Table of Contents

Go Back

A System Dynamics View of the

Theory of Constraints
By
Author Co-Author
James R. Wixson, CmfgE, CVS James I. Mills, PhD.
Advisory Engineer President
Applied Systems Engineering Sustainable Learning Systems, Inc.
Bechtel BWXT Idaho, LLC 311 N. Placer
PO Box 1625 Idaho Falls, ID 83402
Idaho Falls, ID 83413-3634 (208) 522-0378
(208) 526-7784 [email protected]
[email protected]
[email protected]

ABSTRACT

System Dynamics can be used to facilitate the understanding and develop alternatives for a
system. In the same way, The Theory of Constraints (TOC) is based on Eli Goldratt's work on "how to
think"1, System Dynamics (System Dynamics) is based on a way of thinking about systems from a global
perspective. The primary application of TOC embodies a systems thinking approach to manufacturing
systems. By knowing how to think from a systems perspective, we can better understand the system
under study. Through better understanding, the performance of these systems can be improved. The
concepts of ongoing improvement embodied in the TOC are enhanced using System Dynamics. By
providing a way to model and simulate the system under study through System Dynamics and applying
the rules of TOC to the model, alternative solutions to improve system operation can be developed.
System Dynamics software further enhances this process by adding the ability quickly evaluate various
alternatives versus testing these alternatives on live production runs that may take weeks, or months.
INTRODUCTION

The central concept of TOC is the acknowledgement of cause and effect. The core idea of TOC is
that “Every real system must have at least one constraint.”2 Just as the thinking processes of TOC
provides a series of steps that combine cause-effect and our experience and intuition to gain knowledge3,
System Dynamics provides a methodology for studying and managing complex feedback systems, such
as one finds in business and other social systems.4 System Dynamics and TOC provide the tools to
understand system constraints and thus make it possible to improve system performance.
One extraordinary benefit of these thinking processes is that they provide the ability to recognize
the paradigm shifts that occur when times change but our assumptions and rules don't. We cannot
constantly monitor every assumption to be sure we are in line with constantly evolving reality, so the
ability to spot the shifts can be a real advantage. Those who continue their patterns of operation,
regardless of the changing reality, will suffer when the effects of their actions are not those that they
expect5. Eli Goldratt's novel The Goal6 completely exposed the magnitude to which this problem can
exist.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 TOC uses a system perspective to analyze and understand the entire system, not just parts of the
system. This makes it possible to identify elements of the system that constrain, or limit, the output of the
system. By identifying such constraints, it is possible to find ways to alleviate these constraints to
improve overall system performance. As such, it uses a system thinking approach to facilitate the
understanding of how a particular system works. System Dynamics provides the additional steps of
constructing and testing a computer simulation model, and testing alternative policies in the model to test
various hypotheses for eliminating system constraints. As Jay Forester states, “Without a foundation of
systems principles, simulation, and an experimental approach, systems thinking runs the risk of being
superficial, ineffective, and prone to arriving at counterproductive conclusions.” (Jay Forester,
Germeshausen Professor Emeritus and Senior Lecturer at the Sloan School of Management,
Massachusetts Institute of Technology). System Dynamics addresses this concern through computer
simulation of the system at hand.

System Dynamics methodology follows the scientific method in the identification and resolution
of system problems to improve performance. This methodology follows the following steps:
1. Identify the problem
2. Develop a dynamic hypothesis explaining the cause of the problem
3. Build a computer simulation model of the system at the root of the problem
4. Test the model to be certain it reproduces the behavior observed in the real world
5. Devise and test alternative policies that alleviate the unwanted behavior
6. Implement the solution 7

Thus, System Dynamics provides a rigorous approach to solving system problems that can be
applied to systems that have limited output due to constraints that have been built into the existing
system. Furthermore, newly proposed systems, and propose changes to these systems can be modeled to
identify potential constraints before expensive mistakes are made. The Theory of Constraints
methodology provides specific guidelines for modeling system behavior and analyzing systems to
identify and alleviate system constraints. The combination of applying Theory of Constraints
methodology and building a computer model of the system using System Dynamics software provides a
superior tool for analyzing systems and improving their performance. There are several different System
Dynamics software packages available. The one chosen for this project is Stella by High Performance
Systems, Inc. 46 Centerra Parkway, Suite 200, Lebanon, NH 03766-1487

CONSTRAINT MANAGEMENT

To manage constraints (rather than be managed by them), Goldratt proposes a five-step Process of
On-Going Improvement of existing systems. The steps in this process are:
1. Identify: In order to manage a constraint, it is first necessary to identify it.
2. Exploit: Focus on how to get more production within the existing capacity limitations.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 3. Subordinate: Prevent the materials needed next from waiting in a queue at a non-constraint
resource.

4. Elevate: If, after fully exploiting this process, it still cannot produce enough products to meet
market demand, find other ways to increase capacity.

5. Go back to Step 1. 8

This methodology can be compared to the steps of the scientific method that System Dynamics follows.
However, System Dynamics aided by special software, provides a means to quickly test the affects on the
system when these changes are made in order to verify their effectiveness. Evaluating these changes in
real time on a system in operation is much more time consuming and costly. System Dynamics provides
the means to evaluate prospective changes before they are installed. Thus, System Dynamics provides a
much more efficient, and cost effective way of managing constraints by assisting in identification of
constraints and developing ways to manage these constraints through simulation of the system.
SYSTEM OPERATION IMPROVEMENT

