See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228434284

Modeling and Simulating Supply Chain Management

Article · January 2011

CITATIONS READS

5 3,301

3 authors, including:

Abbas Toloie Eshlaghy Maryam Razavi

Islamic Azad University Tehran Science and Research Branch Islamic Azad University, Fars Science Research
146 PUBLICATIONS   591 CITATIONS    1 PUBLICATION   5 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Optimazation Algorithms and Business Intelligence Models View project

Academic Transgression View project

All content following this page was uploaded by Abbas Toloie Eshlaghy on 01 February 2014.

The user has requested enhancement of the downloaded file.

 Applied Mathematical Sciences, Vol. 5, 2011, no. 17, 817 - 828

Modeling and Simulating Supply Chain

Management

Abbas Toloie Eshlaghy

Industrial Management Department

Faculty of Management and Economics
Science and Research Branch
Islamic Azad University, Tehran, Iran
toloie@gmail.com

Maryam Razavi

Industrial Management Department

Faculty of Management and Economics
Science and Research Branch
Islamic Azad University, Tehran, Iran
Mrazavi_ir@yahoo.com

Abstract

Uncertainty in Supply Chain Management can lead to unsuitable services to

customers, but it can be overcome through having an emergency inventory. Due
to high costs of inventory, a general inclination exists towards a minimum
inventory. So, one of the objectives of Supply Chain Management is to coordinate
all activities throughout the chain, so that simultaneously with reducing
commodity inventory, they can be supplied to customers when is needed. In
addition, it is required that negative effects of changes in demand on order rate to
be decreased. If order rate is considered constant, large changes will be seen in the
inventory which will cause increase in inventory costs. Also, if the inventory is
considered is considered constant, it will cause unstable production processes
which will lead to increase in production costs. Existing static models for Supply
Chain Management seem to be insufficient encountering with dynamic
specifications of the supply chain system. The present article uses linear control
method and transfer function analysis to offer a method for modeling Supply
Chain Management in continuous-time condition in order to provide the best
situation for exchanging order rate and inventory. Simulation has been conducted
using Vensim software and results have been described.
818 A. Toloie Eshlaghy and M. Razavi

Keywords: Supply chain management, control, modeling, dynamic systems

Introduction

Supply chain is a network of inventory and distribution factors (producers,

suppliers, distributors and retailers) involving in transforming natural resources,
raw materials and components into a finished product and delivering to the end
customer. The supply chain is specified mainly as a forward process of raw
materials and a backward process of information. Supply chain management is
one of the newly emerge branches of management science and it is developing
day by day. It discusses ways to shorten production processes and delivery times
of supplying finished products to customers, while improving quality of products
and services. To this end, it benefits from cutting-edge breakthroughs in
management and technology sciences. In fact, all countries are inevitable to utilize
supply chain management. Situations which have led to definition and
development of supply chain management are the daily increasing competition
and efforts for the survival of organizations, which have been obtained as a result
of expansion of communication and technology networks.
Enterprises have become recently very interested in improving efficiency of
supply chain management due to rising production and distribution costs,
globalization of markets and demands of customers for diverse products with
short lifecycle. These all boost competition among companies. Uncertainty in all
the supply chain stages has complicated supply chain management process.
Uncertainty can cause halts in a supply chain as wrong predictions, late
information, and transportation problems. Any halt in supply chain can lead to
unsuitable services to customers, i.e. in timely delivery of goods to customers.
Companies overcome the uncertainty usually through holding an emergency
inventory. All the companies maintain the inventory to minimize the effect of
uncertainty and to keep a streamlined, stable and profitable process form suppliers
to customers. But due to high inventory costs, companies are inclined to minimize
the inventory as well.
So, coordinating all different activities across the chain is an important
responsibility of Supply Chain Management, so that goods can be delivered
timely to customers at a time when inventory and costs are low. An efficient
Supply Chain Management will reduce production costs, inventory and supply
costs as well as service costs in each phase.
Different methods have been developed to model supply chains which can be
divided into four categories:
 Deterministic models in which all parameters of the model are determined.
 Stochastic models in which at least one parameter is not determined, but it
follows a probability distribution function.
 Economic games theory models.
Modeling and simulating supply chain management 819

 Simulation based models.

