the tapping schedule are decided by the
casthouse operation management. As a third alternative, there is a hot metal coordinator
in the system and tries to balance hot metal flow between pot room production and cast-
house requirement. In this simulation study, four possible alternatives were modeled.
Push System
In this alternative, the logic is based on the push strategy from electrolysis to the cast-
house. Electrolysis unit decides the hot metal flow scheduling and casthouse reacts to
this schedule. If uncertainties (delays, resource unavailability etc.) occur and hot metal
needs to be buffered, this will take place in front of the furnace. System in the model
approaches to the reality by arranging the production plan according to the data coming
from electrolysis unit. With this concept the production schedule of the casthouse
should be rearranged periodically to respond to uncertainties.
Pull System
In this alternative, the logic is based on the casthouse hot metal filling schedule. This
schedule may contain a pot selection depending on quality of the metal, time of opera-
tion and the amount of tapped metal. The crucible buffer area is either the fluxing sta-
tion or the gate in front of the casthouse. Safety reasons prohibit buffering the crucibles
in the electrolysis area.
Combination of Push and Pull Systems
This concept can be used when the real system has a metal manager on the top who or-
ganizes the hot metal schedule. In this concept, vehicles buffer at the push-pull bounda-
ry [SKS03] of the system and are allocated to the casting furnaces according to a FIFO
rule. In a crucible furnace matching, only arriving times of the vehicles play a role.
Each crucible can be allocated to each furnace without any restriction.
57
Combination of Push and Pull Systems with Limitations
The logic of this alternative has one advanced step compared to “Combination of Push
and Pull Systems”. The aim of this strategy is to create a link in between the pots and
furnaces, based on metal quality specifications. If the preferred furnaces do not require
hot metal in predefined waiting duration then the crucible is allocated to the other avail-
able furnaces. This advanced logic is mostly selected if the casthouse has different
product types or if the system boundary contains more than one casthouse.
4.5.2 Traffic Management in the Simulation Model

The simulation tool has its own traffic management approach based on the predefined
interactions between the paths. It is possible to change the navigation factor which is
defined for the whole path at the beginning of the simulation. According to these fac-
tors, the system calculates the possible shortest path during the simulation. These fac-
tors are defined at the beginning of the simulation so it is not possible to change their
values during the simulation. It is required to control the course of the vehicles in a so-
phisticated way. A navigation tool was developed to solve this problem. And also the
restrictions (like layout perspective) coming from the simulated system specifications
need to be applied in a flexible way in the model, for that reason the zone management
approach is determined.
Vehicle Navigation
It is possible to route the vehicles according to strict traffic rules by programming so-
phisticated logic, which may restrict the flexibility of the model. For that, all of the pos-
sible alternatives should be written in the logic and user should design all decision pos-
sibilities and selections at the beginning. When it is needed during the simulation, the
system selects the correct path according to this predefined logic.
The required navigation concept in simulation models needs more sophisticated logic
than the vehicle control systems provided by the simulation tools. First and probably the
most important requirement is the dynamic approach ([EKP96], [BMND05]) which
cannot be managed by the current vehicle control system of Automod simulation tool.
58
Automod has its own navigation system which also allows defining static weighing fac-
tors through the paths, but it is not possible to change these factors during the running
simulation. It is a big challenge for simulation tools to change these factors and make a
decision according to the new factors at an instant time which introduces dynamics to
the model.
The second demand is related to the flexibility of the system configuration. Path defini-
tion in Automod has the option where the navigation factor can also be attached as an
attribute, but this factor is valid for the whole path. It is not possible to differentiate the
navigation factor for all path connections. The challenge for this issue is to have differ-
ent weighing factors between the decision points along the path. In principal, it can be
solved by defining separate paths between each control points but then the model be-
comes slower due to the high number of defined identities.
The third lack of Automod is that it is not able to differentiate the weighing factors ac-
cording to the different vehicle types. This feature is needed because each vehicle type
has its own restrictions. For example, in some cases one vehicle type cannot have access
to one part of the layout. It is not possible to model it with the vehicle control system of
Automod. Logic has to be written to define that kind of features in the model. After rec-
ognizing the requirements the approach defined in Figure 24 was developed to accom-
plish the lacks of the navigation system of Automod.
Figure 24 The new navigation tool implementation to guide vehicles

59
The logic of the navigation tool is based on the modified Dijkstra’s Algorithm ([Dij59],
[GT96]). The interface between Automod and Dijkstra acts as a decision maker for path
selection according to the output from Dijkstra and the restrictions coming from simula-
tion model. A modified Dijkstra’s Algorithm with neighborhood approach ([ISW09],
[Pri04]) for the vehicle routing problem ([SWW00], [SF96], [Mey03]) finds the opti-
mized node connection by passing the layers in between source and sink points. Figure
25 shows possible two layers connection problem for the navigation tool. The abbrevia-
tions and their explanations on Figure 25 are:
i= Nodes from 1 to n
Li = Decision locations along the path
WLi-L(i+1) = Weighing factor from Li to L(i+1)
Figure 25 Possible two layers node connection
Aim of the system is to reach from L1 to Ln and the steps followed by the navigation tool
are;
1. Check the first possible connection according to the neighborhood sequence ap-
proach;
LiLi+1 i
60
2. Calculate total weight of the possibilities by;
Wtotal  i 1 ( j1 (Wij ))

n n
 i,  j (if Wij <= 0 then break)
(4.7)
3. Check all of the possible alternative connections in between;
L1Ln if Wtotal > “The breaking parameter” then break
4. Select the option having smallest Wtotal as the preferred path.
Example to show the logic behind navigation tool:
According to the example shown in Figure 26, the task is defined to travel from point A
to point E on the layout by using the navigation tool. If the points are read by the system
according to the alphabetical sequence, the navigation check will start from A to B. Af-
ter calculating all the possibilities with this starting arrangement, the model will start to
check other possible alternatives from A to C, then from A to D and finally from A to E.
At the end all the possibilities from A to E are determined and the most optimized path
according to step 5 will be selected.
Figure 26 An example to show the selection logic
The sequence of the calculation is started with A to B, but the defined weighing factor is
greater than “50” (the default breaking parameter) so the connection is broken, and the
system skips this initial arrangement and passes from A to C. With this new arrange-
ment, the system can reach the destination through the connection ACDE according to
61
the neighborhood approach. The total weighing factor is reset from “50” (as predefined)
to “12”. The tool continues to find the other possible alternatives with the breaking pa-
rameter “12” and in the next alternative, it reaches the destination with the arrangement
of ACDBE and the total weighing factors are stored as “4”. It continues with the other
possibilities but after that alternative (ACDBE), the breaking parameter of the calcula-
tion is set to “4”. It is obviously seen at the last check that the system terminates the
calculation without reaching to the destination because the ADB arrangement has “5” as
a weighing factor after reaching to the point B. By the help of this terminating logic, the
speed of calculation is tremendously increased without changing the result.
At the end of this example, the system decides ACDBE as most optimized one and
routes the vehicle through this path. For the next possible demand from A to E, the sys-
tem will again recalculate all these alternatives. Pre-calculated connections are not
stored in the memory, because the connections between the paths are changing during
the simulation dynamically. For example, if one obstacle is located between B to E and
increase the weighing factor, then system can choose the ACE connection or any other
one according to the modified factors for the next travels.
As mentioned, the logic behind the navigation tool is based on quantitative optimization
between possible alternatives found according to a predefined matrix. The system at the
beginning stores all connections and their weights between the decision points. And also
during defining period, neighborhood priority criteria between the connections is creat-
ed. The decision path is shown in Figure 27.
62
Figure 27 Information flow in the decision logic behind the navigation tool
63
Zone Management Concept
In the model for vehicle management, the navigation tool is used to control the traffic
inside and outside of the casthouse. In addition, a zone management approach was de-
veloped and implemented to the vehicle routing control of the model due to safety rules
in the casthouse. The zone management approach has three layers of control. The logic
behind is similar to the work zone approach used in highway’s roadway management
[KA03] which is applied when there is a reconstruction, maintenance or any factor dis-
turbing the traffic on a highway. Mirchandani et al. [MSL03] and Yadlapati et al.
[YP04] performed simulation studies to analyze the decision alternatives taken after
coming across with a work-zone. In the casthouse model, the decision is taken by the
navigation tool when there is a blocked zone along the path. The difference from the
normal work zone management is a multi-stage control of the zones depending on the
layout and system restrictions.
Figure 28 shows the zone layers in front of the casting furnace. The first zone controls
the resource availability. When there is a vehicle based process at the furnace at an in-
stant, this zone prevents the furnace to be claimed by another vehicle. The upcoming
vehicle process is queued until availability of the furnace.
Figure 28 Layers of the zone management approach in front of the casting furnace
64
The second zone controls the traffic just in front of the furnace by obeying the safety
regulations. When there is a vehicle standing in front of the furnace for any reason, this
zone blocks the traffic on the lane in front of the furnace. Depending on the system
characteristics, vehicles may queue at the zone boundary or find another alternative way
to the destination.
The third zone is responsible for traffic of the aisle in front of the furnace. In some cast-
houses, in front of the furnace special traffic rules (such as, overtaking an operating ve-
hicle is forbidden) are executed due to the layout restrictions. This stage is applicable
when there is a two lane way in front of the furnace which can be used by two vehicles
if none of them is in operation. With this zone level the whole access to the furnace area
can be blocked when needed.
4.5.3 Furnace Management

Furnace management is the main decision maker in the process triggering part of the
simulation model. The process cycle (Figure 23) defined in the logic sequences the pro-
cesses and arranges their start time. In the simulation model, the furnace cycle starts
with the warm-up period which helps to create a production schedule for the casthouse
operations. After the warm-up period in the simulation the first process “preparation”
begins to set the batch properties when necessary.
The second process “cold metal addition” depends on the transportation logic in the
model. When the target amount of cold metal is loaded into the furnace then the burners
turn on to melt the metal. Duration of the pre-heating process, excluded the duration of
stirring process, is calculated as shown in equation (4.8). In the middle of the pre-
heating, the stirring process is activated if it is included in the furnace operation cycle.
The stirring process duration is not included in the pre-heating duration because it is
simulated as a discrete event.
65
TScA  TCMA
MD 
AvgMR
(4.8)
where;
MD= Melting duration (min)
TScA= Total amount of scrap metal addition (ton)
TCMA= Total amount of cold metal addition (ton)
AvgMR= Average melting rate of the furnace (ton/min)
After reaching the end of the melting duration, the hot metal filling process is triggered.
The hot metal filling process is ended with the last crucible transport that accomplishes
the required hot metal amount. With this event the holding process is triggered in the
furnace cycle. The duration of the holding process, which also contains the alloying
process duration, is pre-defined in the system depending on the operation and furnace
characteristics. Skimming and sampling are excluded from the holding operation. They
are mapped as discrete event processes.
The start time of the casting process which has a constant duration depends on two cri-
teria, one is the termination of holding process and the other is the availability of the
casting unit. If the casting unit is not ready at the end of the holding process, then the
system continues with the holding process by changing the state of the furnace to
„Waiting for Casting“. When both of these criteria are satisfied then the casting process
is triggered. At the end of the casting process if the start time of the next batch is not
reached then the furnace turns its state to “Idle”. Figure 29 shows the flow of control
logic behind the furnace management concept after the furnace process “holding”.
66
Figure 29 Control logic of the furnace management after the holding process
4.5.4 Storage Management Concepts

The strategies behind storage management used by the facilities depend on criteria such
as material flow characteristics, material type, transportation unit, and storage properties
etc. Logistical simulation studies were performed to analyze some of these strategies in
different industry fields ([DD07], [BPS08], [LJI02], [SER06]). A practical usage of the
storage strategies in aluminium casthouse supply chain can be seen on stocking of
scrap, cold metal and final goods storage concept.
The handling operation for the cold metal may be duplicated due to the drying process
and storage space restrictions. To control this replenishment process in between the in-
termediate storages makes the model complicated. Another important issue for the stor-
age management concept drives from the “door buffering” strategy applied in some
casthouses. This “door buffering”, which is defined in subchapter 4.5.5, enables to fill
the cold metal to the furnace in one step to minimize the hatch opening duration. The
other benefit of door buffering is that it is applicable for special cases like automatic
pusher concept.
67
The replenishment of intermediate storages is triggered by the material sources having

defined capacity, which brings the opportunity to push the material from the source to
the intermediate storages or another source if the amount of storage reaches its limit.
This concept increases the flexibility of material turn over between the storages in the
casthouse. Figure 30 shows the strategy behind the storage management concept.
Figure 30 Strategy behind the storage management concept
The delivery plan, which contains the information like supplier identification, material
category, material type, SKU specifications, amount, delivery date, delivery type etc.,
triggers the incoming material flow. All of the data, specifying the material, is carried
with the material load until the furnace filling. Therefore, system enables the traceabil-
ity of incoming materials to support at the decision gates in the concept.
Incoming Material Stocking Strategies
In the simulation model, several strategies are prepared for the material unloading pro-
cess [Des210], because this process differs from plant to plant depending on the trans-
portation type, plant location, safety regulations, supplier contract etc.
Two main steps require more attention during the incoming material stocking. One is
the unloading from the transportation unit and the other one is the bay selection for the
material in the internal storage.
Two concepts are considered for the material unloading process:

68
1. Always unload at first to a temporary area to enable quality check and create a
decision point for bay selection at the next step.
2. Direct unloading to the internal storage, at which the quality check and bay se-
lection are done before the unloading process. In this concept incoming material
stays on the transportation unit during the acceptance process which causes long
waiting time for the transportation unit.
For the bay selection decision four alternatives are prepared:
1. According to the type of material, to the first available storage based on the in-
dex (sequence of definition) is selected.
2. According to the type of the material, to an available storage selection of which

based on the defined selection criteria; such as supplier identification, quality
etc.
3. To the predefined bay if there are strict storage rules in the system
4. Always to an empty storage, the first available storage from the storage defini-
tion sequence is selected.
Storage Control Alternatives
Three storage control alternatives are established according to the characteristics and
complexity of the system. These alternatives can be seen as the complexity stages of the
control mechanism. Each alternative has its own principles concerning the stocking,
filling and emptying practices [Des110].
• Default: If no details are specified, this concept generates simplified material

flow. The only stocking criterion is the amount of the storage. This concept pulls
material when the amount of the storage reaches its predefined minimum level.
The source of the system has infinite capacity and a delivery plan is not needed
for this concept. Basic FIFO principle is applied to remove material from the
storage.
69
• Intermediate: Enables to control the system better than the default concept by
identifying details about storage management and the system layout. The stock-
ing criterion depends on the material category. The filling strategy is still a pull
concept which does not require detailed delivery plan. For the emptying princi-
ple, a FIFO concept, paying attention to the storage replenishment application, is
applied.
• Advanced: Compared to the intermediate concept the material type is in focus

for the stocking. A detailed delivery plan manages the filling process. With this
principle, the storage areas have to be mapped in detail on the system layout to
be able to define the boundaries. If there is not enough storage places compared
to the material type than system obeys the material category rules (defined in
subchapter 4.4.3.
All of these concepts are prepared to analyze the casthouse supply chain in the simula-
tion environment as close as possible to the reality. Additionally these alternative ap-
proaches enable the user to implement the simulation model to different casthouses.
Flexibility and complexity of the model are increased by providing alternative control
strategies in the forehand.
4.5.5 Scrap & Cold Metal Filling Strategies

Strategies applied for the scrap and cold metal filling into the furnace depends not only
on the buffering concept but also on the transportation type. In most cases the character-
istics of the casting furnace decide for the filling strategy. From the logistical perspec-
tive the material flow in between the main material storage and the casting furnace de-
pends on the material SKU type, distance in between source and sink, availability of the
transportation resources and the filling specifications of the casting unit. Figure 31
shows the common approaches for the furnace filling strategy. The strategy enables to
generate simple transportation in between storage and the unit without “door buffering”.
Although the door-buffering concept doubles the transportation, it eases the resource
allocation. Another advantage of the door buffering concept is the heat loss reduction
70
during the filling operation. The door buffering concept reduces the iterative filling pro-
cess time which forces each time opening the lid of the furnace.
Figure 31 Furnace filling strategy (left) without and (right) with door buffering
Specified by the furnace and material characteristics, in some cases similar to the door
buffering concept, all of the required material are firstly filled into a box. The transpor-
tation equipment can be a front loader with shovel or a simple forklift. Depending on
the system, the box is emptied into the furnace by front loaders or automatic pushers.
This concept is common for the re-melting furnaces or the casting furnaces having
burners with higher heating power. Compared to the other concepts, the filling process
into the furnace is handled in one step. Figure 32 describes the transportation process in
between storage and furnace.
Figure 32 Furnace filling strategy with a box

71
Lean Thinking Approach in the Primary Aluminium Casthouse Supply Chain
5 Lean Thinking Approach in the Primary

Aluminium Casthouse Supply Chain
After preparing the simulation model for the primary aluminium casthouse supply
chain, in this step of the research the results of the model are analyzed. Possible wastes
from logistical point of view are defined and reasons behind these non-value added ac-
tivities are specified. The waste categorization in the lean thinking approach is the main
driver of the analysis. Additionally, possible improvements are highlighted to reduce the
non-value added costs in the supply chain.

