Increasing the Capacity of a Flow ... Technical and Organizational Issues

Increasing the Capacity of a Flow Shop:
Technical and Organizational Issues
by
Gavin J. DeNyse
BA Physics, Kalamazoo College, 1993
BSE Mechanical Engineering, University of Michigan, 1994
Submitted to the Department of Mechanical Engineering and
the Sloan School of Management
in partial fulfillment of the requirements for the degrees of
Master of Science in Mechanical Engineering and
Master of Science in Management
in conjunction with the Leaders for Manufacturing Program at the
Massachusetts Institute of Technology
June 2001
MASSACHUSETTS INSTITUTE
OF TECHNOLOGY
©2001 Massachusetts Institute of Technology. All Rights Reserved.
JUL 16 2001
LIBRARIES
Signature of Author
U
-I
Department of N4hanical Engineering
Sloan School of Management
May 11, 2001
Certified by
Stanley Gershwin, Thesis Advisor
Senior Research Scientist, Department of Mechanical Engineering
Certified by
Janice Klein, Thesis Advisor
Senior Lecturer, Sloan School of Management
Accepted by_
-Ain Sonin, Chairman of the Graduating Committee
Department of Mechanical Engineering
Accepted by
Margaret C. Andrews, Executive Director of the Sloan MBA Program
Sloan School of Management
This page is intentionally left blank
2
Increasing the Capacity of a Flow Shop:
Technical and Organizational Issues
by
Gavin J. DeNyse
Submitted to the Department of Mechanical Engineering and
the Sloan School of Management on May 11, 2001
in partial fulfillment of the requirements for the degrees of
Master of Science in Mechanical Engineering and
Master of Science in Management
Abstract
This thesis addresses the issue of increasing the capacity of a flow shop using existing assets.
Options for increasing capacity are identified and discussed. A network queueing model is
developed to analyze these options. Based on the model outputs, a change to the operation is
suggested. This change would result in a 20% increase in capacity, an 84% reduction in process
cycle time and an 81% reduction in work in process inventory.
The organizational issues of implementing the suggested change are investigated. A framework
for classifying the type of organizational change is introduced. This framework is applied to the
organization being studied to identify and analyze the challenges it faces. Strategies to overcome
these challenges and the key learning points for the researcher are discussed.
Thesis Advisors:
Janice Klein, Senior Lecturer, Sloan School of Management
Stanley Gershwin, Senior Research Scientist, Department of Mechanical Engineering
3
This page is intentionally left blank
4
Acknowledgements
I would like to thank the Leaders for Manufacturing Program. The last two years have been a
tremendous learning experience that has broadened my horizons, deepened my understanding and
introduced me to a group of extraordinary people I am honored to call my colleagues.
I would like to thank H. C. Starck Inc. for providing me this opportunity to learn. I appreciate the
patience of the employees who tolerated my poking, prodding, questioning and calculating. They
taught me more than they know about how best to use the ideas I have learned over the last two
years. In particular, I would like to thank my sponsor, Bob Pereira, whose support and openness
made my work possible.
I would like to thank my advisors, Stan Gershwin and Janice Klein, whose guidance helped make
this thesis what it is.
I would like to thank my family for supporting me as I reach for my dreams. My parents, Joanne
and Reid, were my first teachers and I stand upon their shoulders. My uncle has been a great
friend and supporter. His guidance and encouragement have helped to hatch some of my best
ideas and have helped to smooth the difficult transitions I faced over the last few years.
Finally, I would like to thank my wife, Bernetta. Her presence in my life is a blessing. She is a
continual source of love, inspiration and learning and I feel that anything is possible when we are
together.
5
Table of Contents
ABSTRACT .......................................................................................................
3
ACKNOW LEDG EIM ENTS .................................................................................
5
TABLE O F CO NTENTS ...................................................................................
6
FIG URES ......................................................................................................
.... 9
TABLES .............................................................................................................
10
CHAPTER 1:
11
INTRO DUCTIO N ....................................................................
1.1
Thesis Objective............................................................................................
11
1.2
Problem Statem ent ..........................................................................................
11
1.3
Thesis Overview ...........................................................................................
11
CHAPTER 2:
BACKG RO UND ......................................................................
13
2.1
Tantalum and Its Uses.....................................................................................13
2.2
The Com pany................................................................................................
14
2.3
M elt Shop .........................................................................................................
2.3.1
Inputs......................................................................................................
2.3.2
Products..................................................................................................
2.3.3
Current Processes...................................................................................
15
15
16
17
2.4
Evidence of a problem ...................................................................................
2.4.1
On-Time Performance............................................................................
2.4.2
Underproduction for Current Demand ....................................................
2.4.3
Projected Growth in Demand .................................................................
23
24
25
26
2.5
6
Response...........................................................................................................28
CHAPTER 3:
ANALYSIS .............................................................................
29
3.1
Increasing Capacity .....................................................................................
3.1.1
Identify the "Bottleneck" ........................................................................
3.1.2
Increase Availability ..............................................................................
3.1.3
Offload Work..........................................................................................
3.1.4
Decrease V ariability...............................................................................
3.1.5
Drum-Buffer-Rope.................................................................................
29
30
31
32
33
34
3.2
Queueing Network M odel ............................................................................
3.2.1
M odel Background.................................................................................
3.2.2
M odel Assumptions ................................................................................
3.2.3
M odel Inputs..........................................................................................
3.2.4
M odel Outputs .......................................................................................
35
35
36
36
37
3.3
Scenarios ......................................................................................................
Scenario Assumptions.............................................................................
3.3.1
3.3.2
Scenario Inputs .......................................................................................
Scenario Outputs......................................................................................
3.3.3
38
39
39
44
3.4
Discussion of Results......................................................................................49
3.4.1 Increased Operational Efficiency ...............................................................
Potential Issues.......................................................................................
3.4.2
50
51
CHAPTER 4:
54
IMPLEMENTING CHANGE ..................................................
4.1
Technical W ork versus Adaptive W ork ......................................................
4.1.1
Type I Situation .....................................................................................
Type 11 Situation .....................................................................................
4.1.2
4.1.3
Type III Situation...................................................................................
4.1.4
Summ ary ................................................................................................
54
55
55
56
56
4.2
Application of the Framework ....................................................................
4.2.1
Problem Definition.....................................................................................
4.2.2
Solution ..................................................................................................
4.2.3
Implem entation ........................................................................................
4.2.4
Observations ..........................................................................................
57
57
57
57
58
4.3
Barriers to Change .......................................................................................
4.3.1
Cultural Barriers .....................................................................................
Political Barriers .....................................................................................
4.3.2
4.3.3
Structural Barriers...................................................................................
59
59
61
63
4.4
64
Intern's Role.................................................................................................
7
CHAPTER 5:
CO NCLUSIO NS ...................................................................
66
5.1
Technical Conclusions ..................................................................................
66
5.2
Organizational Conclusions .........................................................................
69
REFERENCES ................................................................................................
70
APPENDIX A
71
: QUEUEING NETWORK MODEL ......................................
A.1
Background..................................................................................................
A.2
Notation............................................................................................................72
A.3
Governing Equations...................................................................................
APPENDIX B
71
73
: M ELT SHO P MODEL ........................................................
78
B.1
Current Arrivals to the System ...................................................................
B.1.1
Route I, Alloys E &D ............................................................................
B.1.2
Route II, Alloy Q ....................................................................................
B. 1.3
Route III, Alloy W & S ............................................................................
78
79
79
80
B.2
Routing Probabilities...................................................................................
81
B.3
Current Processing Rates............................................................................
83
B.4
50% Plasm a Processing Rates.........................................................................88
B.5
100% Plasm a Processing Rates.......................................................................90
B.6
Actual Data: Current...................................................................................
92
B.7
Scenario #1: Current ..................................................................................
93
B.8
Scenario #2: 50% to Plasm a........................................................................
94
B.9
Scenario #3: 100% to Plasma .........................................................................
95
APPENDIX C
8
: ECONOMIC ANALYSIS....................................................
96
Figures
Figure 1: Revenue Contribution by Business ............................................................
15
Figure 2: H. C. Starck Inc. and Its Flow of Tantalum, Carroll (2000).........................
16
Figure 3: M elt Shop Process Steps ...........................................................................
18
Figure 4: Route Ia ........................................................................................................
19
Figure 5: Route Ib ........................................................................................................
21
Figure 6: Route II.........................................................................................................22
Figure 7: Route III........................................................................................................23
Figure 8: Pareto Chart of the Reasons for Late Shipments .........................................
24
Figure 9: Orders and Deliveries, M elt Shop...............................................................25
Figure 10: Demand for Metallurgical Products, Historical and Forecasted.................27
Figure 11: Projected growth in demand for alloy E....................................................28
Figure 12: Cycle Time versus Utilization .................................................................
30
Figure 13: W orkstation Capacities for Each Scenario ...............................................
42
Figure 14: Current Process .......................................................................................
43
Figure 15: 50% to Plasma Process................................................................................
43
Figure 16: 100% to Plasma Process ..........................................................................
43
Figure 17: Throughput for Each Scenario .................................................................
45
Figure 18: Average Cycle Time for Each Scenario ....................................................
46
Figure 19: W ork In Process Inventory for Each Scenario...........................................
47
Figure 20: Utilization for Each Station in M elt Shop ...............................................
48
Figure 21: Revenue Contribution by Business ..........................................................
62
Figure 22: Historic and Projected Demand for M etallurgical Products.......................
67
Figure 23: Planned-for Capacity Results in Lost Sales.............................................
68
Figure 24: Using the Plasma Furnace Reduces Lost Sales.........................................
68
Figure 25: Simple Queueing Network .....................................................................
75
Figure 26: More Complex Queueing Network...........................................................76
Figure 27: Schematic of M elt Shop M odel ..............................................................
78
9
Tables
Table 1: Products of the Melt Shop............................................................................
17
Table 2: Workstation Inputs for All Three Scenarios ...............................................
40
Table 3: Comparison of Actual & Model for Current Processes................................
44
Table 4: Situation Types, Heifetz (1994) ...................................................................
55
Table 5: Routing Probabilities for the Three Scenarios .............................................
82
10
Chapter 1: Introduction
1.1 Thesis Objective
This thesis is the outcome of a joint research project between H. C. Starck Inc. and MIT,
under the auspices of the Leaders for Manufacturing Program. The research for this
thesis is conducted during a six-month period in 2000. The objective of this program is
to improve customer service performance for metallurgical products by increasing the
production capacity of the melt shop.
1.2 Problem Statement
The company is experiencing rapid increases in demand for their products made of
tantalum. This demand is expected to continue for the foreseeable future. Decreased
levels of customer service have indicated that one of the company's business units,
metallurgical products, is unable to meet this growing demand.
The company needs to add capacity to the metallurgical business in order to be able to
participate in the expansion and diversification in this market. Efforts are currently under
way to expand via acquisition. This thesis will not focus on this set of activities,
however. Instead, it will address what can be done to increase capacity with existing
assets. Focusing on existing assets will enable the company to increase capacity in less
time at low cost and little risk. Following this strategy will enable the company to meet
the expanding short-term demands while long-term solutions, i.e. acquisitions and new
facilities, are completed.
1.3 Thesis Overview
This thesis is intended to be an argument for the implementation of a strategy that uses
the current assets of the melt shop more effectively. By using an unused asset, the
plasma furnace, the company will be able to realize additional gross profits of $1.1
11
million annually and enjoy a one-time inventory reduction estimated at $1.3 million.
This recommendation is intended to be complementary to the capacity expansion being
planned. To this end, this thesis is composed of five chapters, the last four of which are
described below.
Chapter two provides the context for the analysis. It will introduce the product, tantalum,
and its uses. Next, it will discuss the company and the operational unit to be studied, the
melt shop. Finally, it will illustrate the motivation for the project and frame the issue to
be addressed.
Chapter three analyzes the melt shop with the intention of identifying a way to increase
its capacity with the assets currently available. To accomplish this, the techniques to
increase capacity are introduced and discussed. An analysis tool, a queueing network
model, is presented. This tool is applied to the melt shop by developing three different
scenarios and examining the resulting outputs.
Chapter four discusses the issues associated with implementing the change suggested by
the analysis. An analytical framework for this discussion is introduced and then applied
to this change effort. The barriers to changes are identified and discussed.
Chapter five presents the conclusions based upon the background of chapter two and the
analysis of chapters three and four. The recommendations focus on actions that can be
taken by the company to meet their objective of growing their metallurgical products
business.
12
Chapter 2: Background
This chapter will provide the necessary background information for the subsequent
analysis and recommendations. First, a description of the product, tantalum, and its uses
is outlined. Next, a high-level view of the company, H. C. Starck, is provided. Third, a
detailed description is given of the operation being studied, the melt shop. Finally, the
motivation for this study and the problem statement are detailed.
2.1 Tantalum and Its Uses
Tantalum, chemical symbol Ta, is an element with atomic number 73. It is in the sixth
period, with elements like tungsten and rhenium, and fifth family, with elements like
vanadium and niobium. Tantalum exists as a metal. It is dense, specific gravity of 16.6.
Its melting point is very high, 2,9960C, and is exceeded only by tungsten and rhenium.
Tantalum is also highly corrosion resistant and ductile.
Tantalum is used by a wide variety of industries with a total worldwide annual
consumption of about 550 tons. The computer and electronics industry uses tantalum
capacitors on circuit boards. These electrolytic capacitors (and the vacuum furnace parts
used to make them) account for about 60% of tantalum's usage. In addition, the
semiconductor industry uses tantalum for sputtering targets used to process their
integrated circuits. In alloys where properties like high melting point, high strength, and
good ductility are important tantalum is a good choice. Examples of these uses include
chemical process equipment, nuclear reactors, and aircraft and missile parts. Also,
because tantalum is not reactive, it is being used in medicine for surgical appliances.
Finally, tantalum oxide is used to make glass with a high index of refraction for camera
lenses.
Tantalum is not readily available. Ores are located in Australia, Brazil, Mozambique,
Thailand, Portugal, Nigeria, Zaire and Canada. The refinement process from ore to pure
13
tantalum is complicated and requires several steps. As a result, tantalum is expensive,
costing $1 00/lb for scrap. Compare this to several other metals. Aluminum costs
$0.70/lb. Silver costs $64/lb. Gold costs $3,841/lb or $263/troy-oz. (WSJ 29-Mar-2001)
2.2 The Company
H.C. Starck Inc., the company, is a producer of tantalum and niobium products. It
employs approximately 400 people at a 325,000 sq-ft complex. The company is part of
H.C. Starck GmbH of Germany, an international company of 2,500 specializing in metals
with high melting points such as tungsten, molybdenum, rhenium, tantalum and niobium
as well as non-ferrous metals like cobalt and nickel. H.C. Starck GmbH is a member of
the Bayer AG group, an international chemical and healthcare company that employs
145,100 people worldwide.
There are six different manufacturing operations within the company: powder, sinter,
wire, melting, flat rolling and fabricating. These operations support four types of
products: powder, wire, mill products and fabricated products. These products are made
from one of three incoming materials. The first, K2TaF 7, is an intermediate in the
refinement of tantalum and represents the largest source of tantalum for the company. It
is received from one of the other H.C. Starck GmbH companies located in Germany,
Thailand or Japan. The second raw material is tantalum ingots. These are usually
purchased at auction from the former Soviet Republics or the US government. The last
type of raw material is scrap tantalum and there are two sources. Some is purchased from
other companies, when they replace their vacuum furnaces, for example. The other
source is the internal scrap flows of the company. See Figure 2 on page 16 for a
schematic of the operations and the flows of material within H. C. Starck Inc.