So, how does TOC improve system operation? Improvement efforts are focused where they will have the
greatest immediate impact on the performance of the system. TOC provides a reliable process that insures
implementation. Stella modeling enables identification of system constraints and exploration of
alternative means of managing these constraints. The first step is to identify the system constraints that
prevent the system from achieving its goal. Next, identify the most constrained resource, called the
Critically Constrained Resource (CCR).
Traditional cost accounting methods demand that each resource is fully loaded to achieve
maximum utilization. However, this causes serious problems due to the build-up of in-process inventory
and wasted machine time. TOC takes a different view of this situation as illustrated by a simple
manufacturing line (figure 1). Or, in order to smooth out production, managers may opt to balance the
line to the critically constrained resource (figure 2).
Raw material flows from left to right through five processes and become finished goods. Note
that process C is the limiting process in that it can only produce 5 parts per day. 9

Raw Finished
Material Goods

A B C D E

11 Parts 9 Parts 5 Parts 8 Parts 15 Parts

per Day per Day per Day per Day per Day
Figure 1, A Simple Manufacturing Line

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Many managers attempt to avoid waste and make everyone efficient by balancing the line. This
balanced line won’t work. It can rarely achieve five parts per day. However, in reality, every process has
some variation. Consider a 50% chance per machine of producing five items per day or more. The
probability two independent machines produce five or more the same day is 0.5*0.5=0.25 or 25% (figure
3). Most processes have more than five steps. And with as few as five steps, there is only a 3% chance
of producing five or more per day.10

Raw Finished
Material Goods

A B C D E

5 Parts 5 Parts 5 Parts 5 Parts 5 Parts

per Day per Day per Day per Day per Day
Figure 2, A “Balanced” Manufacturing Line

Raw Finished
Material Goods

A B C D E

5±2 5±2 5±2 5±2 5±2

Parts Parts Parts Parts Parts
50% 50% 50% 50% 50%
5+ 5+ 5+ 5+ 5+

Figure 3, Problems With a Balanced Manufacturing Line

Note that with a probability of achieving 5 parts per day of only 50%, the cumulative probability
of achieving 5 parts per day for the overall system is 0.5 x 0.5 x 0.5 x 0.5 x 0.5 = .03125, or only about
3%. Thus, there is a high probability, ~97% that the average throughput will be less than 5 parts per day.
The theory of constraints addresses this delemia through the application of the “drum-buffer-rope”
concept of buffer management.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 A proven approach to managing production through the constraint is known as "Drum-Buffer-
Rope" and "Buffer Management."11 TOC encourages every process to work as fast as possible, when
there is work. The constraint is like a drum that beats the cadence of the plant. A time buffer of material
placed between the Raw Material (RM) and the constraint protects it from starvation. A rope throttles the
RM to maintain Work In Process (WIP) at minimum levels (Figure 4). Work completes at predictable

ROPE
Raw DRUM Finished
Material Goods
BUFFER

A B C D E

11 Parts 9 Parts 5 Parts 8 Parts 15 Parts

per Day per Day per Day per Day per Day

flow times.

Figure 4, TOC Process Control

In a production environment, the plant's constraint (bottleneck) must be the driving factor in how it is
managed. In production, the productivity of the constraint is the productivity of the entire plant.
• Drum - The constraint(s), linked to market demand, is the drumbeat for the entire plant.

• Buffer - Time/inventory that ensures that the constraint(s) is protected from disturbances
occurring in the system.

• Rope - Material release is "tied" to the rate of the constraint(s).

The drum, buffer, and rope provide feedback to the production manager for building a production
schedule that is highly immune to disruption, avoids creating excess inventory, and uses small batches to
minimize overall lead time. However, even with "Drum-Buffer-Rope," occasionally disruptions occur
which require special attention. "Buffer Management" is used to mitigate and often prevent those
disruptions.12
Implementations of "Drum-Buffer-Rope" and "Buffer Management" typically result in "lean," low-
inventory production operations capable of consistently 95% (or better) on-time delivery, lead-time
reduction of 35-50%, and inventory reduction of 50%, as well as significantly reduced need for
expediting and rescheduling.13

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 STELLA MODEL OF BUFFER MANAGEMENT

The important distinction between TOC and System Dynamics is that, although TOC involves
systems thinking it does not provide a way to simulate the system under study to test whether the
recommendations developed under TOC will be effective. This is where System Dynamics can aid in
making the TOC concepts more effective, particularly when a SD software package such as Stella is used
to aid with the analysis.

The building blocks of Stella models are "stocks" and "flows."14 Stocks act as "accumulators" of
system objects, and flows act as the transport mechanism between stocks. Converters are System
Dynamics elements that pass information to flows or from stocks to flows. They can also pass
information from one stock to another. A simple example of a stock and a flow is that of a bank balance
as shown in figure 5.

Flow Stock
interest bank balance
added

Converter
interest Connector
rate

Figure 5, A Simple Stock – Flow Stella Model of a Bank Balance

In this example, the bank balance is increased by the flow from "interest added." The bank
balance multiplied by the interest rate controls the amount of "interest added". Therefore, the greater the
interest rate, the more interest is added to the bank balance. A graph depicting this simulation is shown in
figure 6.
1: bank balance 2: interest added
1: 50000
2:

1: 25000
2:

2
1
1:
2: 0 1 2 1 2 1 2
0.00 3.00 6.00 9.00 12.00
Page 1 Months 9:17 PM Tue, Dec 17, 2002
Untitled

Figure 6, Graph of Interest Added to Bank Balance

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 In the Stella model of a simple production system, the process stocks are a special type of stock
called a conveyor. Conveyors are a special version of the stock variable that simulates material moving
through the system in a tightly controlled pattern.15 Work-in-process (WIP) inventories are ordinary
stocks between each operation. Connectors (red arrows) provide feedback to the WIP that tell each flow
between the WIP stock and the next conveyor operation not to send any more material to the next
operation than it can handle. Logic is provided to each flow to limit the amount of material from going to
the conveyor to be processed to no more than is required. This stops any flow from the WIP stock if the
process is full. Figure 7 is a Stella model of the simple production system that consists of a series of
conveyors connected by flows with WIP stocks between operations.