The aim of the lean thinking approach is to increase the efficiency in an application
field. It was developed for the automotive industry which has batch-wise production.
The implementation of lean thinking comprises of three steps. At first, the material and
information flow are mapped to analyze the system which was done in the previous
chapter. Then according to the results of the simulation model, wastes are identified and
finally potential improvements are defined to eliminate these wastes. The improvements
are discussed in the next chapter where the aim of the optimization models is defined.
Wastes are identified in the supply chain by differentiating between value added activi-
ties and non-value added activities in the flow. Wastes or non-value added activities in
this context contain the processes which do not create any value in the production cycle
of the product. There are seven groups of wastes categorized in the lean thinking ap-
proach [Ohn88]. These groups are: Overproduction, waiting time, unnecessary transpor-
tation, poor processing, inventory, extra movements and defective products.
According to Abdullah [Abd03], the application of the lean thinking approach to con-
tinuous processes is not that common compared to discrete manufacturing. It was creat-
ed for the automotive industry which has discrete assembly operations instead continu-
72
ous processes. Therefore, the number of publications concerning lean thinking in metal
industry is restricted (e.g. steel production [Abd03], steel mill facility [AR07] and
chemical process industry [Mel05]). The casthouse supply chain can be defined as a
mixed system which combines continuous and batch-wise production. The continuous
process in electrolysis has an impact on the batch-wise production of casting furnaces.
In the primary aluminium production, cost reduction was studied in different fields of
the smelters. However, the focus of these studies was mostly on process improvement
of the production. A few of the studies in the literature have the aim to improve effi-
ciency of the system by focusing on the non-value added activities. Meier [Mei11] did a
study for an Electrolysis area by focusing logistical activity flow.
Studies done in the casthouse area were mostly focused on a single concept in the sup-
ply chain. For example, Peterson et. al. [PN02] studied to minimize gross melt loss dur-
ing skimming process. Jensson et. al. [JKG05] and Yuan et. al. [YKSBT04] examined
the casthouse production in minimizing the setup times. Maiwald et. al. [ML06] focused
on finding optimum temperature regime in casting furnace which has impact on produc-
tivity and efficiency. Gravel et. al. [GPG02] followed a closer approach to analyze the
scheduling problem by aiming to reduce hot metal waiting time, tardiness and early
production. However, they did the optimization for each waste separately without com-
bining them.
In this thesis, the lean thinking approach is applied to all possible non-value added costs
occurred due to logistical activities.
5.2 Waste Analysis in the Casthouse from a Logistical Point

of View
In this subchapter, the results of the simulation model are analyzed from a logistical
point of view and the wastes are identified. The analysis contains the non-value added
costs depending on logistical activities such as traffic, transportation, buffering, stock-
ing, production planning and scheduling etc. Additionally, it is also described how these
items are quantified and measured in the simulation model.
73
5.2.1 Overproduction in the Casthouse Supply Chain

Overproduction means in lean thinking producing more than demanded. In a primary
aluminium casthouse, there is a considerable surplus production. There are several rea-
sons for the surplus. The first one is the campaign production principle. In this strategy,
uninterrupted production of the same material type is done in the same casting furnaces
and unit. For example, this principle is preferred in casthouses to prevent the wash-up
process due to different alloy usage.
The second reason of surplus production is the continuous hot metal supply from the
electrolysis unit which cannot be stored without casting. Therefore, it is not possible to
stop the casting in casthouse even if there is not any customer order for a batch.
The solution for such inconsistencies is found by matching the surplus with a customer
order after the production. Therefore, surplus material is stored in the casthouse for a
while to be demanded by customers. The difference between customer orders is based
on specifications of product such as dimension of the material and its alloy type. There-
fore, the surplus material is produced with the most common alloy types and in common
dimensions.
There is another reason of overproduction which also causes to produce surplus materi-
al. After splitting customer orders to batches, the amount of the last remaining part of
the customer order can be lower than the production capacity of the casting furnace. The
energy consumption during the furnace holding process does not vary so much with the
amount of material in the furnace. In other words, the energy used to heat up the furnace
and to hold it with a constant temperature does not depend on the amount of material in
the furnace. Therefore, the casthouse unit follows the strategy to produce a batch the
amount of which is closer to the capacity of the furnace. This strategy may cause to
produce more metal than the customer order. A solution for this problem is to split the
customer orders in an optimized way to the casting furnaces to get less amount of sur-
plus material. However, this cannot be a hundred percent solution of the problem, it
only reduces the overproduction.
74
As mentioned, surplus materials are stored until a customer order matches the specifica-
tions. If the surplus does not satisfy any customer requirement within a certain time
frame then it will be re-melted in a casting furnace as cold metal.
In the simulation model, the surplus materials are transported to surplus storages ac-
cording to the data flow from the batch plan. If the customer identification column in
the batch plan is empty or specified as “surplus”, than the type of material is also con-
sidered in the simulation model as “surplus”. The material amount in the surplus storage
is updated periodically and shows the current status of the storage.
5.2.2 Waiting Time in the Casthouse Supply Chain

Waiting time of resources such as transportation units, casting furnaces and casting
units creates not only direct non-value added costs but also inconsistency in the produc-
tion planning and scheduling. Logistical activities in the casthouse have direct impact
on waiting time. Possible reasons of waiting time in casthouse supply chain are dis-
cussed after describing how these waiting durations are measured in the model.
The casting furnaces are the production units in the casthouse, so the analysis on wait-
ing duration has these units on focus. Two reasons of waiting time are specified in the
simulation model after analyzing the status of casting furnaces. The first reason is the
relation between processes such as hot metal flow and furnace charging. And the second
one is the resource unavailability such as waiting for the casting unit or waiting for a
special tool (e.g. skimming, stirring, packaging processes).
Hot metal waiting time is caused by delays in the tapping schedule which can also cre-
ate disturbances in the production schedule of casthouse. On the other way around, poor
scheduling in casthouse may cause also tardiness in hot metal transportation. This wait-
ing duration of casting furnaces is specified in the simulation model by considering the
hot metal filling process duration and the state of casting furnace resource in “Hot Metal
Filling”.
Resource unavailability can also cause the casting furnace to wait for processing. Two
examples in the simulation model are observed for this case casting furnaces wait either
75
for a skimming machine or for the assigned casting unit. The waiting duration of a
skimming machine can be ignored because the process duration of skimming is not that
long. However, the waiting duration for a casting unit cause considerable amount of
delay in the batch production. This waiting duration increases the batch lead time con-
siderably.
Simulation time(min) Furnace ID Batch ID Process Definition Process Duration (min)

630.0 Furnace 1 10001 0_Preparation 0.0
630.0 Furnace 1 10001 1_Cold Metal Addition 21.0
651.0 Furnace 1 10001 2_Pre-heating 52.0
703.0 Furnace 1 10001 3_Hot Metal Filling 42.0
745.0 Furnace 1 10001 4_Holding 20.0
765.0 Furnace 1 10001 5_Alloying 17.0
782.0 Furnace 1 10001 4_Holding 50.0
832.0 Furnace 1 10001 Waiting for Casting 18.0
850.0 Furnace 1 10001 6_Casting 120.0
970.0 Furnace 1 7_Idle 50.0
1020.0 Furnace 1 10002 1_Cold Metal Addition 28.0
Table 4 An example of a furnace status tracing in the simulation model
The batch lead time of casting furnaces is measured in the simulation model by tracing
the start time and end time of batches. Start time is captured when the casting furnace is
triggered for the first process of the batch. End time is captured when the last metal is
transported to the casting unit and the status of the furnace turns to “idle”. In the exam-
ple presented on Table 4 the start time of the batch was recorded as “630” minutes after
simulation was started and end time of the batch is captured as “970” so the batch lead
time is calculated as 340 min for this example.
The waiting durations in the simulation model is also measured by highlighting the re-
source utilization graphically and also by preparing a data table showing the activity of
this resource per time. For each resource, a set of states is attached to specify its current
activity. Vehicles and production units have their own state definition. In other words,
the possible states of the resources are categorized for each resource type.
The figure below (Figure 33) shows an example of resource utilization output from the
model of a production facility. This example highlights the percentage of resource states
for the casting furnace in a casthouse. The states such as casting, hot metal filling, alloy-
76
ing, skimming, stirring, holding, sampling, waiting for casting, cold metal loading, and
scrap loading and burning are some of the possible states of casting furnace. The dura-
tion and percentage comparison help to analyze the furnace efficiency and also unex-
pected high proportion due to delays.
Figure 33 Example utilization graph of a casting furnace in percentage
Vehicle utilization can also be gathered from the simulation model. The states of the
vehicles in the system are defined as waiting, retrieving, parking and delivering. “Wait-
ing” represents the duration without any movement occurred while performing a task. It
contains for example, possible disturbances due to the traffic or possible delays due to
processes or possible bottlenecks due to the resource usage. The second state named as
“Retrieving” represents the empty travel duration spent on the way to the next sched-
uled task. “Parking” shows the time spent at the parking area. The last state “Deliver-
ing” represents the travel duration while carrying a load. Figure 34 shows one example
of forklift performing tasks such as furnace filling, storage reshuffling etc.
77
time
Figure 34 An example utilization graph of a transportation unit in percentage
5.2.3 Unnecessary Transportation in the Casthouse Supply Chain

Waste in transportation contains the relocation of materials which does not add value to
the analyzed system. Other waste groups of the lean thinking approach (such as waste in
processing, waste in inventory etc.) have direct impact on the concept “waste in trans-
portation”. However, storage reshuffling brings extra cost which is not included into
any group of waste.
The storage reshuffling depends on the capacity of storages and inventory strategy of a
casthouse. One reason of reshuffling is that the amount of the assigned product extends
the stock capacity. Then the materials are transported to another stock area. Another
reason is that some storage areas are assigned for specific processes such as drying due
to their location (for example closeness to the furnace). Therefore, incoming materials
transported from outside of the casthouse are firstly stored in this area then distributed
to another storage place.
It is possible to quantify the transportation cost of storage reshuffling in the simulation

model. And also alternative strategies can be selected in the model to map advanced
storage concept. However in this waste, the root of the problem is the operational strat-
egy of the casthouse. The possible improvement forces to change the operational strate-
gy in the casthouse.
78
5.2.4 Inefficient Processing in the Casthouse Supply Chain

In this part of the lean thinking concept, the logistical activities such as production
planning and process scheduling can be included. These activity coordination processes
may cause waste in the supply chain such as longer idle time in furnaces or late delivery
of order. For example, poor production planning in combination with unexpected delays
in the casthouse may cause to ship the material later than the planned delivery date. This
waste is measured in the simulation model in a way that after leaving the last production
station, the delivery date is checked for the material. If the delivery date has already
passed then the late delivery penalty is added to the non-value added costs depending on
the delay and the penalty rate.
Another waste in processing can be the idle time of casting furnaces between two
batches due to unavailability of assigned casting line. This waste can occur when two or
more casting furnaces are assigned to the same casting unit. In this case, scheduling has
to be done by considering the availability of casting unit. If the schedule is not carefully
prepared then the casting furnaces wait longer than planned. The idle time of the fur-
nace is calculated by capturing the trigger time of the last processes of the previous
batch and the first process of the next batch. The duration in between is recorded as an
idle time. For the example presented on Table 4, the idle time is measured as 50 min for
the batch “10001”.
In this research, the focus is only on the discrete logistical events occurring in a cast-
house. Therefore, metallurgical processes are not in the scope of the study. Processes in
the casting furnace are mapped as duration in the simulation model. If a process is not
affected by the traffic in the casthouse or any resource availability then this duration
does not vary during the simulation run.
5.2.5 Inventory of the Casthouse Supply Chain

A poor inventory management in a supply chain requires extra space, extra handling
and extra costs [ITC04]. Therefore, the waste due to inventory management is deter-
mined in the simulation model by two different methods. The first one is highlighting
79
the storage status by measuring the amount in the storage and the total retention time of
the products spent in the storage periodically. The second one is focusing on a specific
product to analyze resulting costs for inventory, handling etc. in the casthouse.
There are three groups of materials that are stored in casthouse. They are:
 Incoming materials containing cold metal, external scrap and alloys
 Internal scrap and surplus materials which are supplied from the production
 Finished goods which are ready for the shipment to customer
Incoming materials enter the supply chain according to a delivery plan provided by a
supplier. In the simulation model, the external truck transport is triggered by these de-
livery plans, which are defined statically at the beginning of the simulation run. Unload-
ing the truck has an effect on the activities in the supply chain due to the resource allo-
cation and the traffic. However, mostly individual resources like forklifts are dedicated
for external truck unloading. In this study, a static delivery plan and a particular re-
source allocation makes it difficult to find any improvement in the incoming material
inventory with the help of the simulation model.
The second group of materials is supplied directly from a casting unit. They are trans-
ported to their assigned storage and according to the batch plan allocated to a furnace.
Scrap materials are created in the simulation model according to the historical quality
deviation of the casthouse which is statically implemented. The inventory cost for the
surplus products is measured in the simulation model by highlighting the storage
amount and measuring the utilization of the allocated transportation unit. These costs
can be reduced by lowering the amount of surplus material.
The last group contains the finished products for shipment to customer. According to
the shipping type specified in the customer order, they are stored in the casthouse. Be-
fore shipment, in most cases some days are needed to prepare the order for the shipment
such as paperwork. However if the product is produced earlier than this time then it cre-
80
ates extra inventory cost for the casthouse. It is possible to measure this inventory and
also the transportation cost in the simulation model. This non-value added cost can be
eliminated or reduced by optimizing the production plan of a casthouse.
5.2.6 Unnecessary Movement in the Casthouse Supply Chain

Waste of movements contains the motions of resources, tools or people which do not
add value to the analyzed system. The furnace charging strategy and the routing strategy
of tapping vehicles in a casthouse may cause extra movements of resources which cre-
ate not only extra handling cost but also possibility to damage the equipment or product
during movement. Strategy in furnace filling such as charging box usage adds also extra
tool handling.
The simulation model concept is developed to map alternative strategies for the furnace
charging. However in this case, the root of the problem is the operational strategy of a
casthouse. Although it is possible to quantify the transportation cost for the box usage,
potential defect cost cannot be measured in the simulation model. And also any im-
provement requires changing the strategy of furnace filling concepts.
The routing of tapping vehicles can create non-value added costs which can happen due
to one-way rule for tapping vehicles in casthouses in terms of safety. It eases the traffic
and helps to prevent tapping vehicles encounter with any other vehicle which may cre-
ate a danger. It is possible to measure the travelled distance and also the utilization of
vehicles in the simulation model. Although finding another routing path for tapping
vehicles may improve this situation and reduce the non-value added costs, it is not al-
lowed in casthouses to change one-way rule if it is requested by safety regulations
5.2.7 Defective Products in the Casthouse Supply Chain

In general, a quality check is done after the casting process directly in the casthouse.
Surface finish, cracks and dimensions are the main characteristics of the products that
are controlled during this process step. The rejected products are called production scrap
which has to be differentiated from the external scrap material. It has high purity and
81
can be melted as cold metal in further batches. The internal scrap is transported to the
scrap storage and then according to the batch allocation reshuffled to a cold metal stor-
age or directly to the furnace.
In the simulation model, the quality deviation of production scrap is parameterized so it

is just defined as a percentage in the logic. However, the percentage does not change
during the simulation so it is not required to measure the internal scrap amount in the
simulation model. Transportation of the production scrap from the casting line to the
storage and to the furnace is mapped as a discrete event in the simulation model.
5.3 Summary of Waste Analysis

Each group of waste in the casthouse supply chain was analyzed and some of them were
specified for the improvement. The process of elimination or reduction of wastes con-
tains optimization models which focus on these non-vale added costs. The categoriza-
tion of wastes depending on improvement potential within the scope of the optimization
model is listed on Table 5. Some of them are kept out of the scope of the optimization
model. The reasons of the exclusion are also defined in the same table, and the extended
explanation for these reasons can be found in the previous subchapter.
In Scope Out of Scope Reason

Waste of overproduction Surplus production
Hot metal waiting time
Waste of waiting time
Longer batch lead time
Waste in transportation Storage reshuffling Problem of the operational strategy
Longer idle time in furnaces
Waste in processing
Late delivery of an order
Inventory of incoming materials Improvement approach not suitable
Waste in inventory Inventory of internal scrap and surplus Integrated into surplus cost
Inventory of finished goods
Furnace charging strategy Problem of the operational strategy
Waste of movement
One-way rule for tapping vehicles Safety aspects
Waste of defective products Rejected products in quality check Improvement approach not suitable
Table 5 Categorization of wastes depending on improvement possibility in the scope of the optimi-
zation model
82
Optimization Model Development
6 Optimization Model Development
After analyzing the non-value added logistical costs in aluminium supply chain (with
the help of the simulation platform and the lean thinking approach), the next step, de-
velopment of the optimization part, is discussed in detail in this chapter. For the optimi-
zation of aluminium casthouse supply chain, two models are developed. The first one is
the short-term production planner and the second one is the production scheduler. These
models are described in detail with their objective functions, constraints and variables.
In addition to those, interfacing platform between simulation and optimization parts is
explained in the last section of the chapter.