Figure 1 below shows the revenue contribution of each business to H. C. Starck Inc.
Powder is the largest at 42%. The second largest revenue source, specialty products, is
made up of items that are imported from other groups within H. C. Starck for sale in the
Unites States. The third largest business is metallurgical products at 19% (when milled
14
and fabricated parts are included). The fourth most important business by sales revenue
is wire products at 12%.
Revenue Contribution by Business
Mill & Fabricated
2/
other
4%/6
Figure 1: Revenue Contribution by Business
2.3 Melt Shop
The melt shop and its tantalum processing operations are the focus of this study. This
section will characterize the melt shop by describing its inputs, its products and the
processes used to make these products.
2.3.1 Inputs
The melt shop is a recycling operation. Figure 2 shows the flow of tantalum into and
within the company and illustrates how the melt shop relates to the other operational
units. There are three items to of particular interest. First, the melt shop receives scrap
tantalum from every operation within H. C. Starck. Second, scrap is also purchased from
outside the company. The split between inside recycling and outside purchases is
15
approximately 50/50. Finally, and most importantly, the melt shop is the only supplier of
inputs to metallurgical products, the mill shop and the fabrication shop.
H.C. Starck, Inc.
Powder
K 2TaF7
Powder
Production
Sinter
Plant
Wre)
Mill
Wire
Ta Scrap
Metallurgical Products
Melt
Shop
Flat
Mill
Mill Products
Fabrication
Shop
Ta Ingot
Fabricated Parts
___________................
)0
........
Material Flows
Scrap Flows
Figure 2: H. C. Starck Inc. and Its Flow of Tantalum, Carroll (2000)
The scrap inputs come in a wide range of forms: powder, machine chips, wire, plate,
sheet, foil, anodes and ingots. While the differing shapes of the input require a variety of
processes, it is the issue of contamination that is most contentious.
2.3.2 Products
There are a limited number of products produced in the Melt Shop: a total of five alloys.
They are denoted E, D,
Q, S
and W. These alloys are transferred to the other departments
in the form of ingot, sheetbar or rod. The total number of products produced is eight
given the shapes and alloys of the materials and is shown below in Table 1. These
16
products are used to make goods for the Ta-Metallurgical, Mill Products and Fabricated
Products businesses. These products accounted for 19% of 2000 revenues for the
company. See Figure 1 on page 15.
Table 1: Products of the Melt Shop
Alloy
E
D
Ingot
Q
X
Sheetbar
X
X
Rod
S
X
X
W
x
x
X
2.3.3 Current Processes
The processes used by the melt shop transform scrap tantalum into sheetbars. The
structure of the process is described best as a flow shop. There are a limited number of
products with low volumes (compared to steel or aluminum manufacturing operations).
There are twelve primary workstations that are separated by buffers. The cycle times for
the processes are measured in hours. It is common for some process steps to take more
than one 8-hour shift.
17
2.3.3.1 Typical Process Steps
Figure 3 shows the typical process steps in the melt shop. Scrap lots of tantalum arrive to
the melt shop and are stored in drums. Different scrap lots are mixed together to make a
blend of 2,250 pounds. Next, the blend is pressed into 50-pound bars to facilitate
material handling. The bars are melted to form ingots. These ingots are cut in two so
they be machined and are to easier handle. If the tantalum is being alloyed, half-ingots
are used. Before being sent to mill shop, all half-ingots are pressed into sheetbars.
Scrap
Tantalum
Units of Analysis:
Blend
Press Bars
Ingot
Ingot
Sheetbar
2,250 lbs
45 bars
50 lbs each
2,000 lbs
-900 lbs
-750 lbs
2 jobs
2 jobs
2 jobs
I job
1job
Figure 3: Melt Shop Process Steps
2.3.3.2 Routings
There are twelve primary machines in the melt shop and three basic routes that use these
machines. The final alloy and the type of raw material determine the route. The routes
will be referred to as I, II and III. Below are detailed descriptions of each route that
material can take from station to station in the melt shop.
18
Route I: Alloy E & D
2.3.3.2.1
Route I is unique because it is the only route that uses the electron beam furnace (7). The
process to make alloy E and D is the same except additional steps are needed for alloy D.
Both versions of route I will are discussed below and are denoted by Route Ia and Route
Ib.
Route Ia takes the following path through the melt shop: 1-2-3-(4 or 5)-7-8-9-10-Out. A
schematic of the route is given below in Figure 4.
Out
A
Mill
(1)
Blend
(2)
4
ress
(3)
ABAR
Sinter
(4
AC/GCA
EB Melt
(7)
Qualiy
(8)
Machine
(9)
Forge
(10)
W01k
Sinter~
5)
Figure 4: Route Ia
The first operation is the mill (station 1). As new lots of material are received they are
milled to break the scrap tantalum down into powder. After milling, the material is set
aside until it is needed.
Next, various lots of previously milled powder are mixed together in the blender (station
2) to form a blend. The lots are chosen to yield an expected chemical composition that
meets the specifications for the blend.
After the lots have been blended, they are pressed into bars (station 3). There are
normally 45 bars weighing 50 pounds each.
19
Next, the bars are sintered (stations 4 or 5). There are four sintering ovens. The type of
material determines which of the sintering ovens can be used. AA can be processed only
in the ABAR furnace (station 4). AA blends make up 20% of the jobs processed. The
ABAR furnace has a slower processing rate than the other three ovens and is also less
reliable. MG and HY bars can be sintered in the other three ovens (station5).
After sintering, the bars arrive to the queue for the electron beam (EB) furnace (station 7)
where they will be melted into ingots. Each lot of bars must be melted twice in the EB
furnace to make an ingot. The first melt is used to consolidate the bars into an ingot and
remove impurities in the metal. The second melt is used to further refine the ingot and
improve its surface quality. The first and second melts have very different processing
times, 20 hours for first melts and 5 hours for second melts. The changeover times
between the melts are long, 4 hours going from first to second and 8 hours going from
second back to first. As a way to manage this process more efficiently, the company
batches the jobs. A batch of jobs typically has 4 ingots.
After being melted in the EB furnace, the ingots are tested by quality control (station 8).
This process takes seven days on average. 25% of the ingots do not pass testing on the
first attempt. Ingots that fail testing are not reprocessed at first. They are retested.
Retesting does not generally result in a different outcome. In most cases, the ingot that
failed the test is downgraded and is used for lower purity ingots, i.e. alloys S and W. The
low first-time pass rate is due primarily to the variability in quality of the incoming raw
material. Remember, the melt shop recycles material and receives many different types
of raw material from a variety of sources. The ingots that pass testing are considered
"pure" tantalum at this point. The name for this alloy is "E."
Following quality control, the ingots are machined on a lathe (station 9) to remove the
outer layer and the associated impurities that are concentrated there. Prior to being
machined, the ingots are cut into two pieces so that they will fit onto the lathes.
20
After being machined, the smooth ingots are pressed (station 10) from cylindrical ingots
to 4" thick sheetbars and sent to the mill shop.
Route Ib: Alloy D
2.3.3.2.2
Route Ib takes is the same as Route la plus some additional steps. It follows the path:
1-2-3-(4 or 5)-7-8-9-10-11-12-9-10-Out.
A schematic of this route is given below in
Figure 5.
In
Out
ABAR
Sinter
-.-
Mill
(1)
Blend
(2)
Press
(3)
-
(4
AC/GCA
EB Melt
(7)
Quality
(8)
Machine
(9)
W~ed
(11)
Forge
(10)
Mrc
-(2
-Sinter
(5
Repeat
Figure 5: Route lb
Route Ib is the same as Route Ia up to station 10. If the ingot of alloy E at the Forge
(station 10) is to be alloyed to make D, it is pressed into a round bar and sent to the
welding operation (station 11). At the welding operation (station 11) leaches of alloying
metal are attached to the side of the bar of alloy E. This bar is then sent to the arc
furnaces (station 12) where it is melted to create alloy D. After melting at the arc
furnaces, the ingots are sent back to the lathes (station 9) where they are machined again
to remove impurities and improve the surface finish of the ingot. Finally, the machined
ingots of alloy D are pressed (station 10) into sheetbars and sent to the mill shop (Out)
for further processing.
21
2.3.3.2.3
Route II: Alloy Q, Alloy S & W (Light Scrap)
Route II takes the following path through the melt shop: 3-11-12-9-10-11-12-Out. Figure
6 shows a schematic of this route.
AHAR
Press
(3)
E.M061
1o
ACGA
()
uai
()
Machine
Forge
(9)
(1)
Weld
M
()
Figure 6: Route UI
Alloy
Q in
the form of a powder and Alloy S & W in the form of light scrap is received
from other departments within H. C. Starck. It is pressed into bars (station 3). The
pressed bars are welded together, end-to-end, to form three rods of equal length (station
11). These rods are melted together to form an ingot in the arc furnaces (station 12).
This ingot is taken to the lathes (station 9) where it is machined to remove impurities and
improve the surface finish. The ingot is then pressed to a smaller diameter rod (station
10). After having a hanger installed on its top (station 11) it is re-melted in the arc
furnaces (station 12) before being sent to another department within H. C. Starck (Out)
for further processing.
2.3.3.2.4
Route III: Alloys S and W (Heavy Scrap)
Route III takes the following path through the melt shop: 11-12-9-10-11-12-9-10-Out. A
schematic of this route is given below in Figure 7.
22
7Out
Fr
SinterArc
(7)
AC/GCA
(8)
Meit
(9)
(10)
(1)
(2
Repa
Figure 7: Route IM
Alloy S and W arrive in the form of thick pieces of scrap. This scrap metal is welded
together (station 11) to form "rods" and a hanger is attached. These rods are melted
together to form an ingot in the arc furnaces (station 12). This ingot is taken to the lathes
(station 9) where it is machined to remove impurities and improve the surface finish. The
ingot is then pressed to a smaller diameter rod (station 10). After having a hanger
installed on its top (station 11) it is re-melted in the arc furnaces (station 12) before being
sent back to the lathes (station 9) for machining. After machining, the ingot is forged
(station 10) into a sheetbar and sent to the mill shop (Out) for further processing.
2.4 Evidence of a problem
The melting operation does not have enough capacity to support the demands being
placed on it. There are three indications of this problem and they will be presented in this
section. First, the on-time performance for metallurgical products is below the company
target and is attributed to a lack of material from the melt shop. Second, there is under
production in the melt shop versus current demand. Finally, and most importantly,
demand for metallurgical products is expected to grow over the next three to five years
and this will place additional demands on the melting operation. The majority of this
growth is expected to be for alloy E while the other alloys stay relatively constant.
23
2.4.1 On-Time Performance
The company has a goal of 90% on-time performance. During 2000 the on-time
performance of metallurgical products dragged the company average below the goal.
While the other business units had performance above the 90% goal, the metallurgical
products business is unable to meet the goal and had several months below 70% on time
performance.
On time shipping performance is a customer service metric. At H. C. Starck Inc., this
measure is reported monthly. It is calculated by identifying the number of orders that are
sent to customers on or before the scheduled shipping date and dividing by the total
number of orders scheduled to ship in the month.
An analysis showed that the primary reason for poor on-time performance is attributed to
the lack of material from the melt shop. The Pareto chart showing this result is given
below in Figure 8.
Reasons for Late Shipments - Mill Products
May &June 2000
Total=1 29
10
125
S1007%
E
C,
75
0
50
25
0%
C-)
0.
A.C
c)
5
0
.0
-h
9
a.
0
C
U)
0)
U)
(
U)
0
C')
Figure 8: Pareto Chart of the Reasons for Late Shipments
24
2.4.2 Underproduction for Current Demand
Delivery of products from the melt shop to the mill shop has not kept up with demand.
The production planner of the mill shop submits requirements on a monthly basis. The
amount request varies little from one month to the next. Ten months out of the eleven
observed, the melt shop did not meet the ordered demand. The monthly demand and the
deliveries are shown in Figure 9.
Orders are stable.
Deliveries are not meeting the demand.
1.5
~0
1.0
0U
0.5
S---Ordered
-u--Delivered
0.0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug* Sep
Oct
Nov
Dec
Figure 9: Orders and Deliveries, Melt Shop
*
Data for August is not shown due to a plant-wide shutdown
The primary reason attributed to the poor delivery performance is low availability of the
electron beam furnace (EB furnace). 76% of the total production of the melt shop passes
through the EB furnace. The average uptime is 68% for the last half of the year. The low
was 38% in June. The high was 75% in December.
The EB furnace is run on a 24/7 basis or 722 hours per month. 68% uptime gives 491
hours of available time. To meet demand, the EB furnace needs 448 hours for production
(first and second melts for 16 ingots and 4 sets of changeovers). The average margin
between the available time and the time needed is 43 hours. In months where the
25
availability is lower than average, the EB furnace can easily get behind and cause the
poor delivery performance to the other departments that depend on its output.
2.4.3 Projected Growth in Demand
The final and most important indication is that demand for the products made by the
metallurgical business unit is projected to grow over the next three to five years.
Historically, this business unit has sold to the chemical processing industry. The markets
in which it participates are changing and will affect the melt shop in two ways.
First, there will be an increase in the absolute level of demand. Sales of consumer
electronics, such as computers, personal digital assistants (e.g. Palm Pilots) and cellular
telephones, have been growing rapidly. The technology trend for these products is
toward miniaturization, smaller devices with higher performance. Tantalum capacitors
perform better than capacitors made of other materials, e.g. ceramics. This affects
metallurgical products because they provide the tantalum parts needed for the vacuum
furnaces used to make tantalum capacitors. The historical shipments and forecasted
demand for mill products is shown below in Figure 10. The forecasted demand is taken
from the sales and marketing group at the company and is assumed to be representative
of the expected growth.
26
Shipments to the mill shop have been growing
and are forecasted to continue.
54
3
CgDre0
F
CDe
O
WO
D W
CD
W
f
CD
-
Met
W
CW
rica
CD
W
-O.
CD
(A
Pr dcW
M
W
-1
CD
W
W
C 0C
C5 03
31
W C
D
CD
co
s
Figure 10: Demand for Metallurgical Products, Historical and Forecasted
Second, the product mix for the metallurgical products business unit will change. There
is projected to be a five-fold increase in the use of alloy E. This increase is due to
expected sales in the computer and semiconductor industries for new products like
sputtering targets. In addition, the purity of alloy E will need to be higher than the
current process consistently can deliver.
The current and projected demand for each of the five alloys is shown in Figure 11.
Notice that there is a significant increase in demand for alloy E. The amount of
demanded of the other alloys is not expected to change. This increase in demand for
alloy E will require that more tantalum be process than today and that they be of
consistently high quality.
27
Demand for alloy E grows 5-fold by 2005
while the other alloys remain constant
m Current
*2005
E
D
Q
S
W
Figure 11: Projected growth in demand for alloy E
2.5 Response
Participating in these new markets will require additional melting capacity and processes
that yield higher purity ingots. To accomplish this, the company has been active in
expanding via acquisition. These new facilities will not be available for approximately
two years, however. Given the forecasted growth in the tantalum market and the capacity
constraints that now limit the company's production, their participation will be limited.