A Stella model of the simple production system would appear like the model in figure 7. The
time to process parts through each conveyor is based on an 8 hr day. Therefore, since A can process 11
parts per day, the process time is 8 hrs / 11 parts, or .72 hrs for one part to be processed. This sets the
transit time for conveyor A at .72 hrs, or about 44 minutes. The Stella software rounds fractions, so, the
time units for this model are best handled in minutes. Likewise, B can process 9 parts per day, or a
transit time of 53 minutes, C is 96 minutes, D is 60 minutes, and E is 32 minutes.

WIP A Processing
WIP B Processing
C To D Processing
B WIP C To C WIP D WIP E
D To E tot time
To B

Raw deliv ering

Material Raw Material A Finished
Goods

To Processing
A A to FG
C
B D E

Orders

ordering

number of
orders

Figure 7, A Stella Model of a Simple Production System

Running this simulation yields some interesting information about the model. Figure 8 shows a
graphical output of the time it takes to produce 5 units. Figure 9 shows that it takes 1,275 minutes, or,
21.25 hours to achieve the Finished Goods total equal to the number of orders if the number of orders are
set to 5.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 1: Orders 2: Finished Goods 3: number of orders
1: 5 1 3 1 3 1 3 1 2 3 1 2 3
2: 2
3:

2
1:
2: 3
3:

1:
2:
3: 0 2
1.00 400.80 800.60 1200.40 1600.20 2000.00
Page 2 Time 7:46 PM Wed, Jan 29, 2003
Untitled

Figure 8, Time to Achieve Finished Goods of 5 equal to 1,275 Minutes, or 21.25 hrs

Figure 9 is a graph of the time it takes to produce 5 units of finished goods.

Figure 9, Graph of Time to Produce 5 units of Finished Goods, Time = 1,275 min

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 The drum-buffer-rope (DBR) concept of TOC states that feedback is provided from the critically
constrained resource (CCR) to the Raw Material delivery source. A time buffer of material placed
between the Raw Material (RM) and the constraint protects it from starvation. A rope throttles the RM to
maintain Work In Process (WIP) at minimum levels of the CCR (Figure 4). Work completes at
predictable flow times. This concept is represented in the Stella model shown in figure 10. Note a
feedback connector represents the “Rope” from C, the CCR, to WIP A. This has the effect of limiting the
flow from the processes prior to C to the same level of C. Next, a buffer is added to Process B so that C
is protected from starvation, yet, A and B are not producing more than C. A feedback loop to ordering
tells the system how many more units are needed to keep C from starving. It would seem that this logic
would cause the model to take longer to produce the 5 units of output. However, in reality, the process is
faster. Now, instead of 1,275.5 minutes to produce 5 units of output, it only takes 910.25 minutes (Figure
11).
Rope

Processing
WIP A WIP B Processing
C To D Processing
B WIP C To C WIP D WIP E
D To E tot time
To B

Raw A Finished
deliv ering
Material Goods
Raw Material

To Processing
A A Feed to to FG
C
B Extra WIP C D E
ordering
Material to
Orders
Buf f er

BUFFER

additional
f raction to
number of WIP C
orders

Figure 10, Stella Model With Feedback Loops and Buffer

Figure 11, Partial Output from Stella Model Without Buffer

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Figure 11 shows the tabular output from the model. This shows Finished Goods of 5 units being
completed at 910.25 minutes. A graph of this output is shown in Figure 12. Note the steeper slope on
the curve and earlier completion of finished goods. Also, note that the number of orders is increased by 1
which accounts for the additional inventory needed to keep process C from starving.

1: Orders 2: Finished Goods 3: number of orders

1: 6 3 1 3 1 3 1 3 1 3
2:
3: 5 2 2
1 2

1: 3
2:
3: 3

1:
2: 0
3: 0 2
1.00 400.80 800.60 1200.40 1600.20 2000.00
Page 2 Time 9:17 PM Wed, Jan 29, 2003
Untitled

Figure 12, Graph of Time to Produce 5 units of Finished Goods with TOC in place

Further confirmation of the buffer effect on the system can be seen in figure 13 and 14. These
graphs compare the output of each process without and with TOC applied. Note the differences in each
process. The processes with TOC applied have a more controlled, gradual slope. The spikes of
production at the beginning of each of the process are leveled out over the process. This shows that the
processes are more in control. This information, coupled with the graph in figure 12 and the numerical
output in figure 11 confirm that the DBR method does control system throughput resulting in a much
more even flow through the system. This in turn improves throughput by allowing every process to work
as fast as possible, when there is work, maintaining WIP at minimum levels, and protecting the
constrained resource from starvation. Furthermore, it confirms that the Stella model is working properly
and effectively models the DBR concept.
Figure 14 shows the various WIP levels and compares them without TOC and with TOC applied.
Note the significantly higher WIP level in WIP A with TOC Applied than without TOC applied. Recall
that process A could produce 11 parts per day. Thus, without TOC applied, its WIP would be about 50%
that of process C which has a throughput of 5 parts per day due to the higher production level. This
shows that the Stella model is working as expected by increasing the amount of material released to
process A which in turn improves throughput by allowing every process to work as fast as possible.
Also, other WIP levels are maintained at minimum levels. Note that WIP B, C, and D are somewhat
lower when TOC is applied. WIP E appears to be about equal. However, in each case except for perhaps
A, the slope of inventory building is less dramatic and the initial spikes are gone. This indicates the
process is more in control.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Without TOC Applied With TOC Applied
1: A 1: A
1: 3 1: 3
1