The optimization part of this study contains two different models which aim to mini-
mize the non-value added costs due to logistical activities. The aim of the first tool is to
create a production plan which is capable of distributing the customer orders to the
batches for casting furnaces and also allocate these batches to the appropriate furnaces.
Tang et. al. [TLRY01] created a production plan in the same direction for the steel cast-
ing plant. Their objective was to increase the productivity and energy saving.
Nonas et. al. [NO05] also focused on production planning of foundry and their objective
was to find an efficient plan which minimizes late delivery. Pacciarelli et. al. [PP04]
created a production plan with the same objective for steel casting plant. Tan et.al.
[TK05] studied rearranging customer orders with the help of computerized method for
the casting unit. They succeeded in reducing scrap metal 20% with the new approach.
The aim of the second tool is to schedule the production in aluminium casthouse by
arranging the operations at the casting furnaces. This part of the study has a close inter-
action with the electrolysis part of the smelter due to the impact on the hot metal flow
83
management. According to Freeman et. al. [FKZM05] the basic of casthouse schedul-
ing can be identified as the combination of problems known as:
 Lot-sizing
 Sequencing
 Scheduling
This definition is based on the combination of continuous operations (e.g. aluminium

production in electrolysis) and batch–wise processes (e.g. pot tapping, crucible transport
and furnace filling). Therefore, batch-optimization formed the main focus of many stud-
ies such as: iron and steel enterprise [TW08], small foundry [AAC08] and aluminium
foundry [Rya98] and [Pie08].
Maticevic et.al. [MML08] dealt with scheduling issues in aluminium foundry by aiming
to minimize the tardiness in production. Gravel et. al. [GPG02] also studied the schedul-
ing problem in an aluminium casting center by aiming to reduce:
 Set-up time
 Hot metal waiting time
 Tardiness
 Early transportation penalty
Tikasz et. al. [TMPB12] highlighted the benefits of testing the production schedule of
aluminium casthouse in the simulation platform. However, after detailed literature sur-
vey, any published study about interfacing a production plan or schedule with simula-
tion platform for the aluminium casthouse could not be found. Therefore, the methodol-
ogy for the interfacing is kept in general which is defined by Matta [Mat08]. According
to this study, simulation model sends “system performance” to optimization model and
receives “system alternatives”.
84
6.2 Model Structure

As mentioned, the optimization part, containing two models, aims to minimize the non-
value added logistical costs in the internal aluminium casthouse supply chain. The first
model is named as “Short-term production planner” which is discussed in sub-chapter
6.3. The second model called “Production scheduler” is detailed in sub-chapter 6.4.
These two optimization models have the same structure. Both of them were modeled in
MS Excel environment and use the “OpenSolver” optimization engine to find the opti-
mum solution.
Figure 35 Optimization platform in MS Excel
Figure 35 shows the optimization platform prepared in MS Excel. The objective func-
tion, constraints and variable definitions are defined in excel solver platform without
paying attention to its limitations. Macros defined in “OpenSolver” are used to solve the
85
optimization models of the study. Integer linear programming method is used to pro-
gram the logic behind the optimization concept.
Input data is received from customer orders and system specifications, both of which
are prepared in MS Excel environment. The first tool distributes the customer orders to
the batches and then allocates them to the furnaces. The data provided from customer
orders contains the information such as:
 Customer identification number
 Sales document number
 Material type
 Dimension of material
 Alloy type
 Date of delivery
 Total delivery weight
 Priority of order
 Region of order
With this data, the optimization tool tries to find the best combination of batches appro-
priate casting furnaces. After solving the optimization problem, the data is converted to
the batch plans which are read by the simulation model.
The second tool schedules the operations for the casting furnaces in the casthouse. The
data gained from system specifications contains the information about:
 Duration of each process
 Process sequence for each furnace
 Furnace characteristics (e.g. capacity, melting rate etc.)

86
 Relation in between casting furnace and casting line
 Hot metal distribution and transportation strategy
 Needed time between two different alloys based production
With this data, the optimization tool seeks the best production schedule for the whole
system.
6.3 Short-term Production Planner

Primary aluminium casthouse which produces the final form of product is the last facili-
ty of primary aluminium supply chain. Therefore, it has direct contact to customers. The
costs due to this contact (e.g. late delivery) may be high and cannot be easily compen-
sated. For that reason the production has to be well planned and followed by the produc-
tion unit.
6.3.1 Aim of the Model

The aim of the short-term production planner model is to minimize the non-value added
costs occurring due to the production planning. The model handles the problem by fo-
cusing on the casting furnaces. The main parameters of the model are number of batches
per shift per furnace, capacity of furnace, allocated casting unit and production period.
Customer orders are distributed with respect to the possible number of production per
furnace in a day. The planning period can be extended by defining more parameters to
increase the time frame. In “OpenSolver” optimization engine, there is not any limita-
tion for the number of parameter so the planning period depends on the user.
Figure 36 shows a snapshot of the optimization model which is prepared as an example

for five casting furnaces (represented as F_1, F_2, F_3, F_4, and F_5) with two casting
lines (represented as CL_1 and CL_2). The example has one batch production per fur-
nace per shift. This figure shows the batch distribution in the matrix as placing “1” un-
87
der the assigned furnace number and across the sales document number which identifies
the customer order. This assignment complies with the constraints:
 Capacity of the casting furnace (batch capacity)
 The relation between casting furnace and casting line
 Casting line availability and utilization
 Capacity of assigned casting line
Day_1
Shift_1 Shift_2 Shift_3
Customer Sales
Identification Document CL_1 CL_2 CL_1 CL_2 CL_1 CL_2
Index Number Number F_1 F_2 F_3 F_4 F_5 F_1 F_2 F_3 F_4 F_5 F_1 F_2 F_3 F_4 F_5
1 106 106101 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1
2 117 117101 0 0 0 0 0 0 1 0 1 0 0 1 0 1 0
3 122 122101 0 0 0 1 0 0 0 0 0 0 1 0 1 0 0
4 105 105101 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0
5 113 113101 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0
6 125 125101 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
7 112 112101 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0
8 111 111101 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0
Figure 36 Snapshot of the first optimization tool
6.3.2 Objective Function

The objective of short-term planner tool is to minimize the costs due to;
 late delivery of the order
 early production causing the inventory cost at the storage of final goods
 surplus material because of more production than customer demand
The indices used in the optimization model are:
 furnace Furnaces in the casthouse
 order Customer orders

88
 batch Produced unit of a customer order
 timeslot Slots that shift duration is equally divided in
 shift Shifts in the optimization period
The parameters used in the optimization model are:
 m Total number of furnaces in the casthouse
 k Total number of customer orders
 n Total number of produced batches
 t Total number of slots
 s Total number of shifts
 FurAvfurnace shift Number of batches assigned to the furnace per shift
 SmFurCap Capacity of the smallest furnace in the system (ton)
 BtAmorder batch Amount of the batch of the order (ton)
 OAmorder Amount of the order (ton)
 OPriorder Penalty rate of the order depending on the customer
 LaProTorder Latest production time that is planned for this order (day)
 BAmbatch Amount of the batch (SKU)
 DelTorder Delivery date of the order (day)
 SalesPr Sales price of the product (Euro /ton)
 PenRate Penalty rate (% per day)
 ManHrC Man hour and equipment cost (Euro/ SKU / day)

89
 InvC Inventory cost (Euro/ SKU / day)
 CMC Market price of external cold metal (Euro /ton)
The decision variables of the optimization model:
 FurCSfurnace timeslot Current status of the furnace per timeslot
 SurpAmorder Amount of over production per order (ton)
 ODelorder Duration of the delay for the order (day)
 ProTorder batch Production time of the batch belonging to the related order
(day)
 BRetbatch Retention time of the batch in the end product storage (day)
The objective function of short-term planner tool:
Minimize COST_LD + COST_EPS + COST_SR
(6.1)
subject to;
FurCSfurnace timeslot  {0, 1}  furnace,  timeslot
(6.2)

t
timeslot1
( FurCSfurnace timeslot) = FurAvfurnace shift  furnace,  shift
(6.3)

n
SurpAmorder = batch 1
( BtAmorder batch) - OAmorder  order
(6.4)
90
SurpAmorder <= SmFurCap  order
(6.5)
SurpAmorder >= 0  order
(6.6)
ODelorder= max[0, (ProTorder batch – LaProTorder)]  order,  batch
(6.7)
BRetbatch = max[0, (DelTorder – ProTorder batch)]  order
(6.8)
where;
COST_LD Punishment cost of late delivery
COST_EPS Inventory cost for waiting at the end product storage
COST_SR Cost of surplus products
6.3.3 Equations of Costs in the Objective Function of the Short-term

Production Planner
Punishment cost of late delivery
This cost contains the extra payments that have to be paid to customer due to the late
delivery. The amount of penalty depends on duration of the delay, amount of the order
and priority of the order. The formula of “Cost_LD” can be seen in Equation (6.9).
Cost_LD = CoeffLDtoCost * order1 ( ODelorder * OAmorder * OPriorder)

k
(6.9)
where;
CoeffLDtoCost Coefficient of the equation for the cost conversion

91
The delay of the order in this frame can be analyzed in detail with respect to production
date and delivery date. The reason of this analysis is to create a connection between
order and batch which is produced later than planned. Equation (6.7) shows the formula
for determination of delay of the order in the optimization model.
End product storage cost:
The inventory cost of early produced batches which depends on waiting duration in the
end product storage forms the main part of “Cost_EPS”. Early produced batches require
not only storage space but also additional handling cost. This extra transportation cost is
also taken into account in the coefficient (shown in Equation (6.13)) of Equation (6.10).
Equation (6.8) shows the detailed analysis of retention time calculation of batch in the
end product storage.

n
Cost_EPS = CoeffEPStoCost * batch 1
( BRetbatch * BAmbatch)
(6.10)
where,
CoeffEPStoCost Coefficient of the equation for the cost conversion
Surplus cost:
As mentioned in Chapter 5, the amount of the batch, which depends on the casthouse
and system characteristics, in most cases is nearly equal to the effective capacity of a
furnace due to system efficiency. Therefore, the produced amount is sometimes more
than the amount of the customer order. If the overproduced amount is not requested by
customer in a defined time frame, the extra production will become production scrap.
The coefficient (shown in Equation (6.14)) of Equation (6.11) contains also the extra
storage and transportation cost in an average value during this waiting time.
92

k
Cost_SR = CoeffSRtoCost * order1
( SurpAmorder)
(6.11)
where;
CoeffSRtoCost Coefficient of the equation for the cost conversion
6.3.4 Equations of the Conversion Coefficients

Equation (6.9) calculates the penalty cost due to late delivery. Without the coefficient,
the formula determines the delay of order in unit “ton * day”, so the formula has to be
converted to cost unit “Euro”. For that reason the coefficient which is calculated by
Equation (6.12) is used.
CoeffLDtoCost = SalesPr * PenRate
(6.12)
Equation (6.10) is used to calculate the cost due to early production. Extra handling op-
eration and storage cost is determined in calculating the coefficient which is used to
convert the unit of equation from “day * SKU” to cost unit “Euro” with Equation (6.13).
CoeffEPStoCost = ManHrC + InvC
(6.13)
To calculate the coefficient used in Equation (6.11) the difference between sales cost
and external cold metal cost is determined with the formula (6.14). The reason behind is
the usage of surplus material as production scrap which can be substituted with required
cold metal in the batch plan.
CoeffSRtoCost = SalesPr - CMC
(6.14)
93
6.4 Production Scheduler

After creating the batches from customer orders and allocating them to the casting fur-
naces, the schedule of processes is coordinated by the tool “production scheduler”. This
model is capable of deciding the start time of the operations of casting furnaces by aim-
ing to reduce the logistical costs occurring due to resource availability, waiting time in
the queue and additional handling operations.
6.4.1 Aim of the Model

Like the first optimization model, this tool also focuses on the casting furnaces. The
sequence of operations, their durations and starting time are the main parameters of the
tool. The logic of the tool divides the time into intervals the length of which depends on
the required sensitivity. The shorter the interval means more sensitive the analysis. The
length of processes is also determined with respect to the length of the intervals. Figure
37 shows the snapshot of the example model which has a half an hour time interval. In
this example the length of the processes are also defined as multiples of 30 minutes that
helps to allocate the process to the interval.
1st Shift
Furnace\ Hour 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0
Fur_1 7 7 7 7 7 7 7 7 1 2 2 3 3 4 5 5
Fur_2 7 7 7 7 1 2 2 3 3 4 5 5 6 6 6 6
Fur_3 1 2 2 3 3 4 5 5 6 6 6 6 7 7 7 7
Fur_4 7 7 7 7 7 7 1 2 2 3 3 4 5 5 6 6
Fur_5 1 2 2 3 3 4 5 5 6 6 6 6 6 6 7 7
Figure 37 Snapshot of the second optimization tool
The processes of each furnace have a sequence, an index and duration definition accord-
ing to the system characteristics of the casthouse. Figure 38 shows another snapshot of
this optimization tool. In this snapshot, the process distribution to time for one furnace
can be seen. The time interval is also taken as 30 minutes for this example and each
process has duration as multiples of 30 minutes. Index of processes and their sequence
are seen on the left hand side of Figure 38. Green circles in the matrix show the time
and process intersection. When there is a green circle across the process then this means
94
that process takes place at a time at the specified furnace. The last process, “7” in this
example, always signifies the furnace idleness.
1st Shift
Furnace\ Hour 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0
Fur_1 1 2 2 3 3 4 5 5 6 6 6 6 7 7 7 7
Process Index Duration (hr) Sq.
Cold Metal Addition 1 0.5 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Pre-heating 2 0.5 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1
Pre-heating 2 0.5 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2
Hot Metal Filling 3 0.5 3 3 3 3 4 3 3 3 3 3 3 3 3 3 3 3 3
Hot Metal Filling 3 0.5 4 4 4 4 4 5 4 4 4 4 4 4 4 4 4 4 4
Alloying 4 0.5 5 5 5 5 5 5 6 5 5 5 5 5 5 5 5 5 5
Holding 5 0.5 6 6 6 6 6 6 6 7 6 6 6 6 6 6 6 6 6
Holding 5 0.5 7 7 7 7 7 7 7 7 8 7 7 7 7 7 7 7 7
Casting 6 0.5 8 8 8 8 8 8 8 8 8 9 8 8 8 8 8 8 8
Casting 6 0.5 9 9 9 9 9 9 9 9 9 9 10 9 9 9 9 9 9
Casting 6 0.5 10 10 10 10 10 10 10 10 10 10 10 11 10 10 10 10 10
Casting 6 0.5 11 11 11 11 11 11 11 11 11 11 11 11 12 11 11 11 11
Figure 38 Another snapshot of the second optimization tool
The needed data for the application of this tool is:
 The time interval meeting the requirements of the analysis
 Sequence of processes for each furnace
 Duration of each process
 Casting furnace to casting unit assignment
 Hot metal distribution
 Desired furnace utilization with respect to time (e.g. per shift)
6.4.2 Objective Function

The objective function for the second optimization model is shown in Equation (6.15).
The aim is to minimize the costs due to hot metal waiting, idle time of the furnace and
longer batch lead time. The duration of processes is determined as constant in the opti-
mization part of the study, so the batch lead time is set as constraint in the optimization
model but can be measured in the simulation model. Therefore, it is integrated to the
objective function of the model but does not have any impact on the calculation.
95
The indices used in the optimization model:
 furnace Furnaces in the casthouse
 castingline Casting units in the casthouse
 batch Produced unit of a customer order
 timeslot Slots that shift duration is equally divided in
 shift Shifts in the optimization period
 process Processes take place at the furnace
The parameters used in the optimization model:
 m Total number of furnaces in the casthouse
 l Total number of casting units in the casthouse
 n Total number of produced batches
 t Total number of slots
 s Total number of shifts
 p Total number of processes take place at the furnace
 FurAvfurnace shift Number of batches assigned to the furnace per shift
 CLCapcastingline Capacity of the casting line
 FurProCapfurnace Process duration capacity of the casting furnace (min)
 HMCapshift Total planned hot metal amount for the shift (ton)
 AmHMfurnace batch shift Amount of hot metal per furnace per batch per shift(ton)
 Capfurnace Capacity of the furnace (ton)

96
 BatAmbatch Amount of the batch (ton)
 PHTP Planned hot metal transportation period in the casthouse
 BatDurbatch Lead time of the batch
 ProDurbatch process Duration of each process in the batch
 PrNoIdle Process number of “Idle process”
 ReqEn Required energy to cast one ton aluminium (kWh / ton)
 EnCost Energy cost (Euro / kWh)
 TotPro Annual production amount of the casthouse (ton/yr)
 SysCap System capacity (total capacity of the casting furnaces)

(ton)
 YrtoHr Unit conversion from year to hour (hr / yr)
The decision variables of the optimization model:
 FurCSfurnace timeslot Current status of the furnace per timeslot
 CLUticastingline Current usage of the casting line
 FurProAllfurnace Total process duration allocated to the casting furnace