This may result in lost sales and a loss of market share.
In order to avoid lost sales and a loss of market share, the company needs to develop a
short to medium term response that will enable them to participate in the growing
demand for alloy E until the capacity provided by the acquisition becomes available.
Doing so will require using the assets currently available in a way that is more effective
and investing in selected projects in order to increase the capacity of the melting
operation.
28
Chapter 3: Analysis
This chapter will focus on the changes that can be made to the melt shop in an effort to
increase its capacity within the next year. As indicated in the previous chapter, demand
for Alloy E is expected to increase five-fold. For this reason, the analysis and subsequent
proposals will focus on increasing the capacity associated with making this alloy.
This chapter will address the analysis in the following way. First, there will be a
discussion of the methods for increasing the capacity of an existing operation. Second,
the application of a queueing network model will be introduced. Third, the operations of
the melt shop will be described in terms of a queueing model. This model will be used to
study the options available to the melt shop. Different scenarios for operating the melt
shop will be described and expressed in terms of the queueing model. Finally, the results
of the models output will be discussed.
3.1 Increasing Capacity
This section will identify strategies to increase the capacity of an existing operation.
Each one will be discussed and related to the melt shop operation. There is a significant
body of work related to increasing capacity of an existing operation. The most famous is
The Goal, Goldratt (1984). In it he outlines the Theory of Constraints. Graham (1998)
combines his work with that of others in the field. Below are some of the steps that can
be taken to increase the capacity or throughput of an operation. Each of these will be
discussed below in relation to the melt shop.
"
"
"
=
*
=
-
Identify the system's capacity constrained operation (bottleneck)
Increase the bottleneck's availability
Offload work from the capacity constrained operation to other stations
Decrease variability
Schedule production based on the capacity constrained resource (Drum)
Buffer the bottleneck operation against starvation (Buffer)
Tie the release of new jobs to the status of the constraint's buffer (Rope)
29
3.1.1 Identify the "Bottleneck"
The bottleneck operation is also known as the capacity constrained resource. It is defined
as the station with the highest utilization. Utilization, u, is defined as the ratio of arrival
rate to the processing rate, ra/ r,. The arrival rate, ra, is the number of jobs arriving to a
workstation per unit time. The processing rate, r,, is the number of jobs completed by the
workstation per unit time. Alternatively, it is the inverse of the time needed to process a
job at the workstation, r, = 1/ta. The operation with the highest utilization is often the
one with the longest queue. Cycle time increases rapidly as the utilization of stations
approach 100%. This fact is captured in the relation below which is plotted in Figure 12
(Hopp & Spearman, 2000).
CT oc
(1- u)
Cycle time increases rapidly as utilization
approaches 100%
120__
_
-
100-
e
80-
E
604020-
50%
60%
70%
80%
90%
% Utilization
Figure 12: Cycle Time versus Utilization
30
100%
In the case of the melt shop, the EB furnace is the bottleneck. Its utilization is the highest
among any station in the melt shop as indicated by several observations. First, there is
often large queue of material in front of it. Second, it is run 24 hours a day, seven days a
week in order to meet the production demands. It is the only operation, with the
exception of the ABAR furnaces, that did so. Third, calculations made in Appendix B
confirm that its utilization is above those of the other station in the melt shop. Finally,
most of the people involved in running the melt shop operation recognized the EB
furnace as the constraint they needed to manage most carefully.
3.1.2 Increase Availability
One of the first steps in increasing the capacity of an operation is to increase the
constraint's availability. Availability is the percent of time that a station is able to process
jobs. The availability can be defined as the ratio of mean-time-to-fail (MTTF) to the total
available time. Total time available is the MTTF plus the mean-time-to-repair (MTTR).
This relationship can be stated as follows (Hopp & Spearman, 2000).
Availability =
MTTF
MTTF + MTTR
There are two ways to increase a station's availability. First, decrease the MTTR. To do
this one must focus on diagnosing and fixing problems with the machine as a quickly as
possible. Second, increase the MTTF. Doing this requires making the station's process
more robust and less likely to fail.
The management and operators of the melt shop suffered a big blow to the availability of
the EB furnace in the year 2000. They made a change to the electron beam guns and its
associated control systems. This change had a major negative effect on the EB furnaces
availability. The MTTR increasedbecause the maintenance people and operators are not
skilled in diagnosing and repairing failures of the new electron beam guns. The MTTF
31
decreased because the new gun system, as installed, is not robust and would fail
frequently. The result is a decrease in this key machine's availability.
The management and operators of the melt shop together with the maintenance group and
electron beam gun vendor are addressing the issue of availability on the EB furnace. The
company uses a measure of availability they call "uptime." It is the number of hours in a
day that a machine is able to produce (but not necessarily producing) divided by the total
number of hours in a day. They have been successful in increasing the average uptime
from a low of 38% in June to 75% in December. This is accomplished by increasing the
MTTF through improved gun maintenance and design changes to the EB guns. MTTR
has been decreased by having the vendor develop and teach repair procedures and by
having a better system for ensuring spare parts availability. Further increases to the
availability will enable increases in the capacity of the operation.
3.1.3 Offload Work
Offloading work from the bottleneck operation means giving work to other machines that
are capable of performing the same task but have lower levels of utilization.
At the outset, there appears to be no other piece of equipment that can do the work of the
EB furnace. It is a highly specialized machine that consolidates sintered bars into ingots
and purifies the tantalum scrap of contaminates. The EB furnace does this in two steps.
The first melt, which takes 18 hours, is used to consolidate the blends from bars into
ingots and remove some contaminates. The second melt, which takes 6 hours, is used to
further purify the ingot and improve the surface quality of the ingot.
There is, however, another machine, the plasma furnace, that might be used to do some of
the work of the EB furnace. There are three reasons why this is appealing. First, it is
especially efficient at removing oxygen, one of the contaminants that cause the first-melt
in the EB furnace to be lengthy. Second, the plasma furnace is not being used. Finally,
32
the plasma furnace is capable of taking un-sintered bars and melting them into an ingot.
If the plasma furnace could be utilized to eliminate sintering completely and perform the
first melts now being done by the EB furnace, this would represent an increase in
capacity. The queueing model will explore the implications of this process change in
later sections.
The plasma furnace is purchased eleven years ago and has not been used very much
since. Several years ago it was configured to produce copper-niobium, an alloy needed
for the super-conducting super-collider project to study sub-atomic particles. That
project did not receive the needed funding and this business evaporated. Since that time,
management has invested to make the changes needed to melt tantalum. There has,
however, been little urgency associated with making the plasma furnace operational and
integrating it into the tantalum refining processes.
3.1.4 Decrease Variability
Increases in variability, like utilization, lead to increases in cycle time. The effect is not
as dramatic as utilization but is still important. The cycle time, CT, is an increasing
function of the variability of the arrivals to a station and the variability of the station's
process. A convenient way of characterizing variability is to use the squared coefficient
of variation (scv), c 2 . The scv is defined as the variance divided by the mean squared:
c2
2
.
In the case of arrival processes, c2 would be composed of the variance
and squared mean of the inter arrival times. For workstation processes, c2 is the
determined by the variance and the squared mean of the time needed to complete a job.
Using these, the effect on CT can be expressed using the following relationship (Hopp &
Spearman, 2000).
CToc (
+cP
2
33
In the context of the melt shop, the EB furnace represents a highly variable process. It
processes work in batches of four ingots. As a result, the next downstream operation,
quality control, receives four ingots (8 jobs) over two days. This is followed by a span of
five days during which the EB furnaces processes its next batch of four ingots. This has
two implications. First, this leads to long queues at the EB furnace given that it is also
the bottleneck operation. Second, the quality control group may be over-utilized during
the days it receives the four ingots and under utilized at other times. This adds time to
the total process and increases the number of jobs that are waiting to be processed by
quality control. Both of these contribute to lengthening the total manufacturing cycle
time and decreasing the system throughput.
Smoothing this part of the production process would help in decreasing the
manufacturing cycle time and increasing the throughput of the melt shop. Reducing
variability of the inter-arrival times is addressed in two ways: decreasing the processing
times and decreasing the change over times. The ideal would be a batch size of one that
requires only one short processing step.
Given that the refinement process currently takes two melts to complete, it seems
unlikely that the melt shop would be able to eliminate one of those melts without
somehow adding another furnace. Remember, however, there is another furnace that is
not being used, the plasma furnace. If a process could be developed to do all first melts
on the plasma furnace and all second melts on the EB furnace, the goal of doing one melt
with no change over could be achieved.
3.1.5 Drum-Buffer-Rope
The drum-buffer-rope method of managing a manufacturing operation, described by
Goldratt (1984), has elements that are applicable to the melt shop. Due to the focus on
increasing the melt shop's capacity, the concept of buffering the capacity constrained
resource against starvation is most applicable. If the bottleneck operation is starved for
34
work, it is a lost opportunity. No more production will flow from a process than what is
produced by the bottleneck operation.
The managers of the melt shop seemed to grasp the importance of buffering the
constraint. There is always a queue of material in front of the EB furnace. This ensured
that when it is operational, there is work available.
The concepts of "drum," scheduling work based on the constraint and "rope," releasing
new jobs to a route based on the status of the bottleneck's buffer, might be helpful to
consider but are not discussed here further.
3.2 Queueing Network Model
As a way to explore changes to the melt shop processes and their effect on the throughput
of the operation, a queueing network model of the melt shop is developed. Doing so
enabled a comparison of several possible process scenarios for the melt shop. This
section will provide the information needed to understand the development of the model
by describing its background, assumptions, inputs and outputs.
3.2.1 Model Background
This model provides a steady state approximation of the throughput and flow time that is
accurate enough to evaluate the changes proposed here. The approach used is
summarized in the paper of Suri, et. al. (1993). Others, whose work is referenced,
include Sur & DeTreville (1991), Whitt (1993), Buzacott & Shanthikumar (1993) and
Hopp & Spearman (2001).
The model uses node decomposition of an open queueing network made up of GI/G/m
queues. This network consists of a twelve separate processing stations and a single class
of material. Arrivals to the network are characterized by the mean and variance of the
inter-arrival times of customers, which are assumed to be generally distributed. The
35
station service times are also assumed to have general distributions characterized by the
mean and variance. Jobs travel from station to station based on transition probabilities.
These probabilities form a set of traffic equations that link the GI/G/m queueing
processes together. Jobs are serviced on a first-come-first-served basis. Service times at
each station are i.i.d (assumed to be independent of one another).
3.2.2 Model Assumptions
The melt shop does not satisfy two of the assumptions identified above. First, there is not
one class of material in the melt shop. There are a variety of raw materials that are used
to make five different alloys. Second, jobs are not always completed on a first-comefirst-served basis. Queues are often rearranged so that similar jobs can be processed
together.
The model itself is not changed to take these violations into account but they are
addressed. In the first case, the input data are calculated in a way that captures the effect
of the differing materials in the mean and variance of the station's processing times. In
addition, only two of the three alloys pass through the EB furnace, alloys E and D.
Demand for alloy D is assumed to be constant over the time period considered.
Regarding the second violation, the queueing discipline of first-come-first-served is
assumed to be "good enough" given the added complexity required of the model for a
marginal improvement in accuracy.
Interested readers are referred to the Appendix A for a more detailed discussion of the
queueing network model.
3.2.3 Model Inputs
Using this model, each station j can be characterized by the 5-tuple, { c
,
,r
,
m }.
cO is the squared coefficient of variation (scv) of the inter-arrival times to the network
36
from the outside. C ,1 is the scv of the processing time of station j. roj is the inter-arrival
rate of jobs to the network from outside. r,, is the processing rate of station j. Finally,
jis the number of servers/machines at station J.
In addition to the characteristics of each station, j, the routings between stations need to
be identified. This is done using transition probabilities. For example, if all jobs at
station one go to station two, the probability would be 1.0. Conversely, if no jobs go
from station one to two, the probability would be 0.0. See Appendix B for a detailed
description of the model inputs used here.
3.2.4 Model Outputs
The output of the model are the inter-arrival rates and the scv of the inter-arrival times to
each station, j, re, and
C
respectively. These are used to calculate the system's
performance characteristics, throughput, TH, cycle time, CT, and work-in-process, WIP.
Each of the metrics is discussed below.
The throughput of the system, TH, is found by adding together the arrival rates of the
three raw material stream. Because of conservation of material, what goes in must equal
what comes out for process in steady state. The arrival rates for the current scenario are
calculated based on available data and are shown in Appendix B. 1. For the other
scenarios, the arrival rate of the raw material for alloy E is adjusted to the point where the
workstation with the highest utilization reaches 85%. This level of utilization was chosen
because it is the utilization of the EB furnace, the bottleneck operation, in the current
scenario.
37
The manufacturing cycle time, CT, is found in two steps. In step one, the time in queue,
CT,,, is calculated using the Kingman equation. (Hopp & Spearman, 2000)
/C2
CT
+
aj
2
2
Pi
i
11- UiI rj
This equation is also known as the VUT equation because the time in queue at a station is
dependent on the variability (V), utilization (U) and average processing time (T) of the
station. The variability term is calculated using the squared coefficient of variation (scv)
of the inter-arrival times, c 2j, and the processing time at the station, ch . The utilization
term is uj = raj / rj. The average processing time is the inverse of the processing rate or
1/rj. In step two, the total manufacturing cycle time, CT, is the sum of all of the queue
times, CTqj, and all of the processing times at each station, 1/rj.
Finally, the total number of jobs in the system, WIP, is calculated using Little's Law.
WIP = (TH)(CT)
3.3 Scenarios
Three scenarios are built using a queueing network model. These are called "current,"
"50% to plasma." and "100% to plasma."
-
The "current" scenario attempts to model the performance of the operation as it
exists presently, i.e. it does not include the plasma furnace in its processes.
* The "50% to plasma" scenario modifies the "current" scenario by eliminating the
use of the ABAR sintering furnace and routing of the pressed bars through the
plasma furnace. The other 1/2 are sent through the GCA/AC sintering furnaces.
" The "100% to plasma" scenario modifies the "current" scenario further to
eliminate sintering completely and routes all un-sintered pressed bars through the
plasma furnace. The sections that follow will describe in greater detail the
assumptions, inputs and outputs of the scenarios identified above.
38
3.3.1 Scenario Assumptions
In addition to the assumptions articulated in Section 3.2.2, Model Assumptions, this
section will identify several additional assumptions that are specific to this analysis.
These assumptions are made to enable a comparison of the scenarios that is direct and
focused on the suggested changes. First, the model is a steady state model. Second, there
are three raw material inputs. They are assumed to have different routes through the melt
shop that ultimately lead to different products. Thirdly, there is one class of material and
the transition probabilities between stations is kept the same between scenarios unless the
stations are directly involved in the change from one scenario to another. The routing
matrix that is formed by these probabilities is designed to capture all three of the raw
material inputs and their associated routes through the melt shop. Fourth, the
development of the scenarios focus on the route used to make alloy E. It is the product
with the largest projected growth and the alloy for which the melt shop needs to increase
capacity. Fifth, the processing rate of the quality control department is adjusted in each
scenario to maintain a total time (queue and process) for quality to be at least 4 days.