Process A Process A
1: 2 1: 2

1
1 1 1 1 1 1 1 1
1: 0 1: 0
1.00 400.80 800.60 1200.40 1600.20 2000.00 1.00 400.80 800.60 1200.40 1600.20 2000.00
Page 3 Time 7:46 PM Wed, Jan 29, 2003 Page 3 Time 9:17 PM Wed, Jan 29, 2003
Untitled Untitled

1: B 1: B
1: 3 1: 3

Process B Process B
1: 2 1: 2

1 1 1 1 1 1 1 1 1
1: 0 1: 0
1.00 400.80 800.60 1200.40 1600.20 2000.00 1.00 400.80 800.60 1200.40 1600.20 2000.00
Page 4 Time 7:46 PM Wed, Jan 29, 2003 Page 4 Time 9:17 PM Wed, Jan 29, 2003
Untitled Untitled

1: C 1: C
1: 3 1: 3

Process C Process C
1: 2 1: 2

1
1
1: 0 1 1 1 1: 0 1 1 1
1.00 400.80 800.60 1200.40 1600.20 2000.00 1.00 400.80 800.60 1200.40 1600.20 2000.00
Page 5 Time 7:46 PM Wed, Jan 29, 2003 Page 5 Time 9:17 PM Wed, Jan 29, 2003
Untitled Untitled

1: D 1: D
1: 3 1: 3

Process D Process D
1: 2 1: 2

1 1

1
1 1 1 1 1
1: 0 1 1: 0 1
1.00 400.80 800.60 1200.40 1600.20 2000.00 1.00 400.80 800.60 1200.40 1600.20 2000.00
Page 6 Time 7:46 PM Wed, Jan 29, 2003 Page 6 Time 9:17 PM Wed, Jan 29, 2003
Untitled Untitled

1: E 1: E
1: 3 1: 3

Process E Process E
1: 2 1: 2

1
1
1 1
1: 0 1 1 1 1: 0 1 1 1
1.00 400.80 800.60 1200.40 1600.20 2000.00 1.00 400.80 800.60 1200.40 1600.20 2000.00
Page 7 Time 7:46 PM Wed, Jan 29, 2003 Page 7 Time 9:17 PM Wed, Jan 29, 2003
Untitled Untitled

Figure 13, Output Graph Material Flow Patterns

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Figure 14 shows the effect TOC has on work-in-process inventories.

Without TOC Applied With TOC Applied

1: WIP A 1: WIP A
1: 5 1: 5
1

WIP A WIP A
1: 3 1: 3
1

1: 0 1 1 1 1: 0 1 1 1
1.00 500.75 1000.50 1500.25 2000.00 1.00 500.75 1000.50 1500.25 2000.00
Page 9 Time 10:15 AM Thu, Jan 30, 2003 Page 9 Time 9:17 PM Wed, Jan 29, 2003
1: WIP B 1: WIP B
Untitled Untitled
1: 3 1: 3

WIP B WIP B
1: 2 1: 2

1
1: 0 1 1 1 1 1: 0 1 1 1
1.00 500.75 1000.50 1500.25 2000.00 1.00 500.75 1000.50 1500.25 2000.00
Page 10 Time 10:15 AM Thu, Jan 30, 2003 Page 10 Time 9:17 PM Wed, Jan 29, 2003
1: WIP C 1: WIP C
Untitled Untitled
1: 2 1: 2

WIP C WIP C
1: 1 1: 1

1: 0 1 1 1 1: 0 1 1 1
1.00 500.75 1000.50 1500.25 2000.00 1.00 500.75 1000.50 1500.25 2000.00
Page 11 Time 10:15 AM Thu, Jan 30, 2003 Page 11 Time 9:17 PM Wed, Jan 29, 2003
1: WIP D 1: WIP D
Untitled Untitled
1: 2 1: 2

WIP D WIP D
1: 1 1: 1

1 1 1 1
1: 0 1 1: 0 1
1.00 500.75 1000.50 1500.25 2000.00 1.00 500.75 1000.50 1500.25 2000.00
Page 12 Time 10:15 AM Thu, Jan 30, 2003 Page 12 Time 9:17 PM Wed, Jan 29, 2003
1: WIP E 1: WIP E
Untitled Untitled
1: 1 1: 1

WIP E WIP E
1: 0 1: 0 1

1
1: 0 1 1 1 1: 0 1 1
1.00 500.75 1000.50 1500.25 2000.00 1.00 500.75 1000.50 1500.25 2000.00
Page 13 Time 10:15 AM Thu, Jan 30, 2003 Page 13 Time 9:17 PM Wed, Jan 29, 2003
Untitled Untitled

Figure 14, Effect of TOC on WIP

Note the initial increase in WIP A in comparison of the two models. This is due to the extra WIP
necessary to keep C from starving. Also, the initial order quantity in figure 12 is for 6, whereas, the
finished goods quantity ends up at 5. This is showing where the extra WIP in A is coming from.
However, WIP at B is somewhat lower and at C WIP is significantly lower in the TOC model than

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
without TOC applied. Finally, WIP D and E are at about the same levels, but, there is indication the
process is more in control since the initial slopes are more gradual.
It is the conclusion of this study that TOC does improve throughput of a production system.
These improvements come from overall reductions in work-in-process inventories and smother flow
through the system.
CASE STUDY EXAMPLE: CASE COIL COMPANY