(min)
 HMTransshift Amount of hot metal transportation in the shift (ton)
 WTHMfurnace batch shift Waiting time of hot metal per furnace per batch per shift
(hr)
 TrHMfurnace batch shift Time of hot metal transportation for the defined batch at
the furnace in the shift (hr at the current shift according to
the shift begin)
97
 IdTFurfurnace Idle time of the furnace (hr)
 Delbatch Delay in the batch lead time (hr)
The objective function of the production scheduler is formulated in Equation (6.15):
Minimize COST_HM + COST_FI + COST_OLD
(6.15)
subject to;
FurCSfurnace timeslot  {0, 1}  furnace,  timeslot
(6.16)

t
timeslot1
( FurCSfurnace timeslot) = FurAvfurnace shift  furnace,  shift
(6.17)
CLUticastingline <= CLCapcastingline  castingline
(6.18)
FurProAllfurnace <=FurProCapfurnace  furnace
(6.19)
HMTransshift = furnace1 ( 
m n
batch 1
( AmHMfurnace batch shift) )  shift
(6.20)
HMTransshift = HMCapshift  shift
(6.21)

p
BatDurbatch = process1
( ProDurbatch process)  batch
(6.22)
98

n
IdTFurfurnace = batch 1
( ProDurbatch process) for process = PrNoIdle  furnace,  batch
(6.23)
WTHMfurnace batch shift = max[0, (TrHMfurnace batch shift – PHTP)]  furnace,  batch,  shift
(6.24)

p
Delbatch = BatDurbatch - process1
( ProDurbatch process)  batch
(6.25)
where;
COST_HM Cost due to the hot metal waiting time of the casting furnace
COST_FI Cost due to the idle time of the furnace in between two batches
COST_OLD Cost of longer batch lead time than planned duration
6.4.3 Equations of Costs in the Objective Function of the Production

Scheduler
Hot metal waiting time
The possible strategies behind the hot metal schedule are discussed in detail in Chapter
4. Uncertainties such as resource availability, operational delays may defer the sched-
ule. These delays generate extra waiting time for casting furnaces or full crucibles and
cause energy loss and additional logistical activities. This part of the objective function
contains the extra cost occurred due to this non-value added waiting time. Equation
(6.26) shows the formula of the cost which determines the waiting time of hot metal and
its amount.
99
Cost_HM=CoeffEnLotoCost * furnace1 (  
m n s
batch 1
( shift1
( WTHMfurnace batch shift *
AmHMfurnace batch shift)))
(6.26)
where;
CoeffEnLotoCost Coefficient of the equation for the cost conversion
Idle time of casting furnaces
In between two batches the waiting time except the preparation for the next batch is
determined in this part of the objective function. Casting furnaces cannot be totally
turned off during inactive state. While there is no production in the furnace, it has to
operate with low temperature. Therefore, there is heat loss which is compensated with
additional energy. Equation (6.27) calculates this non-value added cost occurred due the
furnace idleness.
Cost_FI= CoeffEnLotoCost * furnace1 ( IdTFurfurnace * Capfurnace)

m
(6.27)
where;
Longer batch lead time
As mentioned above the cost occurred due to longer batch lead time is kept in the objec-
tive function but has not any influence on the result. In the optimization model, the du-
ration of processes is assumed to be constant and does not change during the calcula-
tion. However, in the simulation model the duration of the operations may change due
to the dynamics. This variance can be determined by measuring the length of the batch
lead time. The impact of the cost is reflected to the optimization model after interfacing
with simulation model. The data flow from the simulation to the optimization contains
the duration of processes measured in the simulation model.
100
The main disturbances causing costs are delays in material handling, traffic in the cast-
house, transportation unit availability, waiting for the casting line etc. Equation (6.28)
shows the formula to calculate the cost of longer batch lead time.

n
Cost_OLD= CoeffEnLotoCost * batch 1
( Delbatch * BatAmbatch)
(6.28)
where;
6.4.4 Equation of the Conversion Coefficient

The cost calculation for the second tool has the unit “ton * hr” without the coefficient
which is tagged as “CoeffEnLotoCost”. The required total energy for the whole produc-
tion is used to convert this result to real cost. The calculation is done for the whole year
with total production and then split into the unit “ton * hr”. The coefficient can also be
calculated by considering the profit loss. But then the impact of this loss has to be de-
termined for the whole facility which is not easy to calculate without mapping the inter-
nal supply chain of other sub facilities located in an aluminium smelter. In this concept,
the focus is more on losses due to waste in the system boundary. Therefore, the energy
loss is taken into account to calculate the coefficient shown on the Equation (6.29).
CoeffEnLotoCost = (ReqEn * EnCost * TotPro) / (SysCap * YrtoHr)
(6.29)
6.5 Interfacing with the Simulation Model

Interface between the simulation model and the optimization model performs to share
the data via MS Excel. Firstly, the optimization model calculates system alternatives
and then sends them to the simulation model. Secondly, the simulation model responds
with the system performance by converting its output in an appropriate form which be-
comes the input for the optimization part.
101
Figure 39 Data transfer between simulation and optimization models
Figure 39 shows the data flow between the simulation and optimization models. The
main items flowing from the optimization to the simulation are production planning,
operation scheduling and warm-up period for simulation runs. After running the simula-
tion model with this input data, average duration of processes, order status and uncer-
tainties are sent to the optimization part. After each communication, each model reruns
itself to analyze the impact of the output received from the other tool.
6.5.1 Methodology of the Interfacing

As described before, the thesis has two optimization models. The interface between op-
timization models and simulation model is separately set up. Each optimization model
has its own communication path and the data flow has not any influence to the other.
The interface between simulation and optimization models is done in a static way so the
simulation model can receive and transfer data to both models in the same instant. But
the output gained from the simulation model depends on the result of both optimization
models.
The optimization models do not have any data transfer in between. It is possible to run
the simulation model by interfacing with only one of the optimization models. The steps
of the methodology of interfacing shown on the Figure 40 are:
102
1. The optimization model runs with the initial assumed data and determines the
required result with respect to its objective function. The output of the optimiza-
tion model is transferred to MS Excel table which is prepared in the format as an
input for the simulation model. This step is valid for both optimization models.
2. The simulation model reads the input data and builds its logic with this data.
Simulation duration depends on the required time frame. It also depends on the
time interval of the input data received from the optimization model. The calcu-
lated time interval of the input data in the optimization model cannot be shorter
than the simulation run period.
3. The simulation model creates the output and inserts it into MS Excel platform
for the further run of the optimization model.
4. As a last step, the optimization model reads the output data of the simulation
model and reruns its logic for the next round of the interface.
The repetition of the communication loop depends on the accuracy expected by the us-
er.
103
Figure 40 Methodology of the interface between the simulation and optimization models
6.5.2 Architecture of the Interface

After developing the methodology, the structure is built to interface the models. Four
layers of application fields are shown on Figure 41. The first layer which is defined in
MS Excel consists of the data of customer orders and the casthouse facility specifica-
tions. Customer orders contain the information about product properties and delivery
information. Specifications comprise operational control issues, resource properties,
system characteristics etc. Both data is transmitted to the second layer which includes
the optimization model and its macros. This second part is also programmed in MS Ex-
cel platform.
104
Figure 41 Layers of application fields in the interface structure
The third layer of the interface is the output of the optimization model which is received
by the simulation model. This part contains two main MS Excel data, the batch plan of
each furnace and the production schedule for the casting furnaces. The batch plan in-
cluding the data about recipe of the batch, production specifications, and material stor-
age locations gathers information from both optimization tools. On the other hand, the
production schedule which contains the start time of processes and the planned duration
of the batches is formed only by the second optimization tool. When one of the optimi-
zation models is not activated then the presumed specifications are transmitted to the
simulation model.
The simulation model stands at the last layer of the interface which reads the input data
directly from MS Excel. Automod is capable of interfacing with MS Excel to read and
write data. Therefore, the batch plan and the production schedule are prepared in the
format which is converted to the simulation environment by Automod. This feature of
Automod helps to send the output data to the optimization model by writing them in MS
Excel. The simulation model as output changes the status of the customer orders and
modifies the system specifications. The data sent from the simulation model to the op-
timization model is such as;
 The status of the order

105
 The average duration of each process
 The delays due to resource unavailability
 The distribution of hot metal
In the evaluation phase of this research, after the required runs of the models, the cost
values in the objective function are received from both the optimization and the simula-
tion models. The optimization model gives the optimum result according to the parame-
ters defined and modified by the simulation model. The simulation model at the end
reaches a conclusion for the value of the objective function with respect to input gained
from the optimization model. The final result obtained from the simulation model may
differ from the value gained in the optimization model because of the uncertainties. The
simulation model maps the flow interactions and the operations in a dynamic way.
Therefore, the result obtained from the simulation model illustrates what happens if the
batch plan and the production schedule received from the optimization model are ap-
plied to the casthouse supply chain.
106
Description of the Case Study
7 Description of the Case Study
An example case is built to test both the simulation and the optimization models and to
verify the interface. In this chapter, the case study with respect to layout and material
flow specifications is described. Additionally, parameter set for both model concepts
according to assumptions is done. The results gathered from the models are based on the
properties of the casthouse and characteristics of the facility described in this chapter.
7.1 Casthouse System Characteristics

The aim of creating a case study is to validate the gain from interfacing the simulation
and optimization models. For that reason, during the set up period of the case study,
system characteristics of the casthouse are considered to touch each single point in both
model concepts. As a final, one type of end product is selected. The high range of dif-
ferentiation in characteristics of customer orders influences to select Primary Foundry
Alloy (PFA) ingots as a final product type.
Both optimization and simulation models have casting furnaces in the center of their
concepts, so the example created to verify the study is set up by focusing on the casting
furnaces. Two different groups of casting furnaces are considered to see the variances of
the furnace specifications on the results. The layout of the casthouse is built by combin-
ing different characteristics of casthouses owned by Hydro Aluminium. Before discuss-
ing the details of the layout of the system, the production units in the internal supply
chain of the casthouse are identified.
107
Group 1 Group 2
Number of casting furnaces 3 2
Number of casting line 1 1
Capacity of the furnaces (ton) 30 50
Required hot metal per furnace per batch (ton) 20 30
Filling Frequency per furnace per day 3 3
Table 6 Production units in the casthouse
Production units in the casthouse in terms of group dispersion are shown on Table 6.
Number of casting furnaces per group, their capacities, hot metal requirements and fill-
ing frequencies per day are highlighted. Detailed explanation of production units in the
casthouse and their specifications are:
 The casthouse determined for the case study contains two casting lines and two
casting furnace groups with five furnaces in total. The groups are split according
to the capacity and the assigned casting lines. Each group transfers the metal to
its own casting line.
 The first group has three casting furnaces. Each furnace in this group has capaci-
ty of 30 tons. From the planning perspective, two casting furnaces cannot trans-
fer the metal to the casting line at the same time. At a time only one casting fur-
nace fills the casting line, and others have to wait until the end of the process if
they are also ready for casting.
 The second group has two casting furnaces. Each furnace in this group has ca-
pacity of 50 tons. The same planning rule is also valid for this group that two
casting furnaces cannot cast the metal at a same instant.
 The day is split into three shifts which have eight hours duration. This assump-
tion is valid for the simulation and optimization models.
 The filling and casting frequency of each furnace is determined as one time per
shift. This implies that the casting line of the first group has nine and the casting
line of second group has six times filling operation during a day.
108
 The hot metal rate to the batch amount is assumed to be constant for each fur-
nace in the case study. For the first group, 20 tons out of 30 tons of a batch and
for the second group, 30 tons out of 50 tons of a batch are covered by the hot
metal supplied from the electrolysis.
 In addition, the batch contains cold metal, external scrap or internal production
scrap material. Amount of alloy addition is ignored.
According to these specifications, annual and daily transported hot metal amount from
the electrolysis and the capacity of the casthouse are calculated as shown on Table 7.
Total daily capacity of the casthouse is determined as 570 tons which brings annual
production around 208000 tons. Electrolysis supplies 131400 tons of hot metal to reach
this annual production amount.
Casting Line 1 Casting Line 2 Total Capacity

Daily casting capacity (ton) 270 300 570
Annual casting capacity (ton) 98550 109500 208050
Daily required hot metal (ton) 180 180 360
Annual required hot metal (ton) 65700 65700 131400
Table 7 Annual and daily required hot metal and production capacity of the casthouse
7.1.1 Layout Description

The layout of the casthouse established for the case study is shown on Figure 42. Be-
sides the arrangement of the storage locations which are described in detail at subchap-
ter 7.1.2, the figure shows the paths followed by the vehicles which are represented with
purple lines. The casting furnaces and the connected casting lines can also be seen on
the top-left hand side of the layout.
The fluxing process is integrated to the supply chain before filling the hot metal to the
casting furnaces. The location of the fluxing stations is also highlighted on Figure 42.
Although the fluxing process is only mapped as duration in the simulation model, tap-
ping vehicles travel to this location to fulfill the process. During the fluxing process
duration, both the crucible which carries hot metal and also the vehicle that transports
the full crucible are occupied.
109
Figure 42 Layout of the casthouse defined for the case study
Rectangle with black dotted lines on Figure 42 represents the border of casthouse facili-
ty. Outside of its border, there are some storage areas for cold metal and end products
which belong to casthouse internal supply chain. The reasons to keep them outside the
casthouse are their big capacities and their contact to external trucks. Incoming and out-
going materials are transported with external trucks which are not allowed to enter the
boundary of the casthouse.
7.1.2 Storage Specifications

Important logistical activities in the internal casthouse supply chain related to storage
management are stocking of end products, surplus and incoming cold metal (containing
also external scrap material and alloys), their inventory management and relocation
strategies. Locations of the storages have to be well determined to reduce unnecessary
movements. For the case study, the layout of storages is taken from a casthouse owned
110
by Hydro Aluminium. Figure 42 shows the storage locations on the layout. The defini-
tions and properties of the storages are:
 There are four cold metal storages in the casthouse internal supply chain, one is
outside and three of them are located inside of the facility
 The outside metal storage is used to unload the incoming material from the sup-
plier. The incoming material concept contains both cold and external scrap met-
als.
 Cold metals are transported from the outside storage to the drying storage inside
the casthouse. The drying process is considered as duration in the simulation
model. Therefore, each cold metal supplied from the customer spends the drying
duration in this storage before filled into the casting furnace.
 After the drying process cold metal is transported to one of two cold metal stor-
ages located inside the casthouse. These internal cold metal storages are separat-
ed according to the casting furnace groups, so the cold metals are split to these
storage areas depending on the related casting furnaces.
 There are two end product storage locations in the finished goods area (shown
on Figure 43) assigned to each casting line. End products are stocked as a bundle
or containerized depending on the customer order specification. The container
filling operation takes also place in these storages.
 The casthouse layout contains one storage area for the rejected products by the
production. The internal scrap materials are directly transported to the casting
furnaces according to the batch plan.
 There is one surplus storage area for the finished products which are not as-
signed to any customer order. The materials are kept waiting in this storage for a
while to be allocated to the subsequent customer orders.
111
7.1.3 Vehicle Fleet

Vehicle fleet of the casthouse is composed of forklifts with different stack capacities,
skimming machine and front loader. Besides this internal fleet, there are tapping vehi-
cles controlled by both electrolysis and casthouse units. Vehicles in the casthouse, their
tasks and handling capacities are shown on Table 8. Details of their properties are:
 Two “7.5 ton” forklifts assigned to transport cold metal between the storages
and also from storages to casting furnaces
 One “5 ton” forklift” assigned to unload external trucks carrying incoming mate-
rials such as cold metal, external scrap, alloys etc.
 One “32 ton” forklift allocated to transport empty and full containers in the fin-
ished product storage (shown on Figure 42). This forklift is also assigned to load
and unload the external outgoing trucks.
 Two “5 ton” forklifts transport the bundles of PFA ingots from the conveyor of
casting line to storage or directly into the container. The conveyors are placed
behind the casting units. They transport the finished products from casting unit
or packaging station to intermediate storage. Forklifts transport the products
from this storage to permanent end product storage.
 One “5 ton” forklift transport bundles of PFA ingots from production unit to the
surplus or internal scrap storage
 One skimming machine is used for skimming and stirring processes of the cast-
ing furnaces. This vehicle has a special mechanism to handle a skimming beam
for the skimming process.
 One front loader assigned to transport dross bins to and from the dross cooling
area. This vehicle works in coordination with the skimming machine. It trans-
ports the empty dross box in front of the furnace before the skimming process
starts. And after the process it transports the full dross box to the dross cooling
area.
112
 Besides the internal vehicle fleet of the casthouse, four tapping vehicles are allo-
cated to transport hot metal from the electrolysis to the casthouse. These tapping
vehicles have their own equipment on board to empty the crucible into the cast-
ing furnace.
Id No Vehicle Type Task Definition Handling Capacity