This is done because quality control is a pipeline delay. The variability is assumed to be
approximately half as much as EB furnace. Finally, the arrival rates govern the
throughput of the system. What goes into the process will come out, given that this is a
steady state model. In the current scenario, the EB furnace has a utilization of 85%. The
arrival rate of the alloy E raw material is adjusted in the 50% and 100% Plasma Scenarios
so that no major piece of equipment exceeds 85% utilization.
3.3.2 Scenario Inputs
The inputs to the queueing network model for each scenario will be detailed here. The
three scenarios are called "Current Process," "50% to Plasma" and "100% to Plasma".
Remember from Section 3.2.3 each station j can be characterized by the 5-tuple,
C2
{c
2
,,, r 0 , r,, 1, m 1 }. Table 2, shown below, displays these characteristics of the twelve
stations for each of the three scenarios.
39
There are several items of particular interest. In the current process, the plasma furnace
is not used at all. Second, "50% to plasma" scenario adds the use of the plasma furnace
and eliminates the use of the ABAR furnace for sintering. 50% of the jobs from raw
material type I are assumed to be sent to the plasma furnace while the other 50% are
assumed to be sent to the GCA/AC sintering furnaces before going to the EB furnace for
melting. Finally, the "100% to Plasma" scenario assumes that the sintering furnaces are
not used at all. Instead, all jobs of raw material type I are sent to the plasma furnace for
first melts before going to the EB furnace for second melts. Note the increase in
processing capacity of the EB furnace that results from only doing one melt.
Table 2: Workstation Inputs for All Three Scenarios
Current Process
rp
C2
ci
50% to Plasma
100% to Plasma
C2
C2
rp
c ,
ri
c ,
j
Station
M.
[jobs/day]
[-]
[jobs/day]
[-]
[jobs/day]
[-]
1
2
3
Mill
Blend
1500T
1
1
1
2.14
2.46
4.00
0.50
0.50
0.50
2.14
2.46
4.00
0.50
0.50
0.50
2.14
2.46
4.00
0.50
0.50
0.50
4
5
6
ABAR
GCA/AC
Plasma
1
3
0.29
0.69
--
7.00
0.50
--
-0.69
2.77
-0.50
0.50
--2.77
--0.50
7
8
9
10
11
12
EBF
QC
Lathe
2000T
Weld
Arc
1.14
1.14*
1.50
3.81
4.00
2.44
2.92
0.50
0.50
0.65
0.50
0.40
1.14
1.15*
1.50
3.81
4.00
2.44
2.92
0.50
0.50
0.65
0.50
0.40
2.77
1.80*
1.50
3.81
4.00
2.44
0.50
0.50
0.50
0.65
0.50
0.40
*
**
40
1
1
1
2,3**
1
1
1
QC behaves like a pipeline delay. The processing rate is adjusted to keep it total station cycle
time (queue + process) at least four days.
2 lathes are used for the Current and 50% Plasma Scenarios. 3 lathes are used in the 100%
Plasma Scenario. Lathes are not considered a major piece of equipment so one was added to
keep them from becoming the capacity constrained resource.
Each of the parameters were evaluated to determine the sensitivity of the output to
changes in these inputs. The number of machines, m, for a station affects the utilization
of any workstation. The system output is only changed if it occurs on the bottleneck
operation or another near constraint. The number of machines is known with certainty
and therefore not subject to errors. Both the arrival rates to the system, roj, and the
processing rates for each station, rpj, play a large role in the performance of the system.
The arrival rates to the system are variables that determine the throughput of the system.
Changes to the processing rates of the bottleneck and near-constraint operations result in
significant changes to the throughput. The model is not sensitive to the squared
2
2
coefficient of variation (scv) for the arrival rates, co , nor the processing rates, cp1 ,
unless the values are very large, i.e. greater than 2. To address these features of the
model, the parameters for processing rates were chosen carefully. When it was possible,
SAP data from the company was used. Data from direct observation was used when
information from SAP was unavailable. The scv of highly variable processes like the EB
furnace and Quality were found using observational data. The values for the scv of the
other processes were assumed to be 1.0. See Appendix A for a discussion of each of the
parameters for each workstation.
The capacity for each work station is shown below in Figure 13. It is derived from the
input data displayed in Table 2.
41
Capacity of each Workcenter
5
4M
3-
0
cc
0.
0
M
00
m Current
m 50% to Rasma
m 100% to Rasma
Figure 13: Workstation Capacities for Each Scenario
The routings of jobs is different for each of the scenarios. The differences are illustrated
in the figures below. Notice two things. First, the "50% to plasma" scenario (Figure 15)
adds the use of the plasma furnace but does not alter the batch management of the EB
furnace. Second, in the case of the "100% to plasma" scenario (Figure 16) the plasma
furnace replaces both the sintering processes and the first melt in the EB furnace. For
interested readers, the routing matrix for each of the scenarios and the calculations used
to determine the probabilities can be found in the Appendix B Section B.2.
42
Figure 14: Current Process
2"nd
Ml
Ist
Sinter
(,)Melt
Mill / Blend / Press
(1) / (2) / (3)
EB Furnace (7)
Figure 15: 50% to Plasma Process
Mill / Blend / Press
Sinter
(1) / (2) / (3)
(5)
Melt
Furnace (7)
SPlEB
Plasma
(6)
2"
Mt
7Melt 7 7
(
Figure 16: 100% to Plasma Process
Mill / Blend / Press(1) / (2) / (3)
I s Melt
Plasma
(6)
2"nd Melt
EB furnace
(7)
43
3.3.3 Scenario Outputs
The output for the three scenarios, based on the queueing network model, is presented in
this section. The results will be summarized in the tables and figures that follow.
Table 3 shows the key metrics of the queueing model output together with the actual
metrics of the current process. There are differences between the current actual and the
current model scenario. The average cycle time (CT) for the model is 3% higher. The
work-in-process (WIP) is 6% higher. Throughput (TH) is 3% higher. The differences
between the model results and the actual figures for the current scenario are an absolute
indication of inaccuracy. They point out the effect of the simplifying assumptions
mentioned earlier. When comparing the model outputs of the different scenarios, the
inaccuracy is assumed to be trivial. The assumptions of the model are the same from one
scenario to the other. Only the parameters needed to represent the different scenarios are
changed.
Table 3: Comparison of Actual & Model for Current Processes
Scenario
Actual, Current
Model, Current
50% to Plasma
100% to Plasma
44
CT
[days]
70
72
51
11
WIP
[jobs]
101
107
75
20
TH
[jobs/day]
1.44
1.49
1.49
1.79
3.3.3.1
Throughput
Figure 17 shows the throughput for each of the three model scenarios together with the
current actual output. In this plot, higher throughput is best. The model output of the
current processes is 1.49 jobs/day. The actual output is 1.44 jobs/day, a difference of 3%.
Adding the plasma furnace and sending 50% of the press bars through it does not cause
an increase in throughput. The lack of change results from the fact that the EB furnace,
the bottleneck operation, does not have an increase in capacity. The same amount of
material goes through it and two melts are still required of each lot of press bars. Going
to a single melt at the EB would improve throughput. This is what happens in the 100%
to Plasma scenario. The plasma furnace does the first melts consolidating the press bars
to ingots and doing some purification. The EB furnace does second melts, refining the
tantalum further. This enables the EB furnace to now do only one melt per lot. No
batching is required. The "100% to Plasma" scenario has a throughput of 1.79 jobs/day,
an increase of 20% over the current scenario.
Throughput Improvement
2.001.79
1.44---""1.-
(0
1.00C.
0
0.50-
0.00
Actual,
Current
Model,
Current
50% to
Plasma
100% to
Plasma
Figure 17: Throughput for Each Scenario
45
3.3.3.2
Cycle Time
Figure 18 shows the total process cycle time for each of the three model scenarios
together with the current actual cycle time. In this plot, lower cycle time is best. The
model output of the current processes is 72 days. The actual time is 70 days, a difference
of 3%. The 50% to plasma scenario replaces the ABAR sintering furnace with the
plasma furnace. Doing this reduces cycle time from 72 days to 51 days or 30%. The
100% to plasma process is even faster at 11 days of cycle time. The improvement over
the current process is 84%. This scenario benefits from the elimination of the ABAR
furnace as well. The queue time at the EB furnace is nearly eliminated. Batching is no
longer required. Also, there is a reduction in variability that translates into shorter queues
at stations downstream from the EB furnace.
Cycle Time Improvement
90
70-
72
6051
E
3011
0Actual,
Current
Model,
Current
50% to
Plasma
100% to
Plasma
Figure 18: Average Cycle Time for Each Scenario
46
3.3.3.3
Work-In-Process Inventory
Figure 19 shows the work in process (WIP) inventory for each of the three model
scenarios together with the current actual WIP inventory. In this plot, lower WIP is best.
The model output of the current processes is 107 jobs. The actual WIP is 101 jobs, a
difference of 6%. The 50% to plasma scenario replaces the ABAR sintering furnace with
the plasma furnace. Doing this reduces WIP from 107 jobs to 75 jobs or 30%. The 100%
to plasma process has an even lower level of WIP at 20 jobs, an improvement of 81 %
over the current process. The cycle time savings mentioned in the previous section
enable the melt shop to operate with a lower level of WIP and still maintain a higher
throughput.
WIP Inventory Improvement
120
107
100--
75
80
0
C
40-
0.
20
20-
0 1
Actual,
Current
Model,
Current
50% to
Flasma
100% to
Flasma
Figure 19: Work In Process Inventory for Each Scenario
47
3.3.3.4
Utilization
Figure 20 shows the utilization of the twelve stations in the melt shop for each of the
three scenarios considered. There are several items of interest in this figure. First, the
bottleneck process can be identified for each scenario. For the "current" and "50%
plasma" scenario it is the EB furnace. The Quality Control operation becomes the
bottleneck in the case of the "100% to plasma" scenario. Additional capacity increases
would needed to be directed at this resource in order to further improve throughput.
Second, the maximum utilization is 85% by design. It is the maximum utilization
allowed for the scenarios considered because it is the maximum utilization found in the
current situation. Note, a third lathe was added for the "100% to plasma" scenario to
keep the utilization below 85%. Third, higher levels of throughput require higher
utilizations. This is illustrated by comparing the "100% to plasma" with either the
"current" or the "50% to plasma" scenario. Finally, because the throughput is identical,
between the "current" and the "50% to plasma" scenarios many of the workstations have
identical utilizations. The only differences in utilization come in the stations directly
affected by the process changes: sintering, melting and quality.
Utilization for Each Melt Shop Station
100%
85% - - - - - - - - - - - - - - - - --
N
- - - - - - - - - - 85%
50%-
0%
.-j
IE
i
< 0
Current
---
50% to Rasma ----
100% to
Rasrma
- - - - Max Utilizatio
Figure 20: Utilization for Each Station in Melt Shop
48
3.4 Discussion of Results
The results of the model indicate that using the plasma furnace to perform all first melts
can increase the production rate while lowering both the cycle time and work-in-process
inventory of the melt shop. This is not surprising. Adding capacity (plasma furnace) to
the EB furnace (the bottleneck operation), eliminating the need to do two melts, increases
the production rate for melting. Increasing the capacity of melting resulted in a 20%
increase in throughput of the whole process. In addition, the process cycle time and work
in process inventory are reduced by 84% and 81% respectively. This is accomplished by
eliminating the sintering process and the batching at the EB furnace.
Processing all of the pressed bars through the plasma furnace is important critical to
achieving these results. Simply using the plasma furnace for some of the materials will
not result in the benefits mentioned above. There are two primary reasons for this. First,
the EB furnace would be required do two melts on the sintered lots that do not go through
the plasma furnace. This would prevent the increase in capacity. Second, continuing to
do both melts on the EB furnace will require batching. These batches lead to long queue
times and extra inventory.
Based on these observations the recommendation would be to implement the "100% to
Plasma" scenario. This change would affect alloys E and D directly and would require
the following changes:
" Do all first melts in the plasma furnace.
" Do all second melt in the EB furnace.
" Eliminate sintering. Do not sinter the pressed bars before melting them.
Making these changes will increase operational effectiveness of the melt shop. These are
discussed below. There are several issues associated with this change. These are process
49
verification and raw material availability. These barriers to implementation will also be
discussed in this section.
3.4.1
Increased Operational Efficiency
While this proposal does not substitute for the new manufacturing facility planned by the
company management, it does provide a short to medium term solution to the issue of
under capacity of the melt shop. It is capacity that limits the growth of the metallurgical
business.
The suggested change will improve the key operational metrics. Throughput will
increase. Manufacturing cycle time will be reduced. Both of these will be improved with
lower levels of work in process inventory. Furthermore, there will be a decrease in the
inter-arrival variability associated with the sintering and EB furnace operations. This will
enable a smoother flow of work to their downstream operations, like quality control.
Quality may also be improved be sharing the task of purification between the plasma and
EB furnaces. The plasma furnace can be used to upgrade the scrap to a more consistent
but intermediate purity by removing oxygen. The EB furnace can focus on second melts.
As a result, it may be possible to optimize its process parameters and furnace
configuration to produce high purity ingots of alloy E on a more consistent basis.
Given the assumption that the market for tantalum is growing and that every additional
pound of output from the melt shop can be sold, sales will be increased for the company
if more tantalum can be refined. The assets used to create those sales are nearly constant.
Investments in process development, the purchase of another lathe and possible more
testing capacity are what will be needed. Taken together, this provides an increase in the
return on assets (ROA). The expected annual increase in gross profit from increased
capacity is $1.1M. The one-time savings from inventory reduction are $1.3M.
50
Annual Increase in Gross Profit
(1.79 jobs/day -1.44 job/day
f.240yeardays
75Mlbs
job
V lb.20
$LM
One-Time Inventory Reduction
(101 jobs - 20 jobs
job
20
lb
=750lbs
$.3M
The above calculation is based on the production rates generated by the queueing
network model and a conservative estimate of the gross margin for each pound of
tantalum produced.
3.4.2 Potential Issues
Two technical issues have been cited as reasons for not implementing a change of this
type. The first is process verification. The second is raw material availability. Each of
these will be addressed in turn.
3.4.2.1
Process Verification
Making the proposed change represents a significant alteration to the current production
process. The issues of feasibility, yield and quality remain unanswered. At this time
there is simply not enough data to make an informed decision.
Currently, there is an engineering study underway to determine the capabilities of the
plasma furnace in processing tantalum. Three observations are possible at this point.
First, the testing does not appear to have been a consistent, focused effort. It has been
underway for nearly a year without reaching a conclusion. Second, early indications
from the testing are that the plasma can be used in the way described. Finally, in the
51
absence of test data and the need to meet today's demand, management has kept the
production process as is.
Given the results described above, it is important to complete process testing as soon as
possible. An informed decision about whether or not to implement this change cannot be
made without the test data. Not having the data results from a choice about how to
allocate limited resources. Testing will not be completed in a timely manner unless the
value of using the plasma as described is seen as important. The potential benefits
warrant serious consideration and a commitment to completing the testing.
3.4.2.2
Raw Material Availability
Raw material availability is a concern shared by many within the company. Tantalum is
a thinly traded commodity. The demand for tantalum grew much faster than supply last
year, leading to a large increase in the prices of tantalum scrap. Historically, a stockpile
of tantalum has been accumulated to protect against these features of the market.
Processing tantalum scrap at a faster rate could be seen as a threat to this approach.
Faster processing might lead to a reduction in the stockpile of tantalum that protects the
company from the vagaries of the market. Doing this would subject them to more
fluctuation in their primary raw material inputs.