Next, a more complicated case study is presented to further illustrate the concept of Theory of
Constraints and how System Dynamics can improve the process. This case study is that of a hypothetical
company called Case Coil developed by John Tripp of TOC Scotland.16 It is used to illustrate the various
aspects of the Theory of Constraints. The traditional approach to analyzing this case study is to analyze
the given data using the TOC process. However, the application System Dynamics can provide some
valuable insights into the system constraints by providing feedback on testing of various alternatives that
would otherwise take weeks, or months to accomplish using a real system.
The hypothetical Case Coil Company manufactures commodity copper coils by rolling copper
billets into wide rolls, and slitting them into different widths to form coils. The copper coils are available
in five thicknesses, five widths, and four hardness levels. There are seven processing steps for all
products. All raw material is supplied by the smelter in billets weighing one and one half metric tons.
Annealing softens the billets for the rolling process. A degrease operation is next. This is a two-stage
process that uses an acid clean followed by a neutralizing rinse. Heat treat is the next process. This
requires clean material or the process may result in problems. Next is metallurgical testing where both
the gauge and hardness are tested. Typically, there is a 15% failure rate at this process. The next
operation is slitting which cuts the original roll into various widths and trims the outside edges. Slitting
has a 10% failure rate. Each coil is then packed, wrapped with vapor-retardant material and tagged for
identification purposes. Finally, shipping allocates packaged coils to specific orders, palletizes, and ships
the orders to waiting customers. Shipping also inventories the remaining coils.17 Figure 15 is a
breakdown of the throughput for each of these operations.

CASE COIL THROUGHPUT

Total
QUEUE hrs/Part Available
STEP PROCESS SETUP LOAD TIME / # OF w/Queue mach Potential
NUMBER DESCRIPTION/ACTIVITY TIME/Hr TIME TIME parts/hr hrs MACH Time # SHIFTS hrs/wk. Parts/wk COMMENT

10 DRAW FROM RAW MATL STORES - - -

20 ANNEAL TO SPECIFICATIONS 0.200 5.00 11.00 2 11.20 2 160 14.29

30 ROLL 0.303 3.31 15.00 1 15.30 2 80 5.23 CCR

40 DEGREASE 0.640 1.56 15.00 3 15.64 2 240 15.35

50 HEAT TREAT TO GAUGE SPECS 0.745 0.18 0.17 1.34 6.00 1 6.75 3 120 17.79

60 METALLURGICAL TEST 1.447 0.69 2.00 6 3.45 2 480 139.26

70 SLIT TO REQUIRED WIDTH 0.837 1.20 2.00 4 2.84 2 320 112.81

80 PACKAGE AND STORE 0.078 12.77 2.00 5.5 2.08 2 440 211.71

Figure 15, Case Coil Throughput Table

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 From this table, it can be seen that the rolling operation is the critically constrained resource
(CCR) because its potential parts per week is only 5.23 compared to 14.29 for Anneal, 15.35 for
Degrease, 17.79 for Heat Treat, 139.26 for Metallurgical Test, 112.81 for Slit to Required Width, and
211.71 parts per week for Package and Store. Using System Dynamics and the lessons learned in the
Simple Production System a SD model was constructed to depict the throughput of Case Coil. Special
conveyor stocks, called arrays, are used for multiple machine operations. The process time to produce
one unit is used as the transverse time in each conveyor. Additional controls are added to this SD model
to control failure rates, queue times, and possibly other parameters that can be manipulated to improve
system performance.

In this example, controls have been added to control queue time and failure rates. Converters are
added so queue time can be added and controlled between each operation. A queue time reduction
converter is added to control the percent of queue time allowed to affect the system. For simplicity, only
a simple percentage applied to all queue time converters is allowed. All of these features allow testing of
various parameters to see what impacts and improvements can be made on the system. Time is measured
in minutes in this model due to the large disparity between process times and queue time. In order to
input the correct process times, minutes are required.

The layout of this model is shown in figure 16. Running the model establishes a base throughput
time of 19,928 min., or 320.5 hrs to produce 20 units. Output from the model is shown graphically in
figure 17. Tabular output is shown in figure 18.
slitting Met Test Anneal
Rolling
failed Failed WIP
WIP

Raw Annealing
Material Process Rolling Degrease
to degrease
Process WIP
WIP

sending sending to
CCR
to annealing 1 filling rolling sending to
annealing sending to
WIP degrease
Rolling
Orders
queue
Anneal
time reduction Rolling Degrease
Queue
factor Queue
queue Time
Ordering Time
time reduction
factor
to
Number of Orders Met Test Heat Treat Heat Treat
Met Test Met Test sending to to heat
Met Test WIP Process WIP
Queue WIP HT treat WIP
Time
queue
time reduction building sending to
factor Slitting met test
WIP Met Test met test
failure queue Heat Treat
Failed
rate time reduction Queue Time
Package & Degrese
sending to factor Pack & Store
Store Queue
pack & atore WIP queue
sending Time
Slitting WIP time reduction
Slitting to slitting
WIP factor
sending to Pack & Store
package & Queue Time
store
completing
slitting Slitting orders
failed Queue Time
Cycle Time

slitting Completed Orders 20.0

queue
failure rate
time reduction
factor

Completed Orders

Figure 16, Stella model of Case Coil Company.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 1: Completed Orders 2: Number of Orders 3: Orders
1: 20 2 3 2 3 2 3 2 3 1 2 3
2: 1
3:

1:
2: 10
3:

1:
2:
3: 0 1
0.00 4800.00 9600.00 14400.00 19200.00 24000.00
Page 10 Minutes 10:55 PM Sun, Feb 02, 2003
Untitled

Figure 17, Model Output of Case Coil Company Baseline Case

Figure 18, Case Coil Base Case Partial Output – 19,928 min., or 332.1 hrs. for 20 units
Recalling the 5 TOC rules we start with the base case (figure 16), then, go to rule number 2 and
focus on how to get more production within the existing capacity limitations. First a drum, buffer, rope is
added to the system just as in the simpler example of a five step production system. This time the buffer
is added between the Annealing operations and Rolling operation since Rolling is the critically
constrained resource. The feedback to ordering and from rolling and sending to annealing to simulate the
Rope in the DBR scenario is also added. Figure 19 shows the Stella model of this production system.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 queue
Anneal time reduction
slitting Met Test
factor Rolling
failed Failed WIP
WIP