1 Forklift Cold Metal and Scrap 7.5 ton
2 Forklift Cold Metal and Scrap 7.5 ton
3 Forklift Unloading External Trucks 7.5 ton
4 Big Forklift Container Transport 32 ton
5 Small ForkLift End Product Area 5 ton
6 Small ForkLift End Product Area 5 ton
7 Small ForkLift Surplus & Production Scrap 5 ton
8 Skimming Machine Skimming and Stirring 1 SKU
9 Front Loader Dross Bin Carrying 1 SKU
Table 8 Vehicle fleet of the casthouse
The allocation of transportation areas which are split according to transported material
and appointed vehicle types are shown on Figure 43. The areas for incoming materials
involve the transported material types of cold metal and external scrap material. The
first part of it highlights the unloading fields of the materials, another part internal
stocking areas and the last part buffer locations in front of the furnaces. In this area,
three “7.5” tons forklifts operate and according to the workload they can share the tasks.
The second storage area group contains finished good, surplus and production scrap. In
this part of the layout three small forklifts and one big forklift operate. If the surplus or
the production scrap is assigned to any casting furnace to be re-melted again, then mate-
rials are transported into the incoming materials area to be filled into furnace.
The routing of tapping vehicles can be seen schematically on Figure 43. Tapping vehi-
cles enter the casthouse from the left hand side after visiting the fluxing station and
leave from the right hand side. This course helps to prevent any accidents due to the
vehicle turning and encounter passing the gates. One-way rule is a typical safety control
strategy in casthouses. Additionally, Figure 43 shows the parking location of the skim-
ming machine and the skimming tool. The parking position of the skimming machine is
located between the casting furnace groups to access all furnaces in a short time.
113
Figure 43 Areas according to transported material
7.1.4 Casting Furnace Specifications

Some specifications of the casting furnaces in the casthouse can be seen on Table 9.
Assigned casting line and the capacity of the furnaces are discussed in subchapter 7.1.
Melting rate of each furnace can also be seen at the table. First three furnaces belonging
to the same furnace group have the melting rate of “6 tons / hr” and the last two have “8
tons / hr”. The values are considered as average melting rates of cold metal, external
scrap material and production scrap.
Another data gathered from Table 9 is the hatch number of the furnaces and its position
according to the casting line. All of the furnaces in the casthouse contain one hatch
(furnace lid) and this is located at the front side when the orientation of casting line is
considered behind the furnace. These hatches are the access points of the furnaces for
the vehicles to fill metal in, to skim and to stir the furnace.
114
Id No Name Type Casting Line Index Capacity [ton] Melting Rate (ton/hr) Hatch Number Hatch Position
1 PFA1 Casting furnace 1 30 6 1 Front

Table 9 Some specifications of the casting furnaces
Processes of casting furnaces and their sequence in the simulation model are discussed
in Chapter 4 in detail. Their short definitions and sequence in the case study are:
1. Cold metal addition; addition of pure aluminium ingots to balance the produc-
tion. The preparation process is also considered as duration in the cold metal ad-
dition process when needed.
2. Pre-heating; melting the cold metal addition
3. Hot metal filling; filling primary metal coming from electrolysis
4. Alloying; adding alloy depending on the order
5. Holding; holding the temperature constant
6. Casting; end product casting process. Casting furnace empties the metal to the
casting line during this process
After the casting process, if the next batch does not start immediately, the furnace will
turn to inactive state. There is not any production in the furnace but it is also not totally
“off”, so this state is called “idle”.
7.2 Parameter Set for the Simulation Model

The simulation model contains a plenty of assumptions to create a virtual environment
by imitating the whole supply chain. In the simulation model, the decision support sys-
tem controlling the material flow has to be well determined. The hot material flow spec-
ifications, material types used in the study and storage characteristics are discussed in
115
this subchapter in detail. All of the items in the supply chain have to be defined in Au-
tomod as a separate simulation object. The last part of this subchapter contains these
items and the concept highlighting how they are integrated into Automod.
7.2.1 Primary Hot Metal Flow Specifications

Primary hot metal, as described in Chapter 3 in detail, is the metal transferred between
electrolysis and casthouse units. Therefore, it depends also on the specifications of elec-
trolysis to initiate the flow. Table 10 shows the specifications of the electrolysis de-
signed for the case study by considering characteristics of a modern smelter. The elec-
trolysis in the study contains one pot line having 240 pots in total. The tapping cycle
which is used to calculate the number of crucibles per shift is also defined as 40 hours in
the table. Hot metal transportation per shift, day and year compensate the required
amount by the casthouse system specifications.
Potline 1
# of Pots 240
Potline Specification # of Rows 2
# of Pots/Row 120
Tapping Cycle [hr] 40
[ton/Cycle/Pot] 2.5
[ton/Shift] 120
Metal Production (avg)
[ton/day] 360
[ton/year] 131400
Pots/Crucible 2
Tapping per Shift Pots/Shift 48
Crucibles/Shift 24
Table 10 Electrolysis specifications
Hot metal flow specifications and process durations can be seen on Table 11. All of the
data related to hot metal flow specifications are used to calculate the primary hot metal
distribution which is defined in Chapter 4 for the simulation model. Besides the re-
source specifications, operation and transportation durations are also defined in the
same table. This data helps to map the transportation between electrolysis and casthouse
116
in an accurate way. All of the data described on Table 10 and Table 11 is integrated into
hot metal transportation part of the simulation model.
Hot Metal Flow Specifications

Shift Duration [hr] 8
Crucible Design Capacity [ton] 5
# of Vehicles in Operation per Shift 4
# Crucible on Board 1
Number of Pots per Crucible 2
Operation and Transportation Durations
Tapping Duration per Pot [min] 2
Average 30
Driving Duration Round Trip
Shortest 25
outside Casthouse [min]
Longest 35
Average Fluxing Operation Duration (min) 5
Average Tilting Operation Duration (min) 3
Table 11 Hot metal flow characteristics
7.2.2 Material Types and Storage Descriptions

All material types defined in the case study and mapped in the simulation model, except
the primary hot metal are described in Table 12. The table contains the category of the
material type, the amount been carried per transport, the vehicle index number matching
to the fleet defined in Table 8 and the handling duration containing the picking up and
setting down durations of processes.
All of these material types are defined as separate loads in the simulation model. Figure
44 shows the loads definition window of Automod where each load item defined sepa-
rately. Alloys in the simulation model are grouped and they are clustered as one load
type.
117
Material Material Transportation Vehicle Handling

Id No
Type Category Capacity (SKU) Id No Duration (sec)
1 Bundle Ingot 2 1,2,3 45
2 T-Bar Ingot 2 1,2,3 45
3 Sow Ingot 2 1,2,3 45
4 Top-Bottom Pieces Pure Metal 2 7 45
5 Surplus Pure Metal 2 7 45
6 Production Scrap Pure Metal 2 7 45
7 Litho Scrap Scrap 1 1,2,3 60
8 Bale Scrap Scrap 1 1,2,3 60
9 Dross Pigs Scrap 1 1,2,3 60
10 Silicon Alloy 1 1,2,3 30
11 Mangan Alloy 1 1,2,3 30
12 Magnesium Alloy 1 1,2,3 30
Table 12 Material types and their properties
As a storage strategy “Advanced” mode is selected among the storage strategies defined
in the subchapter 4.5.4. The material type and their categories required for the advanced
mode are identified on Table 12. The storage bays are classified according to the mate-
rial type and the filling strategy is applied based on these types. The incoming material
is firstly unloaded to the outside storage and then allocated to the internal storage.
Id No Storage Description Material Type Id No Capacity (SKU) Warm-up level (SKU)

1 Outside Storage 1,2,3,7,8,9,10,11,12 250 150
2 Inside Storage (Drying) 1,2,3,7,8,9 50 50
3 Storage Furnace Group 1 1,2,3,7,8,9,10,11,12 100 100
4 Storage Furnace Group 2 1,2,3,7,8,9,10,11,12 100 100
5 Internal Scrap Storage 4.5 50 20
6 Surplus Storage 5 50 20
7 Finished Product Storage (CL1) - 250 100
8 Finished Product Storage (CL2) - 250 100
Table 13 Storage descriptions and their properties
The storages in the casthouse supply chain and the material types assigned to them are
shown on Table 13. Their capacity and the warm-up level amounts at the beginning of
the simulation can also be found on the table. The internal metal storage classified also
according to the furnace groups that the materials are allocated to. There is a separate
storage for the internal production scrap and also another one for the surplus material
coming from the production. For the finished goods, two storages are considered and
split based on the casting line where the metal is casted.
118
7.2.3 Simulation Model Items in Automod

As mentioned in Chapter 4, Automod enables to define the objects and the control logic
separately. Figure 44 shows the defined items in the simulation model. The main objects
(loads, resources, queues) and the source files containing the control logic are highlight-
ed on the figure. Besides these items, vehicle fleet and paths are defined in the “Path-
Mover” part of Automod.
Figure 44 Simulation model items described in Automod
The “Loads” category contains the materials and the control load which is used to drive
the logic of the simulation model. Each material type has its own load image and attrib-
ute in the model. The control load which is used to control the logic in the simulation
model triggers the source files containing processes and functions. It cannot be visual-
ized during the simulation as a flowing item but can be traced to see any information
attached to it.
The “Resources” group consists four main items that can be claimed by the loads during
the simulation. These four items specify the casting line, casting furnaces, skimming
119
device and crucibles used to transport the hot metal. Their capacity, availability and
status lead the material flow in the model. Item numbers of each resource are also men-
tioned under the column “Dimension” of the “Resources” window in Automod.
The “Queues” part contains the storage units. The container feature of a queue defini-
tion in Automod enables to design the storages in three dimensions. It is possible to give
capacity to axes of container to stack the material in the queue. The exact position of the
load in the storage can be identified by this feature.
The “Source Files” category helps to define control logic of the model in Automod. In
the simulation model several source files are created. These files contain the logic to
control the batch management, production in the casting line, casting furnace cycle
management, metal balance in the furnace, shipping of finished goods, delivery of in-
coming materials, hot metal planning and transportation, warm-up period at the begin-
ning of the simulation, traffic in the casthouse, navigation concept and the results ob-
tained from the model.
7.3 Parameter Set for the Optimization Models

It is required to define the system characteristics of casthouse supply chain of the case
study in detail to set the values of parameters in the objective functions and constraints
of optimization models. Strategies behind the operational management of the casthouse
are also considered during this parameter set. The last part of the chapter contains the
information about the assumptions made for the optimization models which are based
on the specifications of real casthouses.
7.3.1 Parameters for the Short-term Production Planner

Considering the costs in the objective function of the short-term planner, impact of sys-
tem definition of the case study and related assumptions are:
 Casting lines in the casthouse are assumed to cast batch wise which increase the
impact of production scheduling.
120
 Each batch belongs to one customer. In other words, the customer orders have
different specifications so it is not possible to split a batch to the different cus-
tomers.
 Orders from Europe have higher priority due to the penalty rate. This affects the
value of “OPriorder” in Equation (6.9) which is used to calculate late delivery cost
(COST_LD) in the objective function.
 Orders have to be ready three days before the shipment. This period is necessary
to handle the official procedure to send the products out of the country.
7.3.2 Parameters for the Production Scheduler

The production scheduler requires the specifications about the processes handled in the
casthouse supply chain. The first assumption of the production scheduler is about the
hot metal transportation from electrolysis to the casting furnace. The hot metal filling
process at the casting furnace schedule can take place only at the first five hours of the
shift. If the filling process happens in the last three hours then it will create non-value
added cost for the casthouse. This affects the value of “WTHMfurnace batch shift” in Equa-
tion (6.26) which is used to calculate “COST_HM” by setting “PHTP” value as “5” in
Equation (6.24).
The restriction helps to plan the hot metal flow between electrolysis and casthouse
units. The rest of the shift, both facilities have time to compensate their production
schedule. With this restriction, “WTHMfurnace batch shift” can be calculated by using Equa-
tion (7.1).
WTHMfurnace batch shift = max[0, (TrHMfurnace batch shift – 5)]  furnace,  batch,  shift
(7.1)
where;
 WTHMfurnace batch shift Waiting time of hot metal per furnace per batch per shift
(hr)
121
 TrHMfurnace batch shift Time of hot metal transportation for the defined batch at
the furnace in the shift (hr at the current shift according to
the shift begin)
The second assumption is about the duration of processes which has influence on the
optimization model. Process durations of the casting furnaces have to be defined at the
beginning of the interface to be able to run the optimization tool “short-term production
planner” at first. These process durations are changed after interfacing with the simula-
tion model. One of the data flow from the simulation model to the production scheduler
is the new duration of each process.
Pre-assumed Process Durations (min)

Process Name Furnace Group 1 Furnace Group 2
Cold Metal Addition 30 30
Pre-heating 60 60
Hot Metal Filling 60 60
Alloying 30 30
Holding 60 60
Casting 120 180
Batch Lead Time 360 420
Idle 120 60
Table 14 Pre-assumed process durations of the furnace groups
Table 14 shows the pre-assumed duration of each process according to the furnace
groups. The batch lead time of each furnace is calculated at the end, which is shorter
than the duration of the shift for both groups. Therefore, the rest of the shift is assumed
as idle for the furnaces because of “one batch production per shift” restriction. Figure 45
shows the duration of processes and their allocation for each furnace group in a shift.
122
min
Figure 45 The duration of processes and their allocation for each furnace group in a shift
7.3.3 Parameters for the Coefficients of Optimization Models

Casthouse specifications which contain the system capacity and annual production
amount were discussed in subchapter 7.1 in detail. These two parameters are used to
calculate the coefficient of the tool production scheduler. There are some other items
that have not been discussed. Table 15 shows the rest of the required parameters to cal-
culate the coefficient of both optimization models. Their short definitions, values, units,
equations containing the parameter and the source where the value is obtained are pre-
sented on this table.
Abbreviation Short Definition Value Unit Source

SalesPr Sales price of the order 1700 "Euro / ton" [BR12]
PenRate Penalty rate 1 "% per day" [TCE10]
ManHrC Man hour and equipment cost 5 "Euro / SKU / day" Assumed
InvC Inventory cost 5 "Euro / SKU / day" Assumed
CMC Market price of external cold metal 1200 "Euro / ton" [BOR12]
ReqEn Required energy for 1 ton aluminium 2500 "kWh / ton" [BCS07]
EnCost Energy cost 0.04 "Euro / kWh" [DA10]
Table 15 Required parameters for coefficients of optimization models

123
Scenarios & Evaluation
8 Scenarios & Evaluation
In this chapter, the results obtained from the simulation and optimization models are
discussed. Two scenarios are created to analyze the effects of optimization models on
the casthouse supply chain and the impact of their communication with the simulation
model. The description of two scenarios and their analysis are the main focus of this
chapter.
Scenarios handled in this chapter are based on the case described in detail in Chapter 7.
The basic characteristics of the casthouse are kept the same for both scenarios. The
main difference between the scenarios is the hot metal ratio in the production capacity.
This difference enables to change the production amount independently from the
amount of hot metal transported per shift.
8.1 Scenario I: Fixed Production Amount

“Scenario I” represents the case with a fixed hot metal ratio which restricts the produc-
tion amount of the casthouse. The fixed ratio concept is preferred in some casthouses
having constant cold metal supply with high cost. This assumption causes single batch
production per furnace per shift. The ratio of hot metal to total batch amount is deter-
mined around 63%.
The production plan and schedule are prepared for the reference case, without interfac-
ing the optimization models, according to the historical data analysis of a real casthouse.
The aim of the production plan is to sequence the orders according to the ascending
delivery dates. The production schedule intends not to utilize the casting lines by two
furnaces at the same time which enables to reduce the waiting time of the furnaces.
Batches are assigned to furnaces according to the shift length, which arranges the pro-
duction start time at the same hours of the shifts for each furnace. Disturbances on the
124
material flow and delays due to resource unavailability may interrupt this sequence if
the batch lead time exceeds the shift duration.
8.1.1 Analysis of Scenario I with and without Optimization Models

Analysis of scenario I is divided into phases and steps. Phases are specified according to
existence of communication between the simulation and the optimization models. This
analysis is split into four phases:
1. Phase 0: represents the reference case in which the simulation model is run with
the prearranged production plan and schedule prepared for this analysis.
2. Phase 1: represents the case in which only optimization model 1 is integrated in

to the analysis. Data transfer between optimization model 1 and the simulation
model takes place in this phase.
3. Phase 2: represents the case in which only optimization model 2 is integrated in

to the analysis. Data transfer between optimization model 2 and the simulation
model takes place in this phase.
4. Phase 3: represents the next level of the analysis where both of the optimization
models are integrated into.
125
Figure 46 Stepwise visualization of the analysis of Scenario I
In each phase, the values for the analysis are obtained from the simulation model and
the optimization models separately for the non-value costs defined in each objective
function. As described in Chapter 6, the objective functions show the non-value added
logistical costs in primary aluminium casthouse internal supply chain. Figure 46 shows
the stepwise visualization of the analysis of Scenario I. Name of the steps highlights the
phase and the communication step of this phase. For example, step 1.2 represents the
second step of the first phase.
Optimization models provide only the total value of the costs defined in their own ob-
jective functions. On the other hand, the simulation model generates results for the total
cost amount of each objective function. Depending on the phase, the data flows from
the simulation model to the optimization model which is integrated into this phase. For
example in the phase 1, the data transfer takes place only between the simulation model
and the short-term production planner optimization model.
In each phase, step 2 represents the half of the interfacing loop, the concept of which
was defined in detail in the subchapter 6.5. The loop is completed at the end of step 4,
when the data flow takes place in both directions. In step 3.4, the data flows to the simu-
126
lation model from both optimization models. The results obtained at the end of step 3.4
from the simulation model contain the impact of both optimization models.
Opt_1 Opt_2 Sim_1 Sim_2 Sim_Total