Assuming that the melt shop had more capacity, the response to these concerns would be
to choose to operate below the maximum capacity. Doing so would represent a strategic
decision on the part of the company. In the current situation, the decision of how much
to produce is being made for them. Production is limited. If the market demand
increases for metallurgical products, as it did this last year, the company is limited in its
ability to participate.
52
Stated another way, the expanded capacity enables the company to be more competitive.
It will be able to win new customers and buy raw material away from its competitors.
New customers could be won due to the increased capacity that allows the company to
supply their needs. Critical raw materials can be bought using the increased revenues
that result from the expanded capacity, making it more difficult for competitors to find
the inputs they need.
53
Chapter 4: Implementing Change
Recommending solutions is easy. Implementing changes, however, is much more
difficult. Furthermore, a solution that goes unimplemented is no solution at all. Up to
this point, the discussion has focused on developing a technical solution to the problem of
under-capacity in the melt shop. The focus will shift now to look at the work needed to
implement a change such as the one proposed.
This discussion of change will be done in three parts. First, the concept of technical and
adaptive work will be described. Next, this framework will be applied to the melt shop.
Third, the observed barriers to change and adaptive work will be identified and discussed.
4.1 Technical Work versus Adaptive Work
A common model of how change happens is characterized by being top-down, radical,
discontinuous and planned. The thinking is that once the right solution is identified, all
the pieces will fall into place. While this is a common view of change, it is not very
helpful.
Heifetz (1994) proposes a different mental model that is more appropriate. He describes
two kinds of work related to making change: technical work and adaptive work. Given
these two kinds of work, he identifies three types of situations that generally arise. These
are shown in Table 4 below and illustrated in the subsequent sections using the same
example originally used by Heifetz: the doctor-patient relationship.
54
Table 4: Situation Types, Heifetz (1994)
Situation
Problem
definition
Solution &
implementation
Primary
responsibility
for the work
Kind of work
Type I
Clear
Clear
Physician
Technical
Type II
Clear
Requires
learning
Physician &
patient
Technical &
adaptive
Requires
learning
Requires
learning
Patient >
physician
4.1.1 Type I Situation
In the first situation, Type I, the problem definition and the solution are clear and
unambiguous. This leadership situation requires an expert to identify the problem and
propose a solution. The implementation is easy, requiring little change to the way people
behave. This is referred to as technical work. The physician-patient analog is one where
the patient has an infection. The patient comes to the doctor for diagnosis and treatment.
The doctor diagnosis the disease after some testing and gives the patient a prescription.
The patient fills the prescription, takes the medicine and is "cured" with little alteration to
their lifestyle.
4.1.2 Type I Situation
A Type II situation is more complex. In this case, the problem definition is clearly
identified (technical work) but the solution and implementation of that solution require
that work be done on the part of the affected people (adaptive work). In the doctorpatient scenario this might equate to a patient with heart disease. The diagnosis is clear.
The solution is also clear, e.g. eat different foods, get more exercise, stop smoking, etc.
The implementation of the solution, however, requires active participation on the part of
55
the patient (adaptive work). The patient needs to make different life choices in order to
treat the heart disease.
4.1.3 Type III Situation
Situations of Type III are the most complex. All three aspects require learning and
change: problem definition, solution and implementation. Everyone involved must
change his or her current approach in order to effectively deal with this situation.
Returning to the doctor-patient example, this might equate to a patient with terminal
cancer. There may be two possible problem definitions. One problem definition might
be to focus on the cancer. This will lead to actions relating to treatment. Given that
nothing can be done to stop the disease from ending the patient's life, this may not be
appropriate. A second approach might be to help the patient prepare for the end of life.
Following this problem statement recognizes what is occurring and the actions taken are
aimed at preparing the patient and their loved-ones. In this example there are multiple
problem definitions. Each might lead to different solutions. The patient plays the largest
role in deciding which is most appropriate for them. The doctor plays a facilitating role.
No expert opinion will resolve the diagnosis of terminal cancer.
4.1.4 Summary
The situation types described above and summarized in Table 4 provide a framework for
thinking about organizational change and leadership. Type I scenarios require technical
work. Type II situations require both technical work and adaptive work. Type III
situations require adaptive work. There is a continuum represented by these basic
concepts. Knowing where a situation is located along this continuum can be helpful in
formulating an appropriate response.
56
4.2 Applicationof the Framework
The technical/adaptive work framework described previously will be applied now to the
melt shop and the issue of insufficient capacity.
4.2.1 Problem Definition
The problem definition is clear. The melt shop has difficulty meeting current demand for
metallurgical product orders. It will not be able to satisfy the anticipated future demand
for these products without expansion. More melting capacity is required. There is
general agreement on this problem statement.
4.2.2 Solution
The solution to the problem is not clear. There are several options to consider. All of
them are investments of one sort or another. Some of the choices are: buy another
company, build a new plant, use current assets more effectively or a combination of
these.
Given that the management of the company has recently purchased another company and
is planning to move the operation, there is clarity about what they believe to be the best
solution. This solution will take several years to implement. In that time, there may be
significant lost sales due to the present under-capacity. For this reason, a complementary
solution of using the current melt shop more efficiently should be considered.
4.2.3 Implementation
The implementation of any of the solutions mentioned above will require that employees
of the company alter their activities. Consider two different situations.
First, consider the case of the new facility. It will require some people to drive much
further on their commutes or even move their families if they want to keep their jobs.
57
There will be new machines, new processes and new physical surroundings. This change
will no doubt require leadership on the part of management to help employees make the
transition while minimizing the disruption to the business. The workload will be very
heavy during this transition period. These changes are more than a year away, however.
They are not immediate and do not affect employees today.
Consider a second scenario: implementing a major change to the processes in the melt
shop to increase its capacity. The benefits and costs of the change will be more
immediate. If successful, there will be a significant increase in the melt shop's capacity
within a year. Accomplishing this change will require work. Engineering testing needs
to be completed. New SAP recipes need to be written. Money needs to be spent to
upgrade some of the equipment. There is a litany of tasks that would need to be done.
The number of tasks, however, will certainly be less than will be required to move an
entire production operation.
Why the resistance to the second solution? There are two. First, it requires action and
changes in behavior today, not a year from now. Second, and more importantly, it
requires that people change the way they think about the melting operation. For several
decades the melting operation has processed thousands of ingots using the same
technology, the same set of processes and the same physical surroundings. Using the
plasma furnace to replace sintering and the first melts on the EB furnace represents a
departure from the familiar routine. Doing so feels uncertain and can be perceived as a
threat to meeting current production requirements.
4.2.4 Observations
Increasing the capacity in the melt shop is a Type II situation, in the technical/adaptive
work structure of Heifetz. The problem statement is clear: increase the capacity of the
melt shop. The solution and its implementation will require that the company approach
this problem differently. It is not as simple as adding another shift or hiring a new
58
employee. Regardless of the solution chosen, people will need to alter their behaviors in
order for the solution to be implemented successfully.
A Type II situation requires that the leader not only be an expert but also a visionary.
Those affected by the change need to be able to imagine the future desired state that will
result from the change. Remember the patient with heart disease. Why would he or she
give up fried food and their La-Z-Boy if they did not believe that eating more healthfully
and exercising regularly would make their lives better? What is their vision? For a
patient with a life-threatening disease, they might want to see their kids graduate from
college or get married. Similarly, a vision of the future needs to be created by the leader.
This vision helps motivate people to participate in the change process and learn what is
required to make it successful.
4.3 Barriers to Change
The melting operation within H. C. Starck does not appear to have much experience or
interest in "adaptive work." There are significant barriers to change. The evidence to
support this claim will be presented by looking at the company from three different
perspectives. The three lenses are the cultural lens, the political lens and the strategic
design lens. Each lens will be described and examples given from the company.
4.3.1 Cultural Barriers
The cultural lens focuses on the meaning that people assign to their work. (Ancona, et.
al., 1999). From the cultural perspective, people take action based on what a situation
means to them. In this context, change might be prohibited by some of the cultural
beliefs held by members of the company. Here are three observations. First, "the
business is successful." Second, "we are comfortable." Finally, "I don't trust them. We
can do it better ourselves anyway."
59
The business is successful. It is very profitable without the extra effort required by
adaptive work. It participates in niche markets, specialty metals, with few competitors.
These markets have been growing steadily over the last few years. This growth is
expected to continue. The profit margins for their products are generous. There is little
pressure on these margins because demand exceeds supply. While business success is the
desired outcome, organizational changes that may be needed to further improve the
business are difficult in this climate. There is no motivation to change.
People seem comfortable continuing to do things the way they have been done for the
past few years. The culture of the company is stable. It is a 60 year-old company. There
are 400 people on site. Many of them have been there for a decade or more. The
production processes are stable. These processes have been successful at meeting the
demands placed on them until now. Why change? Again, feeling comfortable with
business processes is desirable but it can make organizational change difficult because
there is little motivation to change even though the business would benefit from the
change.
The working relationships between the melting operation and some other key groups are
contentious. These relationships are characterized by a lack of trust. An example of this
occurred during the upgrade of the EB furnace last year. The engineering group and the
melting operation needed to work closely together to design and install the new system.
Communication broke down and the newly installed equipment made the EB furnace run
worse, not better. Production in the melt shop suffered for most of 2000 as a result. The
remedy is for the melting operation to take control of the whole project and fix the
problem. While this breakdown in teamwork is improving, the fact that it exists indicates
that there are issues of people working together to implement change-even change that
is not supposed to affect the way people do their jobs.
60
4.3.2 Political Barriers
The political perspective views an organization as a composite of "stakeholders," both
individuals and groups. Each one contributes resources to the organization and also
depends on the organization's success. Each stakeholder also may have different interests
and goals and bring different amounts and sources of power to bear in her or his
interactions (Ancona, et. al, 1999). In this context, change might be prohibited by the
way stakeholders chose to use their power in the organization. Three examples are given.
First, company management focuses on the powder business. Second, the melt shop is
run on a "command and control" basis with a focus on short-term issues. Finally, the
process engineer in the melt shop does not support using the plasma furnace in the way
suggested.
The company places most of its emphasis on the powdered tantalum business. It
accounts for 42% of the revenue. Metallurgical (Ta-metallurgical and Mill &
Fabricated), on the other hand accounts for a total of 19% of the business. Recently,
however, metallurgical has received much more management attention. There are two
reasons. First, the on-time shipping performance is so bad during 2000 that it dragged
the company-wide metric down. Second, the sales group has identified new markets for
metallurgical products that require more capacity.
61
Mill & Fabricated
2%
Other
F4%
Figure 21: Revenue Contribution by Business
The melt shop is operated by "command and control." Furthermore, the focus seems to
be on "fighting fires" to meet the demands of today. Having this type of focus and
structure does not engage people in thinking about how to do things differently or better.
One example is that there is little effort exerted on how scrap contamination might be
mitigated even though it is one of the biggest sources of variation in the operation.
Another example is that there is no intention to develop a production plan tied to demand
and attempt to build to that plan.
The process engineer in the melt shop is not in favor of making the change to use the
plasma furnace to process all un-sintered bars. His opinion is highly valued by others in
the company. As a result, the urgency to complete the needed installations and testing is
dampened. The plasma furnace is not viewed as a solution to the issue of insufficient
capacity.
62
4.3.3 Structural Barriers
The strategic design perspective focuses on understanding the design of an organization
and how this design helps the organization reach its strategic objectives (Ancona, et. al.,
1999). Here are three examples where the structure of the organization may not be in line
with the larger goals of the company. This mismatch might prohibit needed changes.
First, the incentives of the melting and milling operations are not aligned. Second, there
is not enough information flowing between the two operations. Finally, this internship is
not sponsored by the melting organization.
The flow of material from the melt shop is the life-blood of the milling operation. They
are in the same value stream. From a structural perspective, one would like the
incentives of these groups closely aligned and an easy flow of information between the
groups. The company is grouped by activity, however. Each functional area operates as
a separate organization. There are some cross-functional teams that span these
organizations to solve specific problems. The formal reporting structure is hierarchical.
As a result of the functional structure, there are different managers and organizations
responsible for melting and milling.
The first example is that the two organizations, melting and milling, are evaluated
separately. On one hand, this is effective given that they perform very different
operations. On the other hand, the milling operation is completely dependent on melting.
Poor performance in melting will negatively affect milling. To remedy this it might be
useful to evaluate the two groups as a single entity on metrics related to throughput,
quality and delivery. Tying the performance of the two operations together will make the
role played by the melting operation more visible.
Secondly, there is a lack of communication between the two groups. The melt shop uses
a policy that can best be described as "build-what-you-can." Every month there is a
demand forecast provided by the milling operation to the melting operation. There is no
indication that this is actually used to plan and schedule production. This seemed to be
63
the state of affairs until this past year. The production planner for the milling operation
began taking part in the daily production status ("drumbeat") meetings in the melt shop.
As a result, the information linking the melting and milling operations together now takes
place on an informal basis. This needs to be strengthened.
The supply chain group and not the melting operation sponsored this project. The focus
at the outset is to improve on-time performance of metallurgical products. Analysis of
the issue led to the melting operation and the need to increase its capacity. While the
project is aligned well with the goals of the company at a high level, there are three
functional groups affected by the work. Supply chain is the project sponsor and is
responsible for meeting deliveries to customers but not responsible for melting. Melting
is the department whose operations are most affected by the suggested changes did not
sponsor this project. Finally, milling operations are the largest customer of melting and
most adversely affected by the lack of material from the melt shop.
4.4 Intern's Role
The intern approached the project as if it were a Type I situation. The assumption is that
by pointing out a solution, it would automatically be implemented. This is a mistake.
Obviously, this is unrealistic.
There needed to be greater recognition of the kind of activities that would actually ensure
the adoption and implementation of the solution. There are several activities that would
have led to greater success. Building stronger relationships with the project stakeholders
would have facilitated better communication and mutual understanding. This, in turn,
may have helped the proposed solution gain greater acceptance and led to a better transfer
of knowledge, both from intern-to-company and vice versa. Next, focusing on smaller
wins throughout the internship would have help to build credibility and momentum for
larger recommendations later.
64
Another feature of the internship, similar to the doctor-patient example from Heifetz, is
that of time. The internship is short, 6 months. It is difficult to learn the organization,
identify a problem, devise a solution and implement that solution in this time.
Implementing a solution is particularly difficult because of the adaptive work aspect. It
demands an investment in time and energy that seems beyond the six month of the
internship.
65
Chapter 5: Conclusions
A focused effort designed to bring the plasma furnace on-line as a productive asset is
important to the capacity expansion of the company. Choosing not to do so represents a
wasted opportunity. To date there is no urgency to complete the needed process
verification to use the plasma furnace. This needs to change.
5.1 Technical Conclusions
The central conclusion of this work is that the company should invest the time and effort
to develop a process that utilizes the plasma furnace to increase the capacity of the melt
shop. There are three areas of the operation that are affected by this change. They are
interconnected. Making this change will increase the throughput of the current melt shop
by 20% while simultaneously reducing average cycle time and work-in-process inventory
by 84% and 81 % respectively.