Raw Annealing Anneal

Material Process Rolling Degrease
Queue
Process Rolling WIP
Time

Fraction to
sending CCR
Buffer sending to Annealing
to annealing 1 sending to sending to
annealing
Rolling degrease
Material to
Buffer Buffer feed
to Rolling WIP Rolling Degrease
Queue
Ordering queue
Additional Fraction Time
Orders time reduction
for Rolling Buffer
Buffer factor
to
Number of Orders Met Test Heat Treat Heat Treat
Met Test Met Test sending to to heat
Met Test WIP Process WIP
Queue WIP HT treat WIP
Time
queue
time reduction building sending to
factor met test failure rate
Slitting
met test reduction
WIP Met Test
failure queue Heat Treat
Failed
rate time reduction Queue Time
Package & Degrese
sending to factor Pack & Store
Store Queue
pack & atore WIP queue
sending Time
Slitting WIP time reduction
Slitting to slitting
WIP factor
sending to Pack & Store
package & Queue Time
store
completing
slitting Slitting orders
failed Queue Time
Cycle Time

failure rate slitting Completed Orders 20.80

reduction queue
failure rate
time reduction
factor

Completed Orders

Figure 19, Case Coil – Stella Model with DBR Added

Running this revised model yields a throughput time of 17,699 min., or 294.0 hrs. This is an
improvement of about 11% by simply adding the buffer and feedback loops. The amount of material sent
to the buffer is set at 10%. This can also be varied to see what impact it has on the system. Output from
this model is shown in figures 20 and 21. Note that the orders exceed the Number of Orders and
Completed orders by 6 units. This is due to the feedback loop requesting additional material for the
buffer.
1: Completed Orders 2: Number of Orders 3: Orders
1: 30
2:
3:
3 3 3
3
3

2 2 2 2 1 2
1

1:
2: 15
3:
1

1
1:
2:
3: 0 1
0.00 4800.00 9600.00 14400.00 19200.00 24000.00
Page 10 Minutes 3:26 PM Tue, Feb 04, 2003
Untitled

Figure 20, Case Coil DBR Graphical Output

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Figure 21, Case Coil DBR Partial Tabular Output – 18,340 min., or 305.7 hrs. for 20 units

Next, the failure rates for metallurgical test and slitting were reduced by 50%. This yields a
process time of 11,444 min., or 190.7 hours. This is an incremental improvement of 37.6% in cycle time
to produce 20 units. Graphical and partial tabular output from this run of the model is shown in figures 22
and 23.

1: Completed Orders 2: Number of Orders 3: Orders

1: 30
2:
3:
3 3 3
3
3
1 1
2 2 2 2 2

1: 1
2: 15
3:

1:
2:
3: 0 1
0.00 4800.00 9600.00 14400.00 19200.00 24000.00
Page 10 Minutes 10:05 AM Mon, Feb 03, 2003
Untitled

Figure 22, Reduce Failure Rate by 50% - 190.7 hours to produce 20 units

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Figure 23, Tabular output, Reduce Failure Rate by 50% - 11,444 min., or 190.7 hours for 20 units
The next example looks at reducing the overall queue time in half. Note the failure rates were left
at their original level of 15% at metallurgical test and 10% at slitting. This run yields an overall
production time of 8,678 minutes, or 144.6 hours as shown in figure 24. This is a 35% reduction in cycle
time.

1: Completed Orders 2: Number of Orders 3: Orders

1: 30
2:
3:
3 3 3 3
3

2 2 1 2 1 2 1 2

1:
2: 15
3:
1

1:
2:
3: 0 1
0.00 4800.00 9600.00 14400.00 19200.00 24000.00
Page 10 Minutes 10:21 AM Mon, Feb 03, 2003
Untitled

Figure 24, Reduce Queue Time by 50% - 144.6 hours to produce 20 units

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Figure 25, Tabular Output Reduce Queue Time by 50% - 144.6 hours to produce 20 units

Next, reducing both failure rates and queue time by 50% resulted in 100.7 hrs to produce 20 units as
shown in figures 26 and 27. This is an overall reduction of 43.5% in cycle time with the DBR in effect
and an incremental improvement of 49%.
1: Completed Orders 2: Number of Orders 3: Orders
1: 30
2:
3:
3 3 3 3
3

1 1 1
2 2 2 2 2

1: 1
2: 15
3:

1:
2:
3: 0 1
0.00 4800.00 9600.00 14400.00 19200.00 24000.00
Page 10 Minutes 10:37 AM Mon, Feb 03, 2003
Untitled

Figure 26, Case Coil - Reduce Queue Time and

Failure Rates by 50% - 6,040 min., or 100.7 hours to produce 20 units

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Figure 27, Case Coil Tabular Output - Reduce Queue Time and
Failure Rates by ½ - 6,040 min., or 100.7 hours to produce 20 units

Now, let’s assume that somehow Case Coil was able to obtain 2 additional rolling machines and
somehow reduce anneal queue time by 5%. Leaving the failure rates and other queue times at their
original values would eliminate the bottleneck at anneal by making the throughput 15.68 parts / wk. This
moves the bottleneck to the heat treat process, so, a buffer is added between degrease and heat treat to
control the rate material flows to the heat treat process. Also, feedback from the degrease process to
sending to degrease is added to limit the demand for material at degrease to no more than the degrease

CASE COIL THROUGHPUT

Total
QUEUE hrs/Part Available
STEP PROCESS SETUP LOAD TIME / # OF w/Queue mach Potential
NUMBER DESCRIPTION/ACTIVITY TIME/Hr TIME TIME parts/hr hrs MACH Time # SHIFTS hrs/wk. Parts/wk COMMENT