Most of the models are steady state models based on average performance or
steady state conditions. But, static models encountering dynamic specifications of
Supply Chain Management seem to be insufficient due to demand fluctuations,
delivery time delays compared to order time (lead time) and sale prediction.
The above issue indicates that studying dynamic specifications is a competitive
advantage in modeling supply chain systems. No surprise analyzing and designing
supply chain management systems as a whole has attracted attention of
academicians and industrialists, as well.
Control theory is a suitable mathematical tool for analyzing, designing and
simulating supply chain management systems based on dynamic models. In fact,
the control theory has been used to find solutions for solving problems of supply
chain systems. The article discusses the control theory's application in inventory
and production control issues based on order in supply chain systems.

Control Theory
Using control methods in supply chain management issues dates back to the early
50s. Simon applied the continuous-time theory to manipulate production rate in a
simple system with only one product. This idea was expanded for discontinuous-
time models through using Z Transform methodology for inventory control
purpose.This methodology has been used for dynamic systems such as
production-inventory system as well as for other complicated systems, including
social, planning and strategy developing, educational, medical, biology, energy
and environment and dynamic decision-making systems.
Considering the dire need for establishing a new framework based on control and
feedback regulations in production-inventory systems, a production control
system based on Inventory and Order Based Production Control System
(IOBPCS) can be presented in the form of a block diagram.

Dynamic System Methodology

Dynamic system is a method for understanding behaviors of complicated systems
through simulation and is used to show behavior of a system against different
strategies and policies in order to be utilized in planning and decision-making
processes. This methodology was introduced 35 years ago by Forrester in his first
book titled Industrial Dynamic (1971), focusing on industrial applications. Initial
works paid attention to managerial issues such as instability in production and
employment, deficit or problem in union's growth and decline in market share.
During a short while, this method was used in a broader range of affairs extending
from research and development management to understanding exponential growth
concept in a limited world and reduction in natural resources.
The initial term of industrial dynamic was substituted with a more general term of
"dynamic system". The study of dynamic systems does not focus on a system but
on a problem (Forrester, 1985). Problems in the viewpoint of dynamic system
have at least two specifications. Firstly, they are dynamic. For instance, they deal
with amounts which vary in time. They can be depicted and displayed using time-
820 A. Toloie Eshlaghy and M. Razavi

variant charts. Fluctuations of employment levels in an industry, reduction in

urban municipal taxes base don quality of life, excessive increase of costs in a
construction project, expansion in government's size, and cancer are some
examples of dynamic system problems. Capability in defining dynamic problems
is the first step for learning approaches of dynamic systems (Richardson and Pug,
1981).Secondly, feedback concept is used in dynamic system problems. A good
example for a feedback system is the automatic mechanism in CNC machinery
and closed circuit cameras. All the human systems are basically feedback systems
(Goodman, 1983).

Dynamic system focus on structures and behaviors of combined systems

comprised of mutual feedback loops. The cause-and-effect chart is a simple tool
which helps the modeler in depicting and conceptualizing real world models. In a
cause and effect chart, arcs show the direction of effect and plus and minus signs
show the kind of effect. If a change in a variable causes a change in another
variable in the same direction related to the initial amount of the variable, the
relation between the two variables is positive. If a change in a variable causes a
change in another variable in the opposite direction related to the initial amount of
the variable, the relation between the two variables is negative.

The difference between this chart and other charts, such as flowchart, is in
showing relations which can not be easily depicted in other charts. If a feedback
loop related to a variable causes an increase in the original deviation, it is called a
positive feedback loop. In a positive feedback loop, increase or decrease is
exponential. If a feedback loop changes amount of a variable opposite the
direction of the original deviation, it is called a negative feedback loop. The
negative feedback loop is recognized through a directed behavior toward the goal.
A negative feedback loop is expected to cause an asymptotic increase or decrease
toward the goal (Goodman, 1983).

The abovementioned cause-and-effect is the initial viewpoint toward the problem.