Reference Case Step 0 52680 31513 84193
Step 1.1 14280
With Optimization Step 1.2 21770 31513 53283
Model 1 Step 1.3 13200
Step 1.4 13200 31513 44713
Step2.1 27875
With Optimization Step2.2 53419 34975 88394
Model 2 Step2.3 29125
Step2.4 53471 28363 81834
With Both Models Step3.4 14020 28363 42383
Table 16 Optimization and simulation model results with and without optimization tools
The values of objective functions obtained from both the simulation and optimization
models are shown on Table 16. “Optimization Model 1” represents the short-term pro-
duction planner and “Optimization Model 2” is used for the production scheduler opti-
mization model. The time frame of the analysis is considered as two days and the results
gained from the simulation model are based on this two days run period. The reason of
setting the time frame to two days is that updating of the production schedule takes
place in two days interval which was explained in detail at Chapter 3.3.1.
The abbreviations on Table 16 and their explanations are;
Opt_1= Value obtained from optimization model 1 showing the non-

value added logistical costs defined in the objective function of
optimization model 1 in Euro
Opt_2= Value obtained from optimization model 2 showing the non-

Sim_1= Value obtained from the simulation model showing the non-
127
Sim_2= Value obtained from the simulation model showing the non-
value added logistical costs defined in the objective function
optimization of model 2 in Euro
Sim_Total = Sum of the values of Sim_1 and Sim_2
Figure 47 Non-value added logistical cost supplied from two days simulation run
The results of these phases can be seen graphically on Figure 47. According to the val-
ues of objective functions presented on Table 16 and Figure 47, the following conclu-
sions can be drawn:
 After each step in interfacing the short-term production planner optimization

model with the simulation model, the reduction in non-value added costs in-
creases. After completing half of the loop in the interfacing concept (Step 1.2),
the reduction in the total costs reached in both objective functions is around
30000 Euro for two days. By completing the whole loop of in the interfacing
methodology (Step 1.4) additional 9000 Euro is gained.
 Analysis of phase 1 shows that optimization model 1 does not have any impact
on the costs defined in the objective function of optimization model 2. In step 0,
step 1.2 and step 1.4, the value obtained from the simulation model for the costs
of objective function of optimization model 2 does not change.
128
 If the simulation runs only with optimization model 2 without any data flow in
the direction from the simulation to the optimization model (step 2.2), the costs
in the objective functions are increased. This means that the system creates more
non-value added costs compared to the reference case. The reason of this result
is that the predefined parameters (in step 2.1) for the optimization tool 2 have
considerable difference compared to the parameters obtained from the simula-
tion run (in step 2.3).
 On the other hand, after completing the interfacing concept, by reaching to step
2.4, the result gets better than the reference case by around 3 %.
 After completing the interfacing loop, values obtained from the optimization
models and the simulation model become closer. For example, the difference in
values of Opt_1 in step 1.3 and Sim_1 in step 1.4 is smaller than the difference
of Opt_1 in step 1.1 and Sim_1 in step 1.2. This convergence is also valid for
optimization model 2.
 After completing the interfacing concept with both optimization models (step
3.4), the value of Sim_1 gets higher than the value in phase 1.4. The reason of
this result is that the production scheduler tool reduces the batch lead time which
causes for some batches to be produced earlier than the other cases. This early
production brings extra end product storage cost (COST_EPS) which increases
the objective function of model 1. On the other hand, the value of objective
function 2 is not influenced from optimization model 1.
 The total non-value added logistical costs in the reference case are reduced near-
ly by half after interfacing with both optimization models (step 3.4). Haller et.
al. [HKK07] calculated the aluminium production cost around 1800 “Euro/ton“.
The cost reduction on non-value added logistical activities in the casthouse sup-
ply chain renders around 2 % of this cost.
129
8.1.2 Detailed Analysis of Scenario I with and without the Short-term

Production Planner
The results of detailed non-value added logistical costs analysis done with and without
short-term production planner are shown on Table 17. To remind the cost items in each
objective function, definition of cost abbreviations are:
COST_LD = Punishment cost of late delivery
COST_EPS = Inventory cost for waiting at the end product storage
COST_SR = Cost of surplus product
COST_HM = Cost due to the hot metal waiting time of the casting furnace
COST_FI = Cost due to the idle time of the furnace in between two batches
COST_OLD = Cost of longer batch lead time than planned duration
With Optimization
Reference Case Model 1
Step 0 Step 1.1 Step 1.2 Step 1.3 Step 1.4
COST_LD 0 0
COST_EPS 9280 8200
Opt_1
COST_SR 5000 5000
TOTAL 14280 13200
COST_LD 0 8400 0
COST_EPS 7680 8370 8200
Sim_1
COST_SR 45000 5000 5000
TOTAL 52680 21770 13200
COST_HM 7425 7425 7425
COST_FI 22363 22363 22363
Sim_2
COST_OLD 1725 1725 1725
TOTAL 31513 31513 31513
Sim_Total 84193 53283 44713
Table 17 Detailed cost analysis with optimization model 1 (the short-term production planner)
According to the graph (shown on Figure 48), it can be concluded that the value of
“COST_SR” after implementing optimization model 1 equals to one ninth of the value
of the reference case. This 89 % reduction in the surplus cost affects the total value of
the objective function. Another conclusion drawn from the graph is that the late delivery
130
cost appears in the step 1.2 “with optimization model 1”. This cost is eliminated after
completing the loop of interfacing the simulation and optimization model (step 1.4).
The reduction in the cost evaluation, after implementing a short-term production plan-
ner optimization model, helps to increase the overall profit of aluminium production
(according to Pahladsingh [Pah04]) by around 10 % compared to reference case.
Figure 48 Analysis of cost items in the objective function of tool 1 (the short-term production plan-
ner)
8.1.3 Detailed Analysis of Scenario I with and without the Production

Scheduler
The detailed cost distribution of scenario I with and without the production scheduler
optimization model is shown on Table 18. The cost improvement with the second opti-
mization tool is not as high as the saving obtained with the short-term production tool.
A single batch production per furnace per shift restriction is the main reason of this re-
sult. This assumption prevents to reduce the costs occurred due to furnace idleness and
order lead time which have the biggest portion of the value of the objective function.
Figure 49 shows distribution and the comparison of the cost items in each case of sce-
nario I on the graph.
131
With Optimization
Reference Case Model 2
Step 0 Step2.1 Step2.2 Step2.3 Step2.4
COST_HM 4875 10500
COST_FI 23000 18625
Opt_2
COST_OLD 0 0
TOTAL 27875 29125
COST_LD 0 0 0
COST_EPS 7680 8419 8471
Sim_1
COST_SR 45000 45000 45000
TOTAL 52680 53419 53471
COST_HM 7425 4250 4188
COST_FI 22363 29138 22125
Sim_2
COST_OLD 1725 1588 2050
TOTAL 31513 34975 28363
Sim_Total 84193 88394 81834
Table 18 Detailed cost analysis with optimization model 2 (the production scheduler)
Figure 49 Analysis of cost items in the objective function of tool 2 (the production scheduler)
The achievement of reduction in the non-value added logistical costs with production
scheduler equals to 3 % of the total required energy in the casthouse production. The
amount of total energy cost for the annual production of casthouse is calculated by us-
ing the values presented on Table 15 (in Chapter 7).
Figure 50 and Figure 51 show the distribution of process durations for each furnace in
two days simulation run period. It is expected to produce six batches per each furnace
during this time frame, due to the assumption of one batch production per shift per fur-
nace. The warm-up durations at the beginning of the first batches of each furnace are
excluded from the graphs. Therefore, it can be seen that on both graphs only two fur-
132
naces per each case can succeed to finish six batches production in a two days period.
PFA1, PFA2, PFA3, PFA4 and PFA5 represent the casting furnaces in casthouse. Their
specifications were discussed in the previous chapter.
The percentage of total non-value added waiting time of the casting furnaces in step 2.4
equals to 25.5 % of the overall production time. For the reference case, this value is
calculated as 26.9 %. For two days period 1.4 % reduction in the waiting time of the
furnaces nearly equals to two days more production in a year basis.
One of the assumptions of the case study is that hot metal transportation takes place in
the first five hours of the shift, so the last three hours cause additional cost to the sys-
tem. Figure 52 and Figure 53 show hot metal distribution occurring in the last three
hours of shifts. The data is taken from the simulation model. The improvement in the
step 2.4 compared to reference case appears not only in the amount of the cost value but
also in the number of the furnaces. In other words, the hot metal filling process in the
last three hours of shifts after interfacing with optimization tool 2 takes place in less
number of furnaces compared to the reference case.
133
Figure 50 Process cycle of furnaces as output of the simulation model without optimization models
(reference case)
Figure 51 Process cycle of furnaces as output of the simulation model after interfacing with optimi-
zation model 2 (Step 2.4)
134
without optimization models (reference case)
after interfacing with optimization model 2 (Step 2.4)
135
Figure 54 Batch lead times of furnaces supplied from the simulation model without optimization
models (reference case)
Figure 55 Batch lead times of furnaces supplied from the simulation model after interfacing with
optimization model 2 (step 2.4)
Figure 54 and Figure 55 show the batch lead time distribution of each furnace for six
shifts period. As mentioned before some furnaces as shown on the graphs cannot
achieve finishing the batch production in the last shift. On both graphs it can be ob-
served that some batches exceed the value of the shift duration (eight hours). In this
136
case, at the beginning of the next shift the furnace finishes the batch from the last shift
and then starts the planned batch of the current shift.
To be able to analyze the effect of optimization model on precision of batch lead time,
the statistical parameter “Relative standard deviation (RSD)” can be used. RSD is a
measure of stability which is calculated by Equation (8.1)
RSD = (σ / µ) * 100
(8.1)
The mean value, standard deviation and RSD of batch lead times for the reference case
and the analysis with interfacing the optimization model 2 (Step 2.4) are shown on Ta-
ble 19.
Without Optimization Models With Optimization Model 2

(Reference Case) (Step 2.4)
µ 7.71 7.68
σ 1.45 1.38
RSD 18.81% 17.97%
Table 19 The mean value, standard deviation and RSD values for both steps
RSD value of the reference case is 18.81 % whereas after the interface with optimiza-
tion model 2 (step 2.4) RSD value decreases to 17.97 %. Based on these numbers, it can
be concluded that the case with the optimization model is more stable than the reference
case.
8.2 Scenario II: Variable Production Amount

During the detailed analysis of scenario I with the production scheduler optimization
model, a possible production amount increase was recognized. Fixed ratio of hot metal
to the batch size limits to increase the production capacity in the casthouse. In this part
of the analysis, a new scenario is prepared by removing this restriction. This new as-
sumption has direct impact on the objective function of the second optimization model.
Therefore, the analysis of this scenario is done only with the production scheduler.
137
8.2.1 Definition of Scenario II

After recognizing possible increase in production due to shorter lead time, scenario II
with adding more cold metal to the batch recipe is considered. This allowance enables
the system not to have a fixed ratio of hot metal to the production amount. In scenario I,
the hot metal ratio was assumed to be around 63 %. However, due to the melting rate of
the furnaces minimum hot metal rate is assumed to be 55 % for scenario II. Therefore,
hot metal ratio in the casting furnace may vary between 55 % and 65 % for this new
scenario depending on the situation in the casthouse. On the other hand, this causes
more melting duration compared to scenario I. The restriction of having one batch pro-
duction per furnace per shift is kept for this scenario due to the availability of casting
lines.
With this scenario, it is possible to increase the production capacity by replacing re-
quired hot metal with cold metal. This analysis is useful in the real case when electroly-
sis cannot supply the demand or cold metal prices go down in the market.
8.2.2 Analysis of Scenario II with and without the Production

Scheduler Tool
To compare the results of scenario II with the values of scenario I in the same situation,
Table 20 reminds the results of the analysis of scenario I with and without production
scheduler. The amount of production is 1030 tons in the simulation model for each
phase of the analysis of scenario I. With the variable batch content, it is expected that
the amount of production will increase. Table 21 shows the results of scenario II with
and without the second optimization model. For the reference case “without optimiza-
tion tools” the values are the same as the scenario I because without optimization tools
it is not possible to reflect the dynamic metal ratio to the production schedule.
It can be concluded that with this new scenario, the production amount may increase.
Additional to that, it is achieved having higher production amount with less non-value
added costs. On Table 21, it can be seen that the simulation after interfacing with opti-
138
mization model (step 2.4) has lower values than the analysis without optimization mod-
el (reference case).
By comparing the results with and without optimization model for scenario II, the re-
duction on the non-value added logistical costs is calculated as around “7 euro / ton” for
the case with optimization tool. This reduction is reached with not only due to cost min-
imization but also more production. Additional important output of the analysis is that
hot metal waiting and furnace idleness costs are reduced in the case with optimization
tool while the cost due to batch lead time is increased. The explanation of this increase
is higher utilization of casting line which causes longer waiting time for casting of fur-
naces.
Without Optimization Models With Optimization Model 2, With Optimization Model 2,

(Reference Case) Scenario I (Step 2.2) Scenario I + Interfacing (Step 2.4)
Cost_HM 7425 4250 4188

n
Cost_FI 22363 29138 22125

io
at
Cost_OLD 1725 1588 2050

ul
Sim
Costs in
TOTAL 31513 34975 28363
Objective
Cost_HM 4875 10500
Function
n
Cost_FI 23000 18625

iot
iza
Cost_OLD 0 0
ti m
Op
TOTAL 27875 29125

Production Simulation TOTAL (ton) 1030 1030 1030
Amount
Optimization TOTAL (ton) 1060 1060
Table 20 Results for scenario I with and without optimization model 2 (the production scheduler)
Without Optimization Models With Optimization Model 2, With Optimization Model 2,

(Reference Case) Scenario II (Step 2.2) Scenario II + Interfacing (Step 2.4)
Cost_HM 7425 6663 4513

n
Cost_FI 22363 20200 18000

io
at
Cost_OLD 1725 2600 2850

ul
Sim
Costs in
TOTAL 31513 29463 25363
Objective
Cost_HM 4313 10000
Function
n
Cost_FI 8688 10625

iot
iza
Cost_OLD 0 0
ti m
Op
TOTAL 13000 20625

Production Simulation TOTAL (ton) 1030 1090 1090
Amount
Optimization TOTAL (ton) 1120 1090
Table 21 Results for scenario II with and without optimization model 2 (the production scheduler)
Figure 56 presents the graph of the comparison of results for the scenarios with and
without optimization tools. The benefit gained from interface of simulation and optimi-
zation models can be seen on the graph. The difference in values of the objective func-
139
tion of simulation and optimization tools gets smaller after interfacing the models.
Without the interface, the optimization model delivers lower cost value compared to the
simulation model. As mentioned before, the results obtained from the simulation model
represent more reliable values compared to values from optimization model due to
mapping the dynamic interaction between the operations.
Figure 56 Scenario comparison with and without optimization model 2 (the production scheduler)
As a conclusion optimization models help to reduce the non-value added logistical costs
in the primary aluminium internal supply chain. Completing the loop of the interfacing
concept until the last step (step 4) renders to reduce the cost values in considerable
amount. In scenario I, the short-term production planner optimization model helps to
reduce the costs more than the production scheduler. On the other hand, the results of
analysis with the second optimization model indicate possible increase in production
amount which was analyzed in scenario II.
140
Research Summary & Future Work
9 Research Summary & Future Work
Primary aluminium smelters attempt to reduce operational expenditure to survive in the

current competitive market situation. The technological developments in aluminium
industry focus on how to reduce the production costs. It is also recognized that logistical
activities in the facilities have potential to be improved. Therefore, the focus of the re-
search was to quantify and to reduce the redundant logistical costs in aluminium cast-
house facility.
9.1 Research Summary