10 DRAW FROM RAW MATL STORES - - -

20 ANNEAL TO SPECIFICATIONS 0.200 5.00 10.45 2 10.65 2 160 15.02

30 ROLL 0.303 3.31 15.00 3 15.30 2 240 15.68

40 DEGREASE 0.640 1.56 15.00 3 15.64 2 240 15.35

50 HEAT TREAT TO GAUGE SPECS 0.745 0.18 0.17 1.34 6.00 1 6.75 3 120 17.79 CCR

60 METALLURGICAL TEST 1.447 0.69 2.00 6 3.45 2 480 139.26

70 SLIT TO REQUIRED WIDTH 0.837 1.20 2.00 4 2.84 2 320 112.81

80 PACKAGE AND STORE 0.078 12.77 2.00 5.5 2.08 2 440 211.71

Figure 28, Revised Case Coil Throughput Data

WIP. Finally, feedback from the buffer to ordering through a converter called "added fraction to DBR
buffer." Figure 28 shows the new values and Figure 29 the new Stella model for this situation. Running
the model again we get the graph shown in figure 30.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 anneal queue
slitting Met Test Anneal time reduction factor
failed Failed WIP queue
time reduction
factor Rolling
Raw Annealing WIP
Anneal
Material Process Rolling Degrease sending to
Queue
Process Rolling WIP degrease
Time

sending sending to Annealing

to anneal WIP annealing sending to Fraction to
Rolling Buffer
Buffer

queue
time reduction Rolling CCR
Buffer
factor Queue
Ordering Heat Treat queue
Additional Fraction Time
Orders time reduction Degrease
for Degrease Buffer Queue Time material to
factor buffer
Number of Orders Met Test Met Test Heat Treat sending to fraction to
Queue Met Test WIP Process
heat treating HT heat treat WIP degreasing
Time
queue
time reduction met testing sending to
factor met test Heat Treat
WIP
Met Test
met test
Failedfailure met testing
rate
failure rate Package & Degrese
sending to
reduction Store Queue
sending pack & atore queue Time
Slitting WIP time reduction
Slitting to slitting
WIP factor
sending to Pack & Store
Pack & Store
package & Queue Time
WIP
store
completing
slitting Slitting orders
failed Queue Time
Cycle Time

failure rate slitting Completed Orders 20.38

reduction queue
failure rate
time reduction
factor

Completed Orders

Figure 29, Stella Model for Revised Throughput Scenario

The results of these changes without changing failure rates or queue times makes a significant
improvement in the overall throughput from the base case of 320.5 hours to 265.7 hours for 20 units as
shown in figures 30 and 31.
1: Completed Orders 2: Number of Orders 3: Orders
1: 30
2:
3:
3 3
3
3

3
2 2 2 2 1 2

1: 1
2: 15
3:

1
1:
2:
3: 0 1
0.00 3840.00 7680.00 11520.00 15360.00 19200.00
Page 10 Minutes 10:17 PM Wed, Feb 05, 2003
Untitled

Figure 30, Added Machine Throughput Of 265.7 Hrs For 20 Units

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Figure 31, Improved “Base Case” Tabular Output of Throughput of 15,942 min, or 265.7 hrs for 20 units

From these results it can be concluded that the Drum-Buffer-Rope concept of the Theory of
Constraints does make significant improvements in system throughput. In addition, Stella modeling tools
provide a powerful methodology to evaluate alternatives in system structure to assist in the decision
process of selecting the best alternative.

DISCUSSION

The overall objective of this study was two fold: 1) use SD software to demonstrate how the TOC
principles, in particular the Drum-Buffer-Rope concept can improve throughput of a production system;
and 2) demonstrate how SD software, specifically Stella, can further enhance this process by adding the
ability quickly evaluate various alternatives versus testing these alternatives on live production runs that
may take weeks, or months. Following the 6 steps identified by Jay Forester for SD simulation, it was
possible to show that the DBR concept of TOC is a valid approach to improving production throughput.
The SD software made it possible to observe the effects of testing various hypotheses that were
anticipated to improve performance, and arrive at a quantitative measure of the effects of these changes.
The modeling of a simple 5-step process was designed to demonstrate the effectiveness of the
DBR concept. This test resulted in a 28% reduction in the cycle time for this system. Lessons learned in
the construction of this model were then applied to the more complex model of the hypothetical Case
Coil example. The first test of this model was to show again that the DBR concept also worked to
improve throughput of this model. The base throughput time for the Case Coil system was 332.1 hrs for
20 units. Applying the DBR concept to the Critically Constrained Resources that yielded a throughput
time of 305.7 hrs, or a 7.9% reduction in cycle time. Next, overall failure rates in slitting and
metallurgical test were reduced by 50%. This resulted in a cycle time of 190.7 hrs for 20 units, or an
incremental decrease in cycle time of 37.6%. The next test was to reduce the overall queue time by 50%.
This yielded another reduction of 46.1 minutes, to 144.6 hours.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 Finally, following the TOC concept of exploiting the constrained process, an additional 2 rolling
machines were added to the system and the anneal queue time was reduced by 5% moving the bottleneck
to the Heat Treat process. Then, modifying the Stella model so that the DBR was at the degrease
operation resulted in an overall cycle time of 265.7 hours. This was a 54.8 hr, or, 17% improvement in
cycle time from the original base case of 320.5 hrs. From these results was concluded that the Drum-
Buffer-Rope concept of the Theory of Constraints does make significant improvements in system
throughput. In addition, Stella modeling tools provide a powerful methodology to evaluate alternatives
in system structure to assist in the decision process of selecting the best alternative.
This project has demonstrated what it set out to accomplish. That is that by providing a way to
model and simulate the system under study through System Dynamics and applying the rules of TOC to
the model, alternative solutions to improve system operation can be quickly developed and analyzed.
System Dynamics software further enhances this process by adding the ability quickly evaluate various
alternatives versus testing these alternatives on live production runs that may otherwise take weeks, or
months to perform. Had this been a real production situation, evaluation of these various alternatives
would have taken several months to accomplish without the use of SD software. Whereas, a skilled
Stella modeler who is familiar with the production system under study can construct a model that
represents the production system in a day, or two and run various scenarios the third day. Actual model
simulation times will vary with the CPU speed of the computer the simulation is run on.
Development of the various modeling scenarios for this study required an average of
approximately 10 hours per week over the period of 20 weeks for a total of 200 hours to develop and test
the model scenarios. Significant research went into determining the correct application of conveyors,
arrays, and other modeling elements during this time. The model was prepared on a 233 mHz Pentium
computer using the Stella 7.0 software. In the course of this study it was discovered that the iThink
Software, also by High Performance Systems, would have also worked and had more examples that were
oriented toward the type of problem under study. I was able to obtain a trial copy of iThink and convert
these examples to work on Stella. Student price for both software packages is about $130.00. Proof of
student status is required when ordering the student software. The professional price is about $1,100.00.
There are software packages for either Windows, or Macintosh computer platforms. Actual runtime on
the 233 mHz Pentium computer was about 15 minutes for the complete Stella model with the DBR
added.
From a software features perspective, the iThink and Stella programs are virtually identical. Each
is designed to facilitate the mapping, modeling and simulating of dynamic processes. The major
difference between the two products is in the supporting documentation. The iThink software is targeted
at business users. Applications and sample models cover the gamut of business uses including Business
Process Re-engineering, Strategic Planning, Financial Analysis, Manufacturing, Balanced Scorecard,
Systems Thinking, and Organizational Learning. 18
Stella, on the other hand, is targeted at educators and researchers. Its documentation and sample
models span the curriculum from Literature to Physics, Mathematics to History, and pretty much
everything in between. iThink is the best choice for working with business-oriented issues. Stella is the
best match for an educational or scientific research setting. However, Stella models can be opened using
iThink, and vice versa. 19