Increasing the throughput of the melt shop 20% will enable the company to sell more
product. The gross profit contribution is estimated to be $1.1 million per year. Having
the increased capacity during the next three years in the current production facility is
important for two reasons. First, demand for metallurgical products exceeds capacity
now and is expected to grow (Figure 22). Given this growth and the three years to
complete the new facility, lost sales of approximately $15.1 million will result (Figure
23). Using the plasma furnace as recommended will enable the company to reduce the
lost sales during this time by $3.0 million to $12.1 million (Figure 24). See Appendix C
for the economic analysis that led to these figures.
There will be an 81 % reduction in work in process inventory. This one time reduction is
estimated to be $1.3 million. This represents a decrease in working capital requirements,
more cash and better asset utilization.
66
Finally, there is estimated to be an 84% reduction in process cycle time from 72 days to
11 days. Having a process that is this short would enable the company to be more
responsive to shifting customer demands and schedule production more accurately.
Growing Demand
for Metallurgical Products
CD 2>
CO
F
WD
W
V6
to
(M
sr
CD
a)
CHD Pjcd
WD
WD
-4
O0
CO
W
to
0D
0D
CD
0o
0
0D
K)N
g0
0
C)
0D
0D
f
C
0
3
Figure 22: Historic and Projected Demand for Metallurgical Products
67
Planned-for Capacity
Results in Lost Sales of $15.1M
Lost Sales
of $15.1M
COD
wD
w)
to
_V
O
CO
CM1
C
(0
0)
CD
O
CO
-4
CD
O
CO
O
CCD
CO
CO
CD
0
0,
NP. N
CD
CD
9333
CD
0D
NP
n9
N
Co
w~
N
0
0
--
0
0
CA-~
-n
Figure 23: Planned-for Capacity Results in Lost Sales
Using Plasma Furnace
Reduces Lost Sales by $3.OM
Lost Sales
of $12.1M
CD
CD
w)
-W
CD
-fb
CD
CD
CA
CD
CD
0)
CD
CD
CD
CD
OD
CD
CD
wO CD
D
0D
0
9
C>
0D
3
D
0D
NC)M~0
3
D
0
3
CD
0D
_P.
3
Figure 24: Using the Plasma Furnace Reduces Lost Sales
68
0D
0D
39-
5.2 Organizational Conclusions
Realizing the benefits of the changes recommended here is contingent upon committing
to its implementation. To achieve this commitment, the recommendation needs to be
accepted by the organization and viewed as valuable. There are several dimensions on
which the value of the change can be communicated. First, there is the economic
argument that is presented in the previous section. Second, there are structural changes
that can be made to highlight the interconnected nature of the melting and milling
operations. The pressure that results from tying the two organizations sloser together
could be used to build the consensus needed to find new ways of using existing assets.
Third, the cultural aspects of this change would need to be addressed. It is a Type II
change in the language of Heifetz. Recognizing this highlights the fact that the change
will affect the culture of the melting operation. Culture changes are often difficult
because of the meaning people assign to the their jobs and the way they do their work.
Finally, the key stakeholders of this change will need to be involved and become
advocates for the change. This may be accomplished by forming a cross-functional team
that has a clear goal and timeline and has the autonomy to make the needed changes to
the melt shop processes.
69
References
Ancona, D. G., T. A. Kochan, M. Scully, J. Van Maanen, D. E. Westney, D. E. Kolb, J. E. Dutton and S. J.
Ashford ( 1999). Managingfor the Future: OrganizationalBehavior & Processes. SouthWestern College Publishing.
Buzacott, J. A., and J. G. Shanthikumar (1993). Stochastic Models of Manufacturing Systems. PrenticeHall.
Carroll, T. J. (2000), "Using Postponement to Move from Job-Shop to a Mixed MRP/Job-Shop
Environment," Masters Thesis for the Leaders for Manufacturing Program at the Massachusetts
Institute of Technology, May, 2000.
Gershwin, S. B. (1994). Manufacturing Systems Engineering. Prentice Hall.
Goldratt, E. M., and J. Cox (1984). The Goal, 2 "d edition. The North River Press, 1992.
, and R. E. Fox (1986) The Race. The North River Press.
(1994), It's Not Luck. The North River Press.
(1997). The Critical Chain. The North River Press.
Gomersall, E. R. (1964). "The Backlog Syndrome," HarvardBusiness Review, September-October 1964,
pp. 1 0 5 - 1 15.
Heifetz, R. A. (1994). Leadership Without Easy Answers. The Belknap Press of Harvard University Press.
Hopp, W. J., and M. L. Spearman (2000). Factory Physics: Foundationsof ManufacturingManagement.
McGraw-Hill.
Nahmias, S. (1989). Productionand OperationsAnalysis, 3 'd edition. McGraw-Hill, 1997.
Noreen, E., D. Smith, and J. T. Mackey (1995). The Theory of Constraintsand Its Implicationsfor
ManagementAccounting. The North River Press.
Suri, R., and S. de Treville (1992). "Rapid Modeling: The Use of Queueing Models to Support TimeBased Competitive Manufacturing." OperationsResearch in ProductionPlanningand Control:
Proceedingsof a Joint German/US Conference. G. Fandel, T. R. Gulledge, and A. Jones (eds.).
New York: Springer-Verlag.
Suri, R., J. L. Sanders, and M. Kamanth (1993). "Performance Evaluation of Production Networks." In
Handbooks in OperationsResearch and Management Science, vol 4: Logistics of Production and
Inventory. S. C. Graves, A. H. G. Rinnooy Kan, and P. H. Zipkin (eds.). New York: NorthHolland.
Whitt, W (1983). "Approximations for the GI/G/m Queue." Productionand OperationsManagement,
2(2): 114-161.
Womack, J. P., and D. T. Jones (1996), Lean Thinking: Banishing Waste and Creating Wealth in Your
Corporation,Simon & Schuster, 1996.
70
Appendix A: Queueing Network Model
This appendix will provide background information and a detailed explanation of the
application of queuing networks to a flow shop. The approach described here is an
implementation of the work by Suri, et.al. (1993).
A.1 Background
The goal of modeling a production system in this way is to calculate several important
measure of production effectiveness: cycle time (CT), work in process inventory (WIP)
and throughput (TH). Little's Law relates these three measures to one another: WIP =
(TH)(CT).
One way to determine these measures is through the application of queueing theory. If
we can estimate two of them, we can use Little's Law to find the third. A steady state
queueing network can be used to find the total process cycle time (CT) for a given arrival
rate. For a system in steady state, the arrival rate will equal the throughput (TH). With
these two measures, we calculate the WIP of the process using Little's Law.
The sections that follow will provide the notation and governing equations used in the
queueing model to determine the process cycle time (CT).
71
A.2 Notation
r,,
The rate of arrivals to a station j.
r,,,
The processing rate or capacity of stationj.
t PiThe processing time of station]j. Same as
-.
1
rP
P
-
c2
2
The average time of either inter-arrival time or processing time.
The standard deviation of the inter-arrival time or processing time.
Squared coefficient of variation. It is defined as, C2
ca
Squared coefficient of variation of the time between arrivals to a station j.
ci2
Squared coefficient of variation for the processing time at station j.
c
Squared coefficient of variation of the time between departures from station].
mj
Number of parallel machines at station j.
U,
Utilization of station j.
I
The identity matrix.
P1,
The probability that a job will go from station i to station j.
PO
The probability matrix containing all pg for the network.
72
A.3 Governing Equations
What follows in a description of the method used to determine the throughput (TH),
process cycle time (CT) and the work in process inventory (WIP) in this paper. The
steady state average WIP, TH and CT of a process are found by applying Little's Law.
WIP = (TH)(CT)
Before this can be done, two of the three variables need to be determined. Start with the
throughput, TH. In a steady state system, i.e. no accumulation or generation, the rate of
arrival will equal the output rate or throughput. In our system there are three arrival
streams from outside the melt shop. We denote an arrival stream, ro
1 , as the rate of arrival
to station j from outside the system. Adding all of these arrival streams together gives the
total arrival rate to the system and thus the throughput.
TH
ro
Calculating the cycle time is more complex and requires the use of queueing theory.
Observe the equation below. It is the summation of the queue time and processing time
at each workstation].
CT =X(CT, +t, )
1
The processing time, tpj, is easy to determine. It is the average time it take the
workstation to complete a single job. It is the inverse of the stations processing rate, or
1/rPj.
The time a job spends in queue is determined using the Kingman equation. It is given
below. Notice that it has three terms. The first is an increasing function of variability.
The greater the variation in the arrival of jobs to the system, measured by c , or the
73
variation in the processing time at station j, measure by c2 , the longer the queue at the
station j. The second term is an increasing function of utilization, uj. If workstation j is
busy most of the time, the length of the queue will be long. A bottleneck operation is
defined as the one with the greatest utilization. This term captures this behavior. Finally,
the third term is an increasing function of the processing time. The longer the station
takes to process a job, the longer the queue.
/C2 +
CTj
=I
S
;
2
2
i
'
1 -uj
r.j
Looking at the equation above, there are four variables that determine the cycle time in a
queue at each stationj: rpj, c 2 , u, and c , . Each of these variables will be discussed
below.
The average processing rate of each station, rpj, is a variable that is determined via
observation or by using SAP data. The inverse, 1/ rpj, is the processing time of the
station. In the context of the queue at a station, it makes sense that the time a job will
wait to be processed will depend on the time the station takes to process a job.
The squared coefficient of variation for each station's processing time, c 2 , is determined
via observation or by using SAP data. For most stations a value of 0.50 was chosen.
This indicates that the process has low variability. Some of the workstations are more
variable and the data was readily available. In these cases the C , was calculated and
used in the model.
The next term is the utilization of the workstation, uj. If the station is busy most of the
time, jobs arriving to it will have to wait for the job currently in process to finish before
they can be serviced. The utilization is defined as the ratio of arrivals to the workstation,
or inter-arrival rate, to the processing rate of the station. If there are more than one
74
machine at a station, the processing rate will be the product of the number of machines,
mj, and their individual processing rates, rj. The formula for the utilization is given
below.
U.
-
mr
p
The inter-arrival rate to each station, r,, is the vector F. It is calculated by multiplying
the vector or the arrival rates to each station from outside the system, F, by the transpose
of the matrix formed by subtracting the probability matrix, PO , from the identity matrix,
I . The probability matrix formed by the inter-station transition probabilities. For
example, it all jobs from station 1 go to station 2, the probability is 1.0. The equation for
the inter-arrival rates is given below.
r = F0 (i - P0 )
Finally, the last term we need to determine is the variability associated with the interstation arrivals, c, . The determination of the inter-arrival rate variability is complicated
by the fact that this system is a network. Consider two cases. In the first case the system
is a simple line of stations where the output of one station arrives directly as the input of
the next station. A schematic of such a system is shown below. The arrival variability of
this system is relatively easy to determine. It is discussed by Hopp and Spearman (2000).
Figure 25: Simple Queueing Network
The second case is more complex and requires a more involved set of calculations. The
work of Suri, et. al. (1993) is used to develop the inter-arrival variability. In this case
75
there is one station, G, that is fed by two different stations, D and F. Also, there is station
B whose departures are split between C and E. The ratio of jobs that go to C and E is
determined by the transition probabilities mentioned earlier. If
of the jobs go to C then
the transition probability on this arc will be 0.25. These routing probabilities are given in
Appendix B.2.
C
A
D
B
G
E
F
Figure 26: More Complex Queueing Network
The equations to determine the C for each of the stations in a more complex system
(like the one in Figure 26) are given below. Interested readers are referred to the paper
by Suri, et. al. for a description meaning of each of the terms.
M
c
2
=a
j
+
+
C
- Cb.,
where
a. =I+
(qjc
bU =Wj pi q(1-U
- 1)+
qi [(1- p, )+ piuX,
i2)
and where
ri
x
=1+m- 0 . (max c ,0.2]-1)
vi =
76
2
W
V _u)2
= 1+4(
Departures
c
-2(
=1+(1-u2
(
c -1)+
u
(cP -
Splitting of departure streams
= piC2 +r
iC
P
Merging of arrival streams
I
2
+1-w
where
w = [1+ 4(1
-
u) 2 (
1
and
L
j2
v =
r
Irk
k
77
Appendix B: Melt Shop Model
Figure 27 is a schematic of the melt shop model that is used in this paper. Using this
model, each station j can be characterized by the 5-tuple, { c
,
,rC , r2,m }.
,
c
is the
squared coefficient of variation (scv) of the inter-arrival times to the network from the
outside. c, is the scv of the processing time of station j. ro is the inter-arrival rate of
jobs to the network from outside. rpj is the processing rate of station j. Finally, m is
the number of servers/machines at stationj.
Figure 27: Schematic of Melt Shop Model
The sections that follow will illustrate how each of these variables is determined for each
station. After showing the calculations for each of the variables that characterize the
stations the three scenarios are shown: Current, 50% to Plasma and 100% to Plasma.
B. 1 Current Arrivals to the System
This section describes the arrivals to the system. The calculations for the arrival rate , ro;,
are based on SAP data and are shown. The squared coefficient of variation, c
78
, are
assumed to be 1.0 because the arrival processes have high variability. Sensitivity
analysis indicates that the value of C
affects the performance of the system little.
B.1.1 Route I, Alloys E & D
Based on SAP data, there are an average of eleven process orders for blends each month.
That is eleven arrivals to Route I each month. Each blend, weighing 2,250 pounds, is the
equivalent of two jobs in this analysis because the ingot is cut into two pieces later and
each piece is a single unit of processing. Below is the processing rate that results from
these data.
2 jobs
arrival
arrival rate
11 arrivals
month
arrival scv
= 1.00 (assumed)
1 month
22 days
1.00 jobs/day
B.1.2 Route 11, Alloy Q
Based on SAP data, there are an average of four process orders for alloy
Q each month.
That is four arrivals to Route II each month. Each of these weighs approximately 800
pounds and is the equivalent of one job in this analysis because the ingot is cut into two
pieces later and each piece is a single unit of processing. Below is the processing rate
calculated from these data.
arrival rate
arrival scv
4 arrivals Y I job
=
I=
month
arrival
1month
22 days)
0.20 jobs/day
= 1.00 (assumed)
79
B.1.3 Route IlIl, Alloy W& S
Based on SAP data, there are an average of seven process orders for Route III each
month. That is seven arrivals each month. Each of these weighs approximately 1,000
pounds and is the equivalent of one job in this analysis. Below is the processing rate
calculated from these data.
arrival rate
arrival scv
80
(7 mot
arrivals Y 1 job 1month
ariva0.32
~month
arrival
= 1.00 (assumed)
22 days)
jobs/day
B.2 Routing Probabilities
The figure shown here is a schematic of the melt shop model. Each arc connects two
stations. The flow of material through the model is determined by the transition
probabilities. The transition probability is the percentage of a station's output that flows
to another station. The method for calculating these factors is given in the table below
for each of the three scenarios. The terms "r's" are rates. For example, rj.j is the rate of
material arriving from I to station 1.
81
Table 5: Routing Probabilities for the Three Scenarios
Probability
Current
50% Plasma
(from-to)
PI-2
1.0
P 2 -3
1.0
/(r
+ r,_-3)]
[r,-, /(r,
+ r,,_3)]
P 3 -4
(0.20) r_,
P 3-5
(0.80)
P3-6
0.0
0.0
(0.50) try
/(r,_I
+ r,_-3)]
(0.50) [rr
/(,_
+ r_)1
r,,-3 /r,_,
P3 -1
P 4-7
82
10.0
1.0
1.0
0.0
0.0
P7-8
1.0
P8-9
1.0
P9-Io
1.0
Pio-Il
0.40
P IO-Out
0.60
P1 1-12
1.0
P
2-out
Ir,_- /(rI- + r_13)A
0.0
P6-11
P12-9
0.0
+ r-3
1.0
P5-7
P6.7
100% Plasma
1.0 -
r_3 /(r,_3 + r,,_3 + rI,- I
rI-3 /('-3
+ rII-3 + r,,J--J,
)
B.3 Current Processing Rates
This section describes the processing rates of each workstation in the Current scenario.