It is basically created to establish a link between the modeler and the system.
Formulating the executive model of the system based on structural details such as
policymaking rates or variables, state variables, secondary variables, information
flows and delays is more specific. Flowcharts provide more details and viewpoints
regarding structure of the model (Richardson and Pug, 1981). The dynamic
system model includes state variables, secondary variables, fixed amounts, rate
variables and initial amounts. Cumulative amounts of state variables are totally
different from effective flows over them in terms of concept. Forrester (1985)
described state variables as accumulators inside the system. These variables can
be inventories, products to be manufactured, factory environment and number of
employees.The presented model for production control system based on order and
inventory control have been shown in the below block diagram:
Modeling and simulating supply chain management 821

Figure 1: IOBPCS family model

In which:
AINV: Actual inventory
AVCON: Average consumption
AWIP: Assembly work-in-progress
COMRATE: Completion rate
CONS: Consumption
DINV: Desirable inventory
DWIP: Desirable work-in-progress
EINV: Error in inventory
EWIP: Error in work-in-progress
ORATE: Order rate

This model can be applied to various products or to a single product. Using

control words, the actual inventory (AINV) is a controlled variable. Meanwhile,
the consumption (CONS) is distortion and the order rate (ORATE) is a
manipulated variable. Two integrals have been used to add deficits in inventory
and work-in-progress (WIP) during time. Each of the IOBPCS family members
has been made based on using a portion or all the five factors which are described
below:
 Waiting time: Waiting time is referred to the time period between production
order and delivery dates. In fact, it also includes the production delay time. It
acts as a smoothing factor in production process which, in fact, shows the
adaptability of production with changes in order rate.
 Target stock setting: This value can be fixed or a multiple of the average
current selling rate.
822 A. Toloie Eshlaghy and M. Razavi

 Demand policy: Current demand is a very important factor, because if it is

removed from the planning algorithm, permanent shortages will be created
with step rise in demand. If we use current demand without average-taking
form, it will cause growing changes and fluctuations in the production rate,
which in turn, will raise production costs. The best and the most appropriate
method for this purpose is the exponential smoothing method because it
requires lower data and is precise for short-term prediction. This can be
described using Laplace transform in control theory. The question is how
much weight should be allocated to the recent data in order to minimize
demand fluctuations and simultaneously respond important changes as well?
 Inventory policy: The inventory policy is important because the rate of
inventory changes has a great effect on production changes and fluctuations.
There is a wrong belief in industries which describes production targets as
factors that can improve inventory deficit for a short period of time, while a
much more time was required to show the manufactured product in the
inventory. Once the product is manufactured, a significant rise will occur in
the work-in-progress assembling and inventory will exceed the desirable level.
The problem should be removed through producing less than the market
demand average in order to near the inventory to the desirable amount.
Anyway, some overshoots occur here which affect the production rate. In fact,
the inventory policy is a feedback loop which controls the rate to improve and
modify the inventory shortage or surplus.
 Pipeline policy: The desirable work-in-progress production is a function of the
market demand average and the required time for production (the waiting
time). During times in which work-in-progress production shortage occurs as
a result of an increase in step demand, deficits should be compensated through
increasing demand for production. Contrarily, if the work-in-progress
production increases in comparison with the desirable amount (due to non-
consideration of the waiting time in demand and inventory policies), the
problem should be removed through reducing the production. In fact, the
pipeline policy is a feedback loop which controls the rate that improves and
modifies the inventory shortage or surplus.

Model behavior analysis

The waiting time is a parameter of the system which should be controlled.

Although designer of the system can not manipulate it, but it is important to delay
it as best as possible. A model of the waiting time with 2 parameters has been
mentioned below:
1
GP ( s ) 
((TP / n) s  1) n
n  1 : first order delay
(1) n  3: third order delay
n   : pure delay
Modeling and simulating supply chain management 823

For the two first cases Tp, n=1,3 equals the waiting time average for each unit,
while for the third case Tp, n=∞ is a fixed delay time. For the third case, the
transform function is changed as below:

(2) GP ( s)  e TP s
The system designer should decide how to adjust the target inventory level (a
fixed amount or a multiple of the average sale) as well as the three policies
(demand policy, inventory policy and pipeline policy) so that it optimizes the
system considering the two objectives of the operation.
A- Improving inventory level
B- Weakening effects of changes and fluctuations on order based demand rate
These two objectives are in contrary with each other. So, for a definite supply
chain, the system designer should find the best inventory and order rate and
exchanging these two factors. If a fixed order rate is used, great changes are seen
in the inventory, increasing inventory costs. Also, if a fixed inventory is used,
production plans are greatly changed, increasing production costs.To this end, the
dynamic system has been tested through applying a step signal related to the
demand rate. The response of inventory is analyzed considering performance
parameters such as rise time and settling time and overshoot. (Charts 1 to 4).
Using feed forward related to the demand policy in the model, it can be seen both
empirically and mathematically that if a step signal is applied as the consumption
rate, the zero steady state error can be attained between the real and the desirable
inventory levels. But, if the market demand is applied to the system without the
average-taking from, a rising fluctuation will be seen in the order rate/production.
In fact, this is a factor in non-realization of the second goal.
The inventory policy defines a rate by which the inventory deficit is improved
through manipulating the order rate. The inventory policy should consider the
system's dynamism and waiting time. The applied change in a specific time on the
order rate will cause a change in the real inventory after a specific period of time,
equaling the waiting time. The inventory policy will lead to rise in the work-in-
progress production and eventually the completion rate and the inventory level are
obtained with a sinusoidal behavior.Results of such a dynamic system will raise
production costs due to non-smooth order rate. Also, inventory fluctuations will
lead to rise in inventory costs (in the case of inventory surplus) and reduction in
services to customers (in the case of inventory deficit). So, only a portion of the
inventory deviation should be modified by the inventory policy.
Te pipeline policy is a correction mechanism which does not use the data related
to the real inventory. In fact, the WIP signal undoes the inventory signal and
increases the average consumption's share in reaching the steady state. The WIP
deficit is obtained by deducting the real signal from the desirable signal (based on
the market demand average).
The pipeline policy is in fact effective for reducing the inventory balance which is
considered as a third element besides the inventory policy and the demand policy.
It is used to calculate the order rate.
824 A. Toloie Eshlaghy and M. Razavi

In comparison to the inventory policy, the pipeline policy responds faster to the
requirements of order rate increase and decrease, especially when sharp changes
are seen in the market demand.
Eventually, it can be said that the WIP control loop will reduce rise time and also
fluctuations related to the order rate. Also, it will also reduce the time to reach the
steady state.
Differences of the two cases are shown in the charts (3) and (4).
Various models which are included in the IOBPCS have been mentioned in the
table below:

Table 1: Various models of IOBPCS

The designer can manipulate the four mentioned elements based on the table and
gain different models. It should be noted that the model mentioned in this article
has been simulated using MATLAB and SIMULINK softwares and results have
been analyzed and compared in charts 1 to 4.
For example, for the IOBPCS model, the WIP feedback loop has not been applied,
a first-order delay has been considered for the demand smoothing policy and the
target stock has been set to a fixed level. The production waiting time is a delayed
model from the order rate which has been shown by a first-order lag with Tp as
the time constant. Also, the demand average-taking process (demand policy) has
been shown by a first-order delay with Tp as the time constant.
So, the order rate is obtained by adding average consumption rate and a portion
(1/Ti) of inventory deficit.
The transform function of completion rate variables (COMRATE), actual
inventory (AINV) and consumption fluctuations (CONS) is described as follows:
Modeling and simulating supply chain management 825

AINV TaTP s 2  (Ta  T p ) s

 Ti
CONS (Ta s  1)(TiTp s 2  Ti s  1)
(3)

(4) COMRATE (Ta  Ti ) s  1


CONS (Ta s  1)(TiTp s 2  Ti s  1)

The characteristic equation is a third-order polynomial which is written as the

multiply of a first-order term by a second-order term. Considering the Routh-
Hurwitz criterion, the system will be stable if all the closed system roots are
located at the left of the jw axis. So, because all the constants of the two terms are
positive, the transfer function for all the non-zero adjusting parameters will be
stable. It should be mentioned