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 CONCLUSION

The major conclusions that can be drawn from this analysis are that System Dynamics modeling
enables application of the Theory of Constraints DBR process and exploration of alternative means of
managing system constraints. This study demonstrates that while the TOC Thinking Process represents a
well-structured approach to understanding how to deal with constraints in a production system, Systems
Dynamics modeling provides a supplemental understanding relative to knowledge gained through the
TOC Thinking Process. In addition, System Dynamics and the use of SD software can be a valuable tool
in communicating with management and workers the concept of TOC and how constraints affect the
system they are dealing with. In the same way that The Theory of Constraints (TOC) is based on Eli
Goldratt's work on "how to think"20.
An inherent shortcoming in using the TOC Thinking Process is that it lacks the robust capability
to fully capture the dynamic complexity of even a simple production system.21 System Dynamics can
overcome this shortcoming by adding the capability of experimenting with the dynamic complexity of
these systems through modeling. SD modeling is also based on a way of thinking about systems from a
global perspective. A primary application of TOC embodies a system thinking approach to
manufacturing systems. By knowing how to think from a systems perspective, we can better understand
these systems. Through better understanding, we can improve the performance of the system. Testing of
TOC alternatives using System Dynamics software also provides a valuable tool to understand how
system performance can be improved.

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727
 REFERENCES

1
Web Site: http://www.rogo.com/cac/whatisTOC.html, “What is TOC?“
2
Sullivan, Timothy T., CIRAS/Iowa State University Extension Website,
http://www.ciras.iastate.edu/toc/TOCintroductionWWW/index.htm, TOC Constraints Management Presentation, 8/19/99
3
Rogo.com
4
Web Site: http://www.albany.edu/cpr/System Dynamicss/, “What is System Dynamics?”
5
Rogo.com
6
Goldratt, Eliyahu M. and Cox, Jeff, The Goal, Second Revised Edition, Croton-on-Hudson, N.Y.: North River Press, 1992.
7
Web Site: System Dynamics Society, "What is System Dynamics?" , http://www.albany.edu/cpr/sds/
8
Goldratt, Eliyahu M. and Cox, Jeff, The Goal, Second Revised Edition, Croton-on-Hudson, N.Y.: North River Press, 1992.
9
Holt, James R., Ph.D., PE, Associate Professor Engineering Management, Washington State University, EM 526 Constraints
Management, Week 1 Presentation, 1998
10
Holt, James R., Ph.D., PE
11
Holt, James R., Ph.D., PE
12
Holt, James R., Ph.D., PE
13
Holt, James R., Ph.D., PE
14
Ford, Andrew, Modeling the Environment, An Introduction to Stella models of Environmental Systems, Washington, D.C.,
Island Press, 1999 p. 14
15
Ford, Andrew, p 108.
16
Tripp, John, http://www.goldratt-toc.com/tocworld/CoilCo/Home.html, 1999
17
Case Coil Company case study, Creative Output, Inc., 1986
18
Web Site: http://www.hps-inc.com/ordering/POBusithink.asp#, High Performance Systems, FAQs, 2003
19
Web Site: http://www.hps-inc.com/ordering/POBusithink.asp#, High Performance Systems, FAQs, 2003
20
Web Site: http://www.rogo.com/cac/whatisTOC.html, “What is TOC?“
21
Reid, Richard A. and Koljonen, Elsa L., Validating A Manufacturing Paradigm: A System Dynamics Modeling Approach,
Proceedings of the 1999 Winter Simulation Conference, p759-765

Prepared for the U.S. Department of Energy under contract No. DE-AC07-99ID13727

Back to the Top