The calculations for the processing rate s, r,1 , are based on SAP data or observations and
are shown. The squared coefficient of variation, c 2,, , of most stations are assumed to be
0.50 because the processes have low variability. For workstations that have highly
variable processes, the C is calculated and is shown.
Mill (1):
processing rate =
process scv
16 hours
I day
lot
1 hour
150 lbs
lot
2 jobs
= 2.14 jobs/day
2,250 lbs
= 0.50 (assumed)
Blender (2):
processing rate =
process scv
16 hours
I day
lot
13 hours
2,250 lbs
lot
2 jobs
2,250 lbs
2.46 jobs/day
= 0.50 (assumed)
1500T Press (3):
processing rate =
process scv
16 hours
I day
lot
8 hours
2,250 lbs
lot
I
2 jobs
2,250 lbs
4.00 jobs/day
= 0.50 (assumed)
83
ABAR Sintering Furnace (4): The ABAR sintering furnace has a long cycle time. It is
the only sintering furnace that can process raw material AA. A typical sintering lot is
750 lbs and requires 3 separate runs to complete the batch of 2,250 lbs. Completely
sintering a batch takes one week, with all of the non-productive time taken into account.
This behavior can be characterized by the following set where each number indicates the
number of jobs that are completed on a given day. This set is assumed to repeat. In this
way an estimate of the mean and variance for the process are possible.
Set: {0, 0,
,0, 0,0, 2}.
processing rate=
process scv
=
r3lots
7 days
a-
0.76
2
-
lot
2jobs
2,250 lbs
=750lbs
0.29 jobs/day
)2
=7.00
=
( 0.29
GCA & AC Sintering Furnaces(5): There are other sintering furnaces used to prepare
the raw materials of type MG and HY. These are the GCA and AC sintering furnaces.
There is one GCA and two AC furnaces. Their processing rates are slightly different.
Their variability is observed to be similar. Instead of treating them as separate servers,
they will be treated as a group of three servers, m = 3, whose processing rate is their
weighted average.
GCA furnaces:
processing rate
84
16 hours
1 day
lot
21 hours
750 lbs
lot
2 jobs
2,250 lbs)
A Cfurnaces:
lot
150 lbs
2 jobs
2.75 hours
lot
2,250 lbs
p16 hours
I day
Combined (1 GCA furnace and 2 AC furnaces):
processing rate =
process scv
=
3
(0.51 jobs/day )+ - (0.78 jobs/day ) = 0.69 jobs/day
3
0.50 (assumed)
Plasma Furnace (6): NOT USED
EB Furnace(7): The EB furnace is managed using campaigns. A typical campaign is
four ingots that are melted two times in the EB furnace before being passed to the next
station. This behavior can be characterized by the following set where each number
indicates the number of jobs that are completed on a given day. This set is assumed to
repeat. In this way an estimate of the mean and variance for the process are possible.
Set: {0, 0, 0, 0, 4, 4, 0}.
processing rate
4 lots
7 days I
process scv
-
2,250 lbs
2 jobs
lot
2,250 lbs
= (1.95
p1.14
1.14 jobs/day
2.92
85
Quality Control (8): Quality control is assumed to behave like a pipeline delay. In an
attempt to model this delay, the processing rate is set so that the total cycle time (queue
time and processing time) for this station is at least four days. The variability is assumed
to be less than the EB furnace but still significant.
processing rate
= 1.14 jobs/day (assumed)
process scv
= 0.50 (assumed)
Lathes (9): There are two lathes. Both of them operate with the characteristics identified
below. In the model, both lathes will be used, m = 2.
processing rate
2 shifts
day
process scv
f
0.75 jobs
shift
1.5 jobs/day
= 0.50 (assumed)
2000T Press (10): The 2000T press works on a mixture of job types that have varying
process times. Data compiled from daily shop floor meetings from Aug-Nov 2000 give
the following results:
86
processing rate
= 3.81 jobs/day
process scv
= 0.65
Welding (11): There are two welding stations in the melt shop. The model, however,
will use only one machine, m = 1. This is done for two reasons. First, there is only one
welder. Second the machines do not to work on all incoming jobs. Alloys S and W are
welded in one machine. Alloys D and
processing rate
process scv
Q are welded in the second machine.
2 shiftsY2 jobs~
i
= 4.00 jobs/day
day
shift)
= 0.50 (assumed)
Arc Furnaces (12): There are two arc furnaces in the melt shop. The model, however,
will use only one machine, m = 1. This is done for two reasons. First, only one furnace
can be operated at a time because of a limitation imposed by the plant's electrical
infrastructure. Second the machines do not to work on all incoming jobs. Alloys S and
W are melted in one machine. Alloys D and
Q are melted in the second
machine.
The arc furnaces work on a mixture of job types that have varying process times. Data
compiled from daily shop floor meetings from Aug-Nov 2000 give the following results:
processing rate
= 2.44 jobs/day
process scv
= 0.40
87
BA 50% Plasma Processing Rates
This section describes the processing rates of each workstation in the 50% to Plasma
scenario. Most of the rates, rj, are the same as the Current scenario. For instances where
this is not the case the calculations based on SAP data or observations and are shown.
The squared coefficient of variation, c2, of most stations are assumed to be 0.50 because
the processes have low variability. For workstations that have highly variable processes,
the c is calculated and is shown.
Mill (1): Same as current process
Blender (2): Same as current process
1500T Press (3): Same as current process
ABAR Sintering Furnace(4): NOT USED
GCA & A C Sintering Furnaces(5): Same as current process
Plasma Furnace(6): The Plasma furnace is not currently being used. The figures used
for the calculations below are estimates of the expected performance of the plasma
furnace.
processing rate
process scv
88
=
=
16 hours
I lot
(16
hours
pday shours
0.50
1,950 lbs
1,9
lot
2 jobs
50 lbs
=oo
2.77 jobs/day
2,250bs
EB Furnace(7): Same as the current process. The EB Furnace will need to continue
processing first and second melts in batches. As a result, the queue time will remain long
and the variation in departure times will also remain the same.
Quality Control (8): Same as current process
Lathes (9): Same as current process
2000T Press (10): Same as current process
Welding (11): Same as current process
Arc Furnaces(12): Same as current process
89
B.5 100% Plasma Processing Rates
This section describes the processing rates of each workstation in the 50% to Plasma
scenario. Most of the rates, rp, are the same as the Current scenario. For instances where
this is not the case the calculations based on SAP data or observations and are shown.
The squared coefficient of variation, c2, , of most stations are assumed to be 0.50 because
the processes have low variability. For workstations that have highly variable processes,
the c , is calculated and is shown.
Mill (1): Same as current process
Blender (2): Same as current process
1500T Press (3): Same as current process
ABAR Sintering Furnace (4): NOT USED
GCA & AC Sintering Furnaces(5): NOT USED
Plasma Furnace (6): The Plasma furnace is not currently being used. The figures used
for the calculations below are estimates of the expected performance of the plasma
furnace.
proessngrat
process scv
90
=
( 16phours
dat =
= 0.50
1 day
I lot
10 hours
X
1,950 lbs
lot
2 jobs
-2.77
=
2,250 lbs)jb/a
jobs/day
EB Furnace (7): In the proposed process, the EB furnace would not be operated using
campaigns. Ingots would be melted only once. The figures used below are estimates of
the expected performance of the EB furnace if it are operated in this way.
processing rate =
process scv
=
16 hours
l day
YI
lot
1l hours
1,950 lbs
lot
2 jobs
= 2.77 jobs/day
2,2501lbs)
0.50
Note that the number of hours per day drops from 24 to 16 but the processing rate
increases from 1.14 to 2.77 jobs per day. This is attributable to the elimination of
campaigns with the associated elimination of long queue time and increase in throughput.
Quality Control (8): Quality control is assumed to be a pipeline delay. In the 100%
plasma scenario the processing the processing rate is changed from 1.15 jobs/day to 1.49
jobs/day, a 30% increase in capacity. This is done to keep the utilization below 85%. It
is assumed that this change could be made.
Lathes (9): Same as current process
2000T Press (JO): Same as current process
Welding (1]): Same as current process
Arc Furnaces(12): Same as current process
91
B.6 Actual Data: Current
SAP Data: January 1999 to June 2000
18 month period
22 days/month
Mat'l No
Description
MG
AA
02000053
02000014
02000117
02000118
HY
MG
per month
73
34
36
119
4
2
2
7
262
02000113
02000107
W Scrap
S Scrap
E Ingot
E Ingot
E Ingot
Q Ingot
W Ingot
S Ingot
D Ingot
02000121
02000120
02000015
02000000
02000002
02000001
02000003
02000008
02000083
02000108
02000106
02000114
E
E
W
S
D
Sheet Bat
Sheet Bat
Sheet Bar
Sheet Bar
Sheet Bar
15
Work In Process Inventory
(process orders per month = jobs)
Blends
Scrap
Ingots
Sheetbars
WIP =
15
7
48
32
101
83
35
5
2
118
7
31
244
17
134
174
142
119
2
14
1
7
Throughput
(Sheet Bars Shipped per day = jobs/day)
10
divided by
861
48
140
8
2
9
7
6
32
34
164
122
109
569
92
Calculations
Process Orders
18 mo. total
8
7
TH =
32
22
1.44
jobs
Sheet Bars Shipped
days/month
jobs/day
Cycle Time
Little's Law: CT = WIP/TH
WIP =
101
TH =
CT =
1.44
70
jobs
jobs/day
days
B.7 Scenario #1: Current
1.00
1
707. 00)
1 ,004
1.66
m
~Syria.
72 days
$*MsCTTolal
Modlcn
IN
mm
RO
M..d
Pttr.s
1
~
214.V*00114
&4
0.97
214
45%
0.85
t=
101.1011011 SCV
(13
CO
0.5
P1.ss
CT-To3
72
days
4
~
.l40.'17~.01
*7~5
5.
41o
Aw
2.6
1.17
4m
0.19
39%
29%
67%
0.79
02
0.4
0,91
4.54
029
40.2
3.4
1
.3
Plaoaaa
5
GCAfAC
3
Us
-f -
n4*
n
Procamdng"
ABAR
2
nIm
to
m
U78
DC
EF
149 job./day
Lai"
6
207
37%
ao
0.97
1.14
85%
n
097
1.14
85%
0.99
0.6
14
14.4
0.9
283
1.14
4.1
w
72%
a
02
D.3
0.3
1.26|
07
0.9
I.
1.39
4-O
3%
11 48
2A17
1.81
57%
10D
11
01
12
13
7'
n
217
0.1
Anc
9
O01377
I~7
Wald
Foq.
8
7
.~3
.~.,
N.-,-r
P.,fImanc
107 job.
WPTotal
THTeial
1,39
244
67%
2,60
0.0
0,
N,
knpu Dote 1ewlica
McId.
nd_
ra"
RMI
14h
ION
W1.9
A,.
DC L0th.F.g
9101,4 PrnMI ABACAI E5F
1
2
3
4
5
7
0
9
10
11
41
3
2
1
1
1
4 2.4 10
214 2.46
4
0.29 0.09 1.14 1.14 1.5 3.81
05 0 0.5 7 0.5 2.92 0.5 05 0.65 05 0.4
I
097
0.20
032
1.0
101
1.01
NIl
R
1
1
1
1
rtep
Zw-.
Owl
0.97
0.20
12AWA 13
GCAIAC
EBF
1
1
Mil
P-es
swwn
OC
Lath.
Foggp
Wald 0.32
Arc
OuO
Nevler V1,00abmo mai~a t r i
2
3
4
5
-1,1'eJ
I
I
110
I.O
7
9
1
11
10
1.0
12
Ow_
1rm/o
___
047
1
1-00
2
0
0.62
2
3
I
2
3
1.0
A7
0.06
0
5
10
1
1
I
1
11
070
7
60
1
1.01
.1
- -(%-I)
1
2
4
1
1
1
-qplax I,-m
1
1
i
3
4
5
7
1
7
0.
1012
9
1
1 .
OD ;
1 0D
( 0.0)
10
0
0
1.
1.
)
77
t
10
I
1
fO
10
0.99
645
12
II
13
0.93 0.37
#N#W
12
13
freMI/I.J
3
4
1
11 12
1 13
1
2
0
153.01;
1.0
1.32 0 63
2
3
0.27 0.97 1.17
UImllid
1
045
045
2
0.39
039
4
0.19
3
-5
7
0
0.78 0.97 Q97
2.17
4
67 0.37
067 00
(1.4 D"
11
I
1
4
1.53
W..,
.72
1
01 0.85
001 0.85
4
12
1 14 1.87
9
10
13
0.01
12
It
13
14
' I00.1 1
0.24.
2
3
5
12
1
0.03
a
,
0.12
5
a
9
a
IM
4!
11
10
9
ill
00000
D 15
2.Y7
m
027
0.16
3
2
4
5
0.9
7
.
8
7
9
11
I0
27
0.04
0.29
0.16
0.16
4
5
7
1
11
12 13
072 0.57
035 0.57 015
0.72
057 035 057 015
0.01
0.16
O
1.-ale 1.r la W
a- fib 1. sw a 1-1)
11
12
13
Q2
27
S0.36
1
3
11
0 28
1O
4121
13
13
003
100
10
11
12
.3T
a
7
0.34
0.32 0.01
3
777
7
054
0.95 2.1 035
4.25
R,
S5
1,00
12 13
3
0
0.37 0.
5
I J-
b
1.3
1 49
realm
S
1
2.17 1 39
alec ouae
9
7
0.29
029
9 10
0.20
a
7
.
1.00
1
3
0
1 32
. 0}
5
0 0
13
.83
:
Q ii,
11
wI
1.01
0.27
11
6m.0
4
9
a
10
0.97
03
rm/to
N0.fork
2
.,
3
7
PP.1
A
5
6 7 -1
a
9
10
11
12
1 2 3
A 5 1 8 9 10 11 12 13
91 1M 09 117 0.1 09
0.97 083 07
1 077 0.77 1W9
0.17 01i6 .77177
01 3 4A9 4A9
OD
W2 017 GQffi 0.83 0.83 t 49 1A 077 O
S100
00 OD 53 .53
061 .61
s
1m
OD t 1 .53
061 OBI 1m G
1
.3
61 6 .'00 .1 2
.O12 153 '.53
1.53
1.53
61 0 61
to
O.S3 I 5 061 0.61 1.O
a 11
132 1.32 153 153 1.01
1
. - WP,
2
I
0.05
3.10
0
F.W1m.,Ial
i
5
4
1.55
000
7
12
213
7
1.3
1
O
5
3
0.61
3
0
0
1.31
1.2
1.0
02
7
.
9
7
12.97
0.07
2
10
II I I. - P)
o
/l/
J
1
2
1
10
fl
3
2
1.0
.
3
4
5
7
0.1
6rm/toJ
17
1
- mafto
5
6
1
P)
2
019
7
01
9
12
13
3
m.qwt
1
7
200.