The main objective of this thesis was to develop a new concept that controls the non-
value added logistical costs in primary aluminium casthouse supply chain. Steps that
were followed to reach this main objective and the summary of the results obtained
from the thesis are:
1. It was verified that the casthouse unit has potential to eliminate non-value added
logistical costs in its internal supply chain.
2. The alternative approaches to control logistical activities such as hot metal flow,
casting furnace control and storage management enabled to develop a universal
simulation platform which is capable of being implemented to any casthouse.
3. The boundary of the primary aluminium casthouse internal supply chain was
simulated without mapping the hot metal flow in electrolysis unit. For this, a
primary hot metal distribution concept was developed for the simulation model
covering the metal flow occurring in the electrolysis.
4. The navigation concept that dynamically manages the traffic in the casthouse
simulation model was developed and implemented. The required decisions in the
141
vehicle routing were taken by using this tool. It enabled the user to model and
control complicated traffic situations in the simulation program.
5. Defining the input data of the simulation model in MS Excel created flexibility
to modify the model without changing the programmed logic in the simulation
tool. Casting furnace control, vehicle fleet of the casthouse, customer orders, de-
livery plan for incoming materials, the storage control strategies and the batch
plan of the casting furnaces can be modified in these MS Excel sheets.
6. It was verified that the simulation method was an efficient evaluation tool for the
casthouse aluminium supply chain. Some standard results (e.g. resource utiliza-
tion, storage analysis, traffic density, safety and bottleneck analysis etc.) which
are valid for each casthouse can be obtained from the simulation model.
7. It can be concluded that it is possible to apply a lean thinking approach to the

continuous process sectors like aluminium industry.
8. With the combination of simulation and lean thinking approach, non-value add-
ed logistical costs in the aluminium casthouse supply chain were specified. Sim-
ulation as an analyzing tool enabled the user to elaborate the details in the sup-
ply chain.
9. A short-term production planner was developed to split customer orders to the

batches and dispatch the batches to the casting furnaces. It was achieved to re-
duce the non-value added logistical costs occurred due to poor production plan-
ning such as late delivery, early production and surplus rate.
10. A production scheduler optimization model was developed to sequence the pro-
cesses of the casting furnaces. Reduction in hot metal waiting time, furnace idle
time and batch lead time were succeeded after applying this model.
11. A new concept was developed to interface the discrete event simulation with lin-
ear optimization. Data transfer was achieved between two models. The output of
one of the models served as input of the other.
142
12. After interfacing with the simulation model, both of the optimization models
separately succeeded in improving the casthouse supply chain by reducing the
non-value added logistical costs specified by lean thinking approach. The study
gave the best result after interfacing both of the optimization models at the same
time with the simulation model.
13. Besides the reduction in non-value added costs, the interface between models
provided more accurate results of the optimization models compared to the anal-
ysis without simulation interface. And also the values of the objective functions
determined in simulation and optimization models were converged.
14. A real based case study helped to test the models. It also highlighted the possible
increase in casthouse production amount besides the reduction in the logistical
costs. The scenario for more production was also developed and tested in the
case study.
9.2 Recommendations for Future Work

In this thesis, it was verified that the interfacing between the simulation and optimiza-
tion models helps to reduce the non-value added logistical costs in the primary alumini-
um casthouse supply chain. However, further investigations can be done for cases con-
taining the extended supply chain boundary or different industrial requirements. The
potential points for the further research are recommended as:
1. A simulation model can be prepared for the electrolysis unit which will interface
with the casthouse simulation model to identify and analyze the effects of other
logistical activities occurring in electrolysis on the hot metal flow.
2. The analysis period was kept as two days for each model. This timeframe can be
increased because the simulation model is capable of running limitless when the
required data is supplied.
143
3. The interfacing between the simulation model and the optimization models was
done manually in the research. The data flow may be progressed automatically
by using a computer program providing a dynamic interface.
4. The selected final product type for the case study can be changed to analyze the
possible impacts on the optimization models with different product types. Espe-
cially the short-term production planner may be affected due to the restrictions
of product dimensions.
144
References
References
1. [AAC08] Araujo S., Arenales M., Clark A., Lot-sizing and Furnace Scheduling
in Small Foundries, Computers and Operations Research, Vol. 35-3, 2008, Pp.
916-932
2. [Abd03] Abdullah F., Lean Manufacturing Tools and Techniques in The Process
Industry with A Focus on Steel, Ph.D Thesis, School of Engineering, University
of Pittsburgh, 2003
3. [AR07] Abdulmalek F.A., Rajgopal J., Analyzing the Benefits of Lean Manu-
facturing and Value Stream Mapping via Simulation: A Process Sector Case
Study, International Journal of Production Economics, Vol. 107, 2007, Pp. 223–
236
4. [ARA10] Awheda A., Rahman M.N., Arshady H., Key Factors Influencing the
Effectiveness and Efficiency of Supply Chain Management in Small and Medi-
um Enterprises (SMES) in Malaysia, 15th International Symposium on Logis-
tics, 2010
5. [BADW04] Buxmann P., Ahsen A., Diaz L.M., Wolf K., Usage and Evaluation
of Supply Chain Management Software – Results of an Empirical Study in the
European Automotive Industry, Info Systems Journal 14, 2004 Pp. 295-309
6. [Ban04] Banks, J.: Getting Started with AutoMod. Second Edition. Chelmsford,
MA: Brooks Automation, Inc., May 2004.
145
References
7. [BB99] Beddoes J., Bibby M.J., Principles of Metal Manufacturing Processes,

Butterworth-Heinemann, 1999
8. [BBB11] Berlioux J., Baudrenghien J.L., Bourgier A., New Casthouse Smelter
Layout for the Production of Small Nonalloyed Ingots: Three Furnaces/Two
Lines, Light Metals, TMS(The Minerals, Metals & Materials Society), 2011, Pp.
635-639
9. [BBJLM02] Banks J., Buckley S., Jain S., Lendermann P., Manivannan M.,
Panel Session: Opportunities for Simulation in Supply Chain, Proceedings of the
2002 Winter Simulation Conference, Vol. 1, 2002, Pp. 1652-1658
10. [BBTM09] Baxter R., Bouk T., Tikasz L., McCulloch R.I., Simulation-Based
Approach for Validating a Lean Anode Plant Configuration, Light Metals, TMS
(The Minerals, Metals & Materials Society), 2009 , Pp. 1043-1048
11. [BC02] Borge G., Cooper P., Recent Studies of Compacted Additives Including
Industrial Scale Tests, Light Metals, TMS(The Minerals, Metals & Materials
Society), 2002
12. [BCS07] BCS, U.S. Department of Energy, U.S. Energy Requirements for Alu-
minum Production: Historical Perspective, Theoretical Limits and Current Prac-
tices, Industrial Technologies Program Energy Efficiency and Renewable Ener-
gy, 2007
13. [Bea98] Beamon, B. M., Supply Chain Design and Analysis: Models and Meth-
ods, International Journal of Production Economics, 55(3), 1998, Pp. 281–294
14. [Bee01] Beeley P.R., Foundry Technology, Edition 2, Elsevier, 2001

146
References
15. [Ben97] Benson D., Simulation Modeling and Optimization Using ProModel,
Proceedings of the 1997 Winter Simulation Conference, Vol. 1, 1997, Pp. 587-
593
16. [BM00] Barnett M.W., Miller C.J., Analysis of the Virtual Enterprise Using Dis-
tributed Supply Chain Modeling and Simulation: An Application of e-SCOR,
Proceedings of the 2000 Winter Simulation Conference, 2000, Pp. 352-355
17. [BMND05] Branke J., Middendorf M., Noeth G., Dessouky M., Waiting Strate-
gies for Dynamic Vehicle Routing, Transportation Science, Vol. 39, No. 3,
2005, Pp. 298–312
18. [BOR12] Benefits-of-Recycling.com, Aluminum Recycling Prices,

www.benefits-of-recycling.com/aluminumrecyclingprices.html, 2012
19. [BPS08] Böse F., Piotrowski J., Scholz-Reiter B., Autonomously Controlled
Storage Management in Vehicle Logistics - Applications of RFID and Mobile
Computing Systems, International Journal of RF Technologies: Research and
Applications , iFirst Article, 2008, Pp. 1–20
20. [BR12] Brandtzeg S.R., Rostrup J.C.A., First Quarter 2012: Lower Aluminium
Prices, Strong Energy Result, Hydro Aluminium, 2012
21. [Bra10] Braziotis C., Towards the Extended Enterprise: Clarifying Paradigms
and Concepts in Supply Chain Management, 15th International Symposium on
Logistics, 2010
22. [BW10] Babulak E., Wang M., Discrete Event Simulation, Discrete Event Simu-
lations, Sciyo, 2010
147
References
23. [CHH06] Chatfield D.C., Harrison T.P., Hayya J.C., SISCO: An Object-oriented
Supply Chain Simulation System, Decision Support Systems, 42, 2006, Pp. 422-
434
24. [Chi01] Chinneck J.W., Practical optimization: A Gentle Introduction, Electron-

ic Document, http://www.sce.carleton.ca/faculty/chinneck/po.html, 2001
25. [CM11] Campuzano F., Mula J., Supply Chain Simulation: A System Dynamics
Approach for Improving Performance, Springer, 2011
26. [CWD09] Chesonis D.C., Willliams E.M., DeYoung D.H., Meeting Environ-
mental Challenges in the Casthouse, Light Metals, TMS (The Minerals, Metals
& Materials Society), 2009 , Pp. 653-657
27. [DA10] Driscoll K., Adams R., The Strategic Impact of Changing Energy Mar-
kets on the Aluminium Industry, Light Metals, TMS(The Minerals, Metals &
Materials Society), 2010
28. [DD07] Dong L., Dingwei W., A Simulation Based Research on Loading –
Unloading Strategy in Railway Container Terminal, Proceedings of ICM’ 2007,
The 6th International Conference on Management, 2007, Pp. 180-185
29. [DD09] Dey T.K., Dinakar R., Linear Programming, Advanced Algorithms,
Department of Computer Science and Engineering, The Ohio State University,
2009
30. [Des110] Desjardins J., Storage Mode Overview, Concept Description, Internal
Documentation, IDEcraft, 2010
148
References
31. [Des210] Desjardins J., Technical Considerations for Further Step, Internal
Documentation, IDEcraft, 2010
32. [Dij59] Dijkstra E.W., A note on two problems in connexion with graphs,
Numerische Mathematik 1, 1959, Pp. 269–271.
33. [DK97] Drexl A., Kimms A., Lot Sizing and Scheduling - Survey and Exten-
sions, European Journal of Operational Research Vol. 99, 1997, Pp. 221-235
34. [DN08] Damiron C.,Nastasi A., Discrete Rate Simulation Using Linear Pro-
gramming, Proceedings of the 2008 Winter Simulation Conference, 2008, Pp.
740-749
35. [DO05] Duda J., Osyczka A., Multiple Criteria Lot-sizing in A Foundry Using
Evolutionary Algorithms, Lecture Notes in Computer Science, Vol. 3410, 2005,
Pp. 651-663
36. [DY00] Detty R.B., Yingling J.C, Quantifying Benefits of Conversion to Lean
Manufacturing with Discrete Event Simulation: A Case Study, International
Journal of Production Research, Vol.38-2, 2000, Pp. 429- 445
37. [EAA12] European Aluminium Association, http://www.alueurope.eu/, 2012
38. [Ebb01] Ebben M., Logistic Control in Automated Transportation Networks,

Ph.D. Thesis, University of Twente, 2001
39. [EKP96] Eklund P.W., Kirkby1 S., Pollitt S., A Dynamic Multi-source Dijks-
tra’s Algorithm for Vehicle Routing, Australian and New Zealand Conference
on Intelligent Information Systems, 1996, Pp. 329 - 333
149
References
40. [EVB01] Eick I., Vogelsang D., Behrens A., “Planning Smelter Logistics: A
Process Modeling Approach”, Light Metals, TMS(The Minerals, Metals & Ma-
terials Society), 2001, Pp. 393-398
41. [Fer12] Ferguson T.S., Linear Programming: A Concise Introduction, Electronic

Document, www.math.ucla.edu/~tom/LP.pdf, 2012
42. [FKM09] Fischer W.K., Keller F., Mannweiler U., Anode Plants for Tomor-
row’s Smelters: Key Elements for the Production of High Quality Anodes, Alu-
minium International Today, Volume 21 No. 1, 2009
43. [FKZM05] Freeman N., Kelly J.D., Zhang W., Mann J.L., Scheduling Optimiza-
tion for Aluminium Smelter Casthouses, Light Metals, TMS(The Minerals, Met-
als & Materials Society), 2005, Pp. 843-848
44. [Fri04] Friedrich B., Aluminium-Primärgewinnung, IME Institut für

Metallurgische Prozesstechnik, RWTH Aachen, 2004
45. [Fu01] Fu M.C., Simulation Optimization, Proceedings of the 2001 Winter Sim-
ulation Conference, 2001, Pp.53-61
46. [Fu02] Fu M., Optimization for Simulation: Theory vs. Practice, Journal on
Computing 14 (3), 2002, Pp. 192–215
47. [GAN06] Guan Y., Ahmed S., Nemhauser G. L., A Branch-and-Cut Algorithm
for the Stochastic Uncapacitated Lot-Sizing Problem?, Mathematical Program-
ming, Springer, 2006
150
References
48. [GH95] Ganeshan, R., Harrison, T. P., An Introduction to Supply Chain Man-
agement; Technical Report; Department of Management Science and Infor-
mation Systems, Penn State University, 1995.
49. [GK93] Grjotheim K. , Kvande H., Introduction to Aluminium Electrolysis Un-

derstanding the Hall-Héroult Process, Aluminium-Verlag, 1993
50. [GW88] Grjotheim K., Welch B.J., Aluminium Smelter Technology, A Pure and
Applied Approach, 2nd Edition, Aluminium-Verlag, 1988
51. [GKL99] Glover F., Kelly J.P., Laguna M., New Advances for Wedding Opti-
mization and Simulation, Proceedings of the 1999 Winter Simulation Confer-
ence, 1999, Pp. 255-260
52. [GNS03] Ganapathy S., Narayan S., Srinivasan K., Simulation Based Decision
Support for Supply Chain Logistics, Proceedings of the 2003 Winter Simulation
Conference, 2003, Pp.1013-1020
53. [GPG02] Gravel M., Price W.L., Gagne C., Scheduling Continuous Casting of
Aluminium Using A Multiple-objective Ant Colony Optimization Metaheutistic,
European Journal of Operational Research, Vol. 143, 2002, Pp. 218–229
54. [GS02] Gupta A.K., Sivakumar A.I., Simulation Based Multi-objective Schedule
Optimization in Semiconductor Manufacturing, Proceedings of the 2002 Winter
Simulation Conference, Pp. 1862-1870
55. [GT96] Goldberg A.V., Tarjan R.E., Expected Performance of Dijkstra’s Short-
est Path Algorithm, Technical Report TR-96-062, NEC Research, 1996
151
References
56. [Guo03] Guo X.J., Process Modeling and Intelligent System in Copper Smelter
– The Concept of Future Smart Smelter, 5th International Copper Conference,
Pyro metallurgy of Copper, 2003
57. [Har10] Harton T.W., Virtual Smelter Modeling for Metal Flow Management,
Proceedings of the 45th Annual Conference of the ORSNZ, 2010, Pp. 50-59
58. [Har99] Harland C.M., Supply Chain Management: Relationships, Chains and
Networks, British Journal of Management, Vol. 7, Special Issue, 1996, Pp. 63-
80
59. [Hie98] Hieta S., Supply Chain Simulation with Logsim-Simulator, Proceedings
of the 1998 Winter Simulation Conference, 1998, Pp. 323-326
60. [HKK07] Haller M., Kettle J., Kendrew G., Primary Aluminium Smelting and
Power Costs, Summary and Analysis, Vol. 3, Brook Hunt & Associates Ltd. ,
2007
61. [HMSW53] Hoffman A.J., Mannos M., Sokolowsky D., Wiegmann N., Compu-
tational Experience in Solving Linear Programs. Journal of the Society for In-
dustrial and Applied Mathematics. 1953, Pp. 17–33
62. [Hoe04] Hoefler J., Aluminium Primärerzeugung, Vorlesungsreihe an der