0.03
1
4
3
1,32
76
12
x4S- l.,~
II.
11
3
.
11
10
4
0.19
3
458 1.32
0
1640 15
038
V2 0.66
13 0
13
1.0
D 13
3087
039
1.0
10
1.40
9
IDD
7
10
12 JM
8
7
0.95
0.16
061
U.2B
2Z29
ImO 1
1O0
4
2
5
4
1 O
101
12
t0
3
870.62
12
13
0 02
0.78
S04102
O
014
to
0.31
1 2
095 0.79 0.91
3
4.54
4
099
5
2.83 1.14
7
1.26
107
D.W
O.B
8B
2.6
1.48
9
F.W9
1011 11
12
93
13
B.8 Scenario #2: 50% to Plasma
100
100
1.8
03100
0.1
MNo-r
S~stam
P.rfIwmanc
CTueaw 51 days
76
1 49 jb/d7y
jobs
WAPTotal
THT1t.l
Ml lod
RU
M.Acn.
1
P-
GCA/AC
5
AAR
3
2
01
06
-
IOU
.
4
Pha.
EF
6
7
0.49
O.R7
1.14
AC
a
La..
Farg.
9
10
W.1d
11
2.17
139
out
A..
12
13
S'e-ne
-
p
Mwte
0.97
14
2.46
45%
39%
0.05
0.76
a
0
0
0
51 days
Robe
Anriva
I
P,c,
D.97
2.
gRaue
UtOl....
laleawfloalSCV
Quues
CTTP100.ia
CO7T411l
1.17
4.00
0A49
29%
2.07
23%
0.57
0
0.99
0
0.97
2.17
85%
84%
72%
18.96
39
1
391
a9d
2.77
111%
0
57%
0.59
0
1
3.64
1.49
39
12"4
35%
57%
16%
155
0
069
099
0
0
1.1'ad Det tnoic.4
RM
MI
0.97
0.20
Pres
M0.45, RM MUM Woo.d
INu
1 2
1
1
1
11.-P
-- p
fwft11
2.14
0.5
0.5
0.97
P...
3
4
(0
C03
246
10
181-&
GCA/IPI,...EF QC
5
1
0.39
0,5
LA.e Forg.
7
1
3
1.14 1.16
2.92 &S
6
2.77
0.5
1011
1
1
.5
0.5
3.81
0.0
W.M
A.m Oat
12
13
2
T
1
4 2.44
10
0.4
50.
0.32
GCA/AC
I
1
I
EBF
()C
Forgp
18M
032
WoMd
Amc
1.81
wu
Netb..k
Pmebabftl
IrDmIltoj 1
E
ltw
a
2
3
-b 5
1p)
7
6
9
S
10
1-
q
Ow som
12
11
Faft A
toml/I.J
0.47
1.O
I'M
IM
I.8
1.00
1.00
2
2
I-o1m
0.41
3
0.41
6
181
1
II,8oDAR
Ill
1.00
O
0.13
087
12
i
lJ
1m
2
1
3
11
13
-
5
3
6
8
:)
.I00t
7
9
1$x I
181
-
.....
Mf
1
7
1-m/IJ
1
tI
12
93179
67
10
2
8.1.00;1.5
7
Irm/~
10
73
4r
.
.
.10.300
01000
7
a
9
0.00
1.02
0.97
645
6
5
0S0
2.92
6
5
0.00
a 3=
tP .uq L,
in
2
UU7
61.0
1.0
6
12
13
1.00
04
061
5 /.0J
1112
6
0.70718
003.0
0 0103
1.00
11CID*
1.09
1,01
1.00
5
1 9 1o
1001
2,
2
1
0000
1.W
2
2
3
5
5 6
4
a
a
1
10
I1
12
7
3
4
.. 5.
6
7
S
3
5
.DD
0.41 U93 U.83
01 O 0 041 0.1 0.
0.83
0.21 0,l1 0.83 0.83
100
0.41
I. OD
".
IOU
It
1
I .0
1
Il
5
I4
"
149
1.49
1
149
I
13
v1
Q77
149
77 03
D.77
0.77
77
U.1111raft"
Lw I
FIP.a .
IL
2
3
.03 0.37 0.1
0.00
0.64 0.5301
2
i0xl
I
1.14 1.87
7
6
II
19
12
11
12
13
020
10
I
001
9
2
10
13
13
12
19
014
5.14
5 14
0.0
0 00
5
10
0
1.0
6
0.35
5
3
7
7
6
1
12
11
47t
32
1
9
8
13
099
10
I
12
13
0(13
073
0,29
037
005
015
000
2.07
000
028
1.0
3
1
034
2
fib
-
5
1
1.9)
6
49
8
7
am D
9
10
11
13
12
003
673
073
1
7
pF.)
9
5
0.82
054
am1 000
1.00C
0.6 073 V .0
.03)
5
133
16
0,21
003
D
6
01
029
0.37
*
3
9
005
0.00
1001
11
207
00
0.23
12
13
1
2
085 0.87 090
94
13
12
44A3
0.34
B
: , I.5 .5 ".6 061 IM
.53
0.:1 90,61 1.00
OD 00 .53
10
Io 1OD 53 :.53 0. , t 0 1: ':I1
1 53 1 53 M.61 12 0.61 1IM
13
D.61 I.M
0.63 1.53 0.61
132 132 1.53 1.53 1M
1.32 1.32 0.53 1.53 ia
s00mavoloft1
l
2
FomIIIOJ
'00 1
)I
7
-
7
0.2D
7
12 13
Int-mal Traffc Arrfval Rates Noall 0) N
0
9 111 11 12
2
3
5 6
1
0.97 097 1.17 0.49 0.49 0.97 0.9 2.W 7 17 1.39 t.39
oft ech Wortd-~e (. flo g
1
2
3
5 6 7
9 to 11 -12 0.15 13 02 1 .23 11: 1,85 1.8 0.72 11W 03 .5.17 .I
O3 29 .23
9 .5
0.4 02 DS7 0.35 0.57 0.15
BN
r-
i.
.,,e
731.0
1
3
':'
0 40
024
bIIIl.J 1
067
1 0
it
037
026
2
'.
to
0.93
035
1.93
.A
I
9
3
028
I -0
1 0.7
1.0031.00
.
.2:.
7
0.50 0.65 0.65 000
3
'07
7
7
tA
A
003
13
1.45
10
11
.80
100 (t.41
015
6
*
3
6,
1
0.03
2
3
5
6
1
5
3
132 OA
2IL
0.68 0.61
toonm
1
complvx
1I
2)
3
2
D00 0.00 0.0 0.71
0.19
131
Mar0 -
611.D
q
1.32
1 .-
}
1
'5
0.38
000
- 10*.IA0I.p)
12
w
0.39
12
Is
1
51
121
2
0.99
I2
O
4.58
9d2ft
12
i
016
0.10
7
1
1I
IS
096
1w
181
100
9
8
7
6
2.43
I.1O
I
5
3
2A3
1.00
100
2
240
5
6
1.00
0.17
7
toI
1
0.62
0.62
3
l.0
1D
SI~m
8
2
5
3
3.47
100
6
~
0.64
1 18
7
0,87
a
1.50
I
160
099
i
11
12
13
13
B.9 Scenario #3: 100% to Plasma
0..86
014
11 mSysnew
RM
Blend
o
Pfass
nas i
2
2.46
40
52%
37%
CTe e e
0.77
0.4
0.70 0.05
03 01
CTPeoees
CT Total
0.01
11 days
04
SCV
46%
8
8
40
269 1.79
381
71%
60%
1.11
221
0.85
05
0.3
1.3
0.707
10
9
24
2.69
40S
85%
3.2
out
Amc
7
1.27
1.49
0.92
0.2
(.4
0.4
13
Weld
6
1.27
2.77
46%
1.03
02
Forge
1M t
27
1.27
2.77
147
59%
Logn
days
jobs
jobs/day
5
4
Z0
1 27
I.27
2.14
C
EBF
3
2~~1
Illy 1
p
later Anvel Rae
Processing Ra.
Utelz
n
Into AfIfWl
Ploenoo
GCAfAC
ABAR
1
11
20
1.79
W5iPToa
THTotal
Machine
lOW
Meassres
Peroonce
CT_-Total
1.42
04
1.59
244
65%
0
1.11
.9
02
0.3
1.59
4.05
40%
00
01
304
Input Data werl6cali
Machine rate 0
Rid
MA 1.27
RE
Macb.
to#
M
olep
MIII
I
serlp
2.14
05
1
Bled Prosslawe
1
2
I
2.46
0.
se%.
1
2.77
0.5
4
0.5
0.20
1.00
1.27
1.0
'ale-D
EBF
QC
3
LaheForg
6
7
1
1
2.77 1.492 1.5 3.81
0.5 0.5 0.5 0.65
Blend
WeId Arc
8
9
Oa
1
1
4
Z44
10
0.5 0.4
0.32
1.0)
10
3
020
Press
Plot"a
11
13
12
EOF1
0C
Lathe
Forgo
1
1
1
W.d0.32
A.:
"evmork ProbabilityMbx ( P)
I f to J 1
2
fo.
3
E
6
7
9
S
10
1.00
10
I - W.P IWOe A
fom l11tto 12
3 1 Out2
0 47
1101
O
.
1.
1
2
3
1.05
0.14
0.05
IS
1.05
1.05
1.05
1
1.05
0.0
0.11
col|-
eem(
1
3
6
9
N
7
1
6
13
11
1
x
13
I
1.32
1
1
030
3
I
2
105
1
0.37
7
8 1
050 050 071 065
0.50
3
172
0056
0.0
B
645 0.94
7
029 057
6
3
00
077
1
Metrix
6
7
1.39
t07
1.05
1
6
1
3
2
6
1
2
3
11
7
9
11
1
12
13
t01.05:
035
SR
11
12
10
11
12
04
0D0
8
1.10
13
45.09
16
9
1.43
13
020
12
1i
13
001
to
I
11-:<
1.0
13
frm ,
1
1o J
1
1.05
2
3
6
5
4
3
0.05 0.05
0.5 0-5
0.5 0.05
1.05 105
1.0
6
0.05
005
006
105
105
105
5
6
1.53
1.53
153
155
155
155
16
0
7
S
D 10
S
10 12
11
13
Interelt
.11
3
I
ToMe
1
1.27
Rules(r
3
6
Arriwal
2
127
1.47
6
7
1
0.75
0.75
0.75
062
0.62
062
0.62
7
1.53
1.53
1.53
1.55
105
155
10.5
5.155
138
138
1 38
138
0.75
075
0.75
0.62
0.62
0.62
0.62
1.62
1.55
0.55
7
127 2.69
127 1.27
9_
8
9
1
1.05
011
9
013
0.24
0.05
0.28
0.30
- IF. a.qe
1
2
11M;0.32
0.77
105
1
.
I.w I
/1,J
J.w
3
0.62
0.12
1
0.49
D
6
2,!,:
3
0.65
6
1.47
9
11
1N
0.70
8
9
0.99
--.
11
10
014
0
1.34
12
070
swunven
Inerb
1
rMI/IoJ
11
13
0.02
007
1.59
12
0.55
11
to
13
12
1.29
0.79
1
In sme al
2
0.38
1
2
3
2.69
0.37
7
105
105
9
a
0.11
2
11 0.35
2 10
11
12
13
3
6
7
7
10
10
105
0.62 100
10i5
100
1055 10
a)N
8
-(Fate
13
12
025
bi - IIP,q
1.05
PF
11
011
11
12
f1
04.-,
l
1
N- (I -
10
/
11
obork Matrix
2
1
2
1.05
1.05
1.05
1.05
1.05
9
N
*
I i frp?
05
7
0
x
067
8
01
q
7
1.05OU.321
0
V!)
11.
6
FoodseonotalNe
1
=oI/loJ
u,
lIP,
2
3
6
0.12
1
004
026
in*
2
3
I.P
6o/toJ
tamn
comple
1
1
2
3
4
1.32 064
038 0.15
1 .64
I
123
from
039
0,0
6
3
msv p
1
2
0.5
0.50
-
11
I3
o
11
||
q
0
408
Mabix I)
1
0.99
1.05
1.05
2
020
1.05
0,4
1
1
186
1.05
0.05
Ideety
Im IlloJ
1
2
3
0 13
1.
1.0
105
105
11 12 13
0095
1.0
I
91
12
10
1 .0
0.62
1.0
7
1
8 69
71
am
125
6
5
1.00
6
0 62
087
1
2
3
1.00
-
f0b
,
sew a
3
0.05
-1)
6
7
9
8
1
11
13
12
105
024
6 12
1.59
0.01
13
1.79
081
05
1.12
Uilizatein
1
0.59
0.59
09 each Worlsaten
2
3
0.52
0.37
046 0.46
0.52 0.37
0.6
0.46
6
It"
I
-
r
/
(ride
pIen
)
7
19
11
9
12
085 0.60
0.71
0.40
0.05
0.18
065 0.18
0,71
0.40
0 85 0.1
SLV a
012
13
0.
1.91
0.81
1
077
2
0.70
3
005
505
6
7
8
9
112
092
1,15
2,21
10
085
11
1.42
12
13
1.05
099
95
19
12
Appendix C: Economic Analysis
ASSUMPTIONS
Gross Margin
Discount Rate
$20
10%
DEMAND (Ibs)
Year
Demand
CAPACTY (lbs)
Current
% diff
Current
New Facility
+ Plasma
1993
1994
1995
1996
1997
1998
1999
2000(P)
2001(F)
2002(F)
2003(F)
2004(F)
2005(F)
201,271
233,943
264,453
345,658
286,228
256,116
309,205
361,999
419,289
475,950
642,784
707,063
777,769
REVENUE ($000)
Year
Potential
Sales
2000
2001
2002
2003
2004
2005
NPV
7,240
8,386
9,519
12,856
14,141
15,555
$51,706
Lost Sales
300,000
300,000
300,000
300,000
300,000
300,000
300,000
300,000
300,000
300,000
300,000
300,000
300,000
16%
13%
31%
-17%
-11%
21%
17%
16%
14%
35%
10%
10%
100%
360,403
360,403
360,403
360,403
360,403
600,000
600,000
6,000
6,000
6,000
6,000
6,000
6,000
$28,745
6,000
6,000
6,000
6,000
12,000
12,000
$36,568
6,000
7,208
7,208
7,208
7,208
7,208
$33,324
Current
+Plasma
+New Facility
6,000
7,208
7,208
7,208
12,000
12,000
$39,573
($22.962)
($15,138)
($18,382)
($12.134)
Current
+New Facility
Current
Current
+Plasma
$3,004
Additional Sales Possible by adding Plasma to Current+New Facility:
INVESTMENT ($000)
Year
20%
Current
+New Facility
Current
2000
2001
2002
2003
2004
2005
NPV
0
0
0
0
0
0
$0
0
2,000
2,000
2,000
2,000
2,000
$7,582
Current
+Plasma
0
1,000
0
0
0
0
$909
Current
+Plasma
+New Facility
0
3,000
2,000
2,000
2,000
2,000
$8,491
SUMMARY
96
Current
Current
+New Facility
Current
+Plasma
Current
+Plasma
+New Facility
NPV of Project
$28,745
$28,987
$32,415
$31,082
Subtracting lost sales
$5,783
$13,849
$14,033
$18,948