Fachhochschule Bonn-Rhein-Sieg, Hydro Aluminium Deutschland GmbH, 2004
63. [HS07] ten Hompel M., Schmidt T. Warehouse Management: Automation and
Organization of Warehouse and Order Picking Systems, Springer, 2007
64. [IAI03] International Aluminium Institute, The Aluminium Industry’s Sustaina-

ble Development Report, 2003
152
References
65. [ISW09] Imran A., Salhi S., Wassan N.A., A Variable Neighborhood-Based
Heuristic for the Heterogeneous Fleet Vehicle Routing Problem, European Jour-
nal of Operational Research Volume 197, Issue 2, 2009, Pp. 509–518
66. [ITC04] Industrial Technology Center, Principles of Lean Thinking Tools and
Techniques for Advanced Manufacturing, National Research Council Canada,
2004
67. [Jai03] Jain P.L., Principles Of Foundry Technology, Edition 4, Tata McGraw-
Hill Education, 2003
68. [Jao11] Jaouen G., Use of Process Simulation to Design A Billet Casthouse,
Light Metals, TMS(The Minerals, Metals & Materials Society), 2011, Pp. 641-
646
69. [JKG05] Jensson P., Kristinsdottir B.P., Gunnarsson H.P., Optimal Sequencing
of Tasks in An Aluminium Smelter Casthouse, Computers in Industry Volume
56, Issue 2, 2005, Pp.221-234
70. [JWCEL01] Jain S., Workman R.W., Collins L.M., Ervin E.C., Lathrop A.P.,
Development of a High-level Supply Chain Simulation Model, Proceedings of
the 2001 Winter Simulation Conference, 2001, Pp. 1129-1137
71. [KA03] Karim A., Adeli H., Radial Basis Function Neural Network for Work
Zone Capacity and Queue Estimation, Journal of Transportation Engineering,
ASCE, 2003, Pp. 494-503
72. [KKKF85] Karmarkar U.S., Kekre S., Kekre S., Freeman S., Load-sizing and
Lead-time Performance in A Manufacturing Cell, The Institute of Management
Sciences, 1985, Pp.1-9
153
References
73. [Kle05] Kleijnen J.P.C., Supply Chain Simulation Tools and Techniques: A
Survey, International Journal of Simulation and Process Modeling, Vol. 1, 2005,
Pp. 82–89
74. [KPM04] Kadar B., Pfeiffer A. and Monostori, L., Discrete-event Simulation for
Supporting Production Planning and Scheduling Decisions in Digital Factories,
Proceedings of the 37th CIRP International Seminar on Manufacturing Systems,
Digital Enterprises, Production Networks., 2004, Pp. 441-448
75. [KR04] Kaufman J.G., Rooy E.L., Aluminum Alloy Castings: Properties, Pro-
cesses, and Applications, American Foundry Society, ASM International, 2004
76. [KU07] Kaufmann H., Uggowitzer P.J., Metallurgy and Processing of High In-
tegrity Light Metal Pressure Castings, Fachverlag Schiele & Schoen, 2007
77. [Kva11] Kvande H., Introduction to the Industrial Hall-Héroult Process, Hydro
Aluminium, 2011
78. [KVHBD06] Koning H., Verver J.P.S., Heuvel J., Bisgaard S., Does R.J.M.M.,
Lean Six Sigma in Healthcare, Journal for Healthcare Quality, V. 28, N. 2, 2006,
Pp. 4–11
79. [KVKMP03] Kovács, A., Váncza, J., Kádár, B., Monostori L., Pfeiffer, A., Real
Life Scheduling Using Constraint Programming and Simulation, Proc. of IMS
7th IFAC Workshop on Intelligent Manufacturing Systems, 2003, Pp. 233–239.
80. [Lar09] Larrivée Y., Hot Metal Transport: A MES Role in Inter-Sector Data Ex-
change, Aluminium International Today, 2009
154
References
81. [LJI02] Liu C., Jula H., Ioannou P.A., Design, Simulation, and Evaluation of
Automated Container Terminals, IEEE Transactions on Intelligent Transporta-
tion Systems, VOL. 3, NO. 1, 2002, Pp.12-26
82. [LMKO01] Lakhal, S., Martel A., Kettani O., Oral M., On the Optimization of
Supply Chain Networking Decisions, European Journal of Operational Re-
search, 129(2), 2001, Pp. 259–270
83. [LPM01] Lendermann P., Ping G. B., McGinnis L.F., Distributed Simulation
with Incorporated APS Procedures for High-fidelity Supply Chain Optimization,
Proceedings of the 2001 Winter Simulation Conference, 2001, Pp.1138-1145
84. [Mat08] Matta A., Simulation Optimzation with Mathematical Programming

Representation of Discrete Event Systems, Proceedings of the 2008 Winter Sim-
ulation Conference, 2008, Pp. 1393-1400
85. [MDKMNSZ01] Mentzer J.T., DeWitt W., Keebler J.S., Min S., Nix N.W.,
Smith C.D., Zacharia Z.G., Defining Supply Chain Management, Journal of
Business Logistics, Vol.22, No. 2, 2001
86. [Meg91] Megiddo N., Linear Programming, Encyclopedia of Microcomputers,

School of Mathematical Sciences, Tel Aviv University, 1991
87. [Mei10] Meijer M., New Logistic Concepts for 400 and 500 kA Smelters, Light
Metals, TMS (The Minerals, Metals & Materials Society), 2010 , Pp. 345-348
88. [Mei11] Meijer M., Aluminium Smelter Logistics- Can These Bring Cost Sav-
ings?, Light Metals, TMS (The Minerals, Metals & Materials Society), 2011 ,
Pp. 421-423
155
References
89. [Mel05] Melton T., The Benefits of Lean Manufacturing: What Lean Thinking
has to Offer The Process Industries, Chemical Engineering Research and Design
Vol. 83-A6, 2005, Pp. 662–673
90. [Mey03] Meyer U., Average-case complexity of single-source shortest-paths al-

gorithms: lower and upper bounds, Journal of Algorithms 48, 2003, Pp. 91–134
91. [ML06] Maiwald D., Leisenberg W., Charging and Melting Management in
Primary and Secondary Casthouses, Innovatherm-Paper Casthouse, 2006
92. [MML08] Maticevic G., Majdandzic N., Lovric T., Production Scheduling
Model in Aluminium Foundry, Journal of Mechanical Engineering 54, 2008, Pp.
37-48
93. [MP00] Miller S., Pegden D., Introduction to Manufacturing Simulation, Pro-
ceedings of the 2000 Winter Simulation Conference, 2000, Pp. 63-66
94. [MRFK03] Mason S.J., Ribera P.M., Farris J.A., Kirk R.G., Integrating the
Warehousing and Transportation Functions of the Supply Chain, Transportation
Research Part E 39, 2003, Pp. 141–159
95. [MSL03] Mirchandani P., Syal R., Lucas D., Traffic Assignment Using Iterated
Route-based Simulation, Transportation Research Board 82nd Annual Meeting
Compendium of Papers (CD ROM), 2003.
96. [NO05] Nonas S.L., Olsen K.A., Optimal and Heuristic Solutions for A Sched-
uling Problem Arising in a Foundry, Computers & Operations Research, Vol.
32, No. 9, 2005, Pp. 2351-2382
156
References
97. [Ohn88] Ohno T., Toyota Production System: Beyond Large-Scale Production,
Productivity Press, 1988
98. [Oke84] O’keefe R.M., Developing Simulation Models: An Interpreter for Visu-
al Interactive Simulation, Ph.D. Thesis, University of Southampton, 1984
99. [Pah04] Pahladsingh R.R., Introduction of an Integrated Aluminium Production

Plant with High Economical Profit, Minimal Energy Consumption and Maxi-
mum Environmental Protection, Surinam Industrial Engineering, 2004
100. [PBNN02] Peres S.D., Brahimi N., Najid N.M., Nordli A., Uncapacitated
Lot-Sizing Problems with Time Windows, Research Report, 02/4/AUTO, Ecole
des Mines de Nantes, 2002
101. [PBTM11] Pires R.L., Baxter R., Tikasz L., McCulloch R.I., Integrated Ap-
proach for Safe and Efficient Plant Layout Development, Light Metals, TMS
102. [Pie08] Piehl T.S., Cell Batching Optimization for the New Zealand Alumin-
ium Smelter, Annual Conference of ORSNZ, 2008
103. [PN02] Peterson R.D., Newton L., Review of Aluminium Dross Processing,
Light Metals, TMS(The Minerals, Metals & Materials Society), 2002
104. [PP04] Pacciarelli D., Pranzo M., Production Scheduling in A Steelmaking-

Continuous Casting Plant, Computers and Chemical Engineering, No. 28, 2004,
Pp. 2823–2835
157
References
105. [Pri04] Prins C., A Simple and Effective Evolutionary Algorithm for the Ve-
hicle Routing Problem, Computers & Operations Research 31 ,2004, Pp. 1985–
2002
106. [PW92] Potts C.N., Wassenhove L.V., Integrating Scheduling with Batching
and Lot-sizing: A Review of Algorithms and Complexity, Journal of the Opera-
tional Research Society Vol. 43, 1992, Pp. 395-406.
107. [RS03] Rother M., Shook J., Learning to See: Value Stream Mapping to
Create Value and Eliminate Muda, Lean Enterprise Institute, 2003
108. [Rya98] Ryan D.M., Optimized Cell Batching for New Zealand Aluminium
Smelters Ltd, Annual Conference of ORSNZ, 1998
109. [Sau02] SauerJ., Multi-Site Scheduling Hierarchisch Koordinierte

Ablaufplanung auf Mehreren Ebenen, Habilitationsschrift, Fachbereich
Informatik, Universität Oldenburg, 2002
110. [SB10] Schriber T.J., Brunner D.T., Inside Discrete-Event Simulation Soft-
ware: How it works and Why It Matters, Proceedings of the 2010 Winter Simu-
lation Conference, Pp. 151-165
111. [SB99] Sadowski D., Bapat V., The Arena Product Family: Enterprise Mod-
eling Solutions, Proceedings of the 1999 Winter Simulation Conference, 1999,
Pp. 159-166
112. [Sch07] Schmitz C., Handbook of Aluminium Recycling, Vulkan-Verlag

GmbH, 2007
158
References
113. [SER06] Soriguera F., Espinet D., Robuste F., A Simulation Model for
Straddle Carrier Operational Assessment in a Marine Container Terminal, Jour-
nal of Maritime Research, Vol. 3,No. 2, 2006, Pp. 19-34
114. [SF96] Shekhar S., Fetterer A., Path Computation in Advanced Traveler In-
formation Systems, Proceedings of the 6th Annual Meeting and Exposition of the
Intelligent Transportation Society of America, 1996, Pp.304-312
115. [Shi89] Shingo, S., A Study of the Toyota Production System from An In-
dustrial Engineering Viewpoint, Productivity Press, 1989
116. [SKS03] Simchi-Levi D., Kaminsky, P., Simchi-Levi, E., Designing and
Managing the Supply Chain: Concepts Strategies and Case Studies, 2nd Edition,
McGraw-Hill ,2003
117. [Sum08] Summerauer W., Solution Methods for Lot-Sizing Problems – Mul-
ti-Level Models with and without Linked Lots, Diplomarbeit, Wien University,
2008
118. [SWW00] Schulz F., Wagner D., Weihe K., Dijkstra's Algorithm On-Line:
An Empirical Case Study from Public Railroad Transport, Journal of Experi-
mental Algorithmics, Volume 5(12), 2000
119. [TA03] Truong T.H., Azadivar F., Simulation Based Optimization for Sup-
ply Chain Configuration Design, Proceedings of the 2003 Winter Simulation
Conference, 2003, Pp.1268-1275
120. [TBPM10] Tikasz L., Read C.M., Baxter R., Pires R.L., McCulloch R.I.,
Safe and Efficient Traffic Flow for Aluminium Smelters, Light Metals, TMS
159
References
121. [TCE10] Tokyo Commodity Exchange, Aluminum Delivery Detailed Rules,

www.tocom.or.jp, 2010
122. [TK05] Tan R.B.H., Khoo H. H., An LCA Study of a Primary Aluminum
Supply Chain, Journal of Cleaner Production, 2005, Pp. 607–618
123. [TLRY01] Tang L., Liu J., Rong A., Yang Z., A Review of Planning and
Scheduling Systems and Methods for Integrated Steel Production, European
Journal of Operational Research No. 133, 2001, Pp. 1-20
124. [TM03] Totten G.E., Mackenzie D.S., Handbook of Aluminum: Alloy Pro-
duction and Materials Manufacturing, CRC Press, 2003
125. [TMPB12] Tikasz L. G., McCulloch R.I., Pentiah S.D., Baxter R. F., Simu-
lation Tools to Complement Cast House Design and Daily Operation, Light
Metals, TMS (The Minerals, Metals & Materials Society), 2012 , Pp. 993-997
126. [Tod01] Todd M.J, The Many Facets of Linear Programming, Springer-
Verlag 2001
127. [TS98] Tompkins J.A., Smith J.D., The Warehouse Management Handbook,
Second Edition, Tompkins Associated, North Carolina, 1998
128. [Tum96] Tumay K., Business Process Simulation, Proceedings of the 1996
Winter Simulation Conference, 1996, Pp. 93-98
129. [TW08] Tang L., Wang G., Decision Support System for the Batching Prob-
lems of Steelmaking and Continuous-casting Production, Omega, The Interna-
tional Journal of Management Science 36, 2008, Pp. 976 – 991
160
References
130. [UAKEN07] Upenieks V.V., Akhavan J., Kotlerman J., Esser J., Ngo M.J.,
Value-Added Care: A New Way of Assessing Nursing Staffing Ratios and
Workload Variability, The Journal of Nursing Administration, V.37, N.5, 2007,
Pp. 243-252
131. [VW03] Voss S., Woodruff D.L., Introduction to Computational Optimiza-

tion Models for Production Planning in A Supply Chain, Springer, 2003
132. [WEDS09] Winkelmann A., Eick I., Droste C., Segatz M., Logistic Simula-
tion Discrete Material Flow and Processes in Aluminium Smelters, Light Met-
als, TMS (The Minerals, Metals & Materials Society), 2009 , Pp. 523-527
133. [WG03] Williams E.J., Gunal A., Supply Chain Simulation and Analysis
with Simflex, Proceedings of the 2003 Winter Simulation Conference, 2003, Pp.
231-237
134. [WHP98] Whitman L., Huff B., Palaniswamy S., Commercial Simulation
over the Web, Proceedings of the 1998 Winter Simulation Conference, Vol. 1,
1998, Pp. 335-340
135. [Wis11] Wisner J.D., Principles of Supply Chain Management: A Balanced

Approach, Cengage Learning, 2011
136. [WJR91] Womack J., Jones D., Roos D., The Machine that Changed the
World., Harper-Perennial, 1991
137. [WW58] Wagner H.M., Whitin T.M., Dynamic Version of The Economic
Lot Size Model, Management Science Vol.5, 1958, Pp.212-219.
161
References
138. [YKSBT04] Yuan X.M, Khoo H.H., Spedding T.A., Bainbridge I., Taplin
D.M.R., Minimizing Total Setup Cost for A Metal Casting Company, Proceed-
ings of the 2004 Winter Simulation Conference, 2004, Pp. 1189-1194
139. [YP04] Yadlapati S. S., Park B., Development and Testing of Variable
Speed Limit Control Logics for Work Zones Using, Center for Transportation
Studies, the University of Virginia, Research Report No. UVACTS-13-0-43,
2004
140. [ZKB99] Zeigler B.P., Kim D., Buckley S.J., Distributed Supply Chain Sim-
ulation in A DEVS/CORBA Execution Environment, Proceedings of the 1999
Winter Simulation Conference, 1999, Pp. 1333-1340

Ilgin Diss

Uploaded by

Copyright:

Available Formats

Ilgin Diss

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ilgin Diss

Uploaded by

Copyright:

Available Formats

Concept Development to Control Non-value Added Logistical Costs

in a Primary Aluminium Casthouse

Von der Fakultät für Ingenieurwissenschaften,

zur Erlangung des akademischen Grades

Doktors der Ingenieurwissenschaften

Gutachter: Univ.-Prof. Dr.-Ing. Bernd Noche

Tag der mündlichen Prüfung: 06.09.2013

This thesis is dedicated to

Table of Contents ...................................................................................................................... IV

List of Figures ...........................................................................................................................VII

List of Tables ............................................................................................................................. IX

2 Literature Review ..........................................................................................................13

3 Casthouses in Primary Aluminium Smelter ................................................................24

4 Simulation Model Development for Aluminium Casthouse ......................................41

5 Lean Thinking Approach in the Primary Aluminium Casthouse Supply Chain.....71

6 Optimization Model Development ...............................................................................82

6.4.2 Objective Function ..............................................................................................94

7 Description of the Case Study .....................................................................................106

8 Scenarios & Evaluation ...............................................................................................123

9 Research Summary & Future Work ..........................................................................140

Figure 37 Snapshot of the second optimization tool.................................................................................. 93

Table 1 The major objectives of supply chain studies [BADW04] ................................................................ 4

1.1 Introduction to Fundamental Concepts of the Thesis

 supply chain as a focus area,

 primary aluminium casthouse as a core concept,

 discrete event simulation as an analyzing tool and

 linear optimization as an improvement method.

1.1.1 Supply Chain in General

Figure 1 Different supply chain definitions and their boundaries [MDKMNSZ01]

Major Objectives Authors and Year of the Study

Table 1 The major objectives of supply chain studies [BADW04]

1.1.2 Aluminium Smelter Casthouse

Figure 2 Primary aluminium production capacity in the last 10 years [HKK07]

Aluminium is mostly used in industrial fields such as construction, transportation, pack-

• Strong and light

• Highly corrosion resistant

• Easy to form and process

The facility called smelter is normally composed of three sub facilities:

• Carbon plant; supplies anodes to the electrolysis unit

Figure 3 Aluminium production and life cycle [IAI03]

1.1.3 Discrete Event Simulation

• New manufacturing processes, at the design and evaluation phase

• Existing processes, to improve the performance

• Operational area, to establish optimum policies

• Production field, to support planning and scheduling

1.1.4 Linear Optimization

• Variable bounds: represents the range of values of the variables

Standard representation of linear optimization is [Fer12]:

subject to a 11x1  a 12 x 2  ...  a 1n x n  b1

The objective function is represented as maximization problem in Equation (1.1). This

1.2 Research Motivation

• Electrolysis has still potential in its production division for improvement

• Energy consumption in electrolysis unit forms ~ 25 % of the production cost

• Complex operation techniques

1.3 Research Objectives

• Identification of material and information flow in the primary aluminium cast-

• Development of flexible discrete event simulation model which can be imple-

• Discovering the reasons of the non-value added logistical costs by approaching

• Development of linear optimization model reducing these redundant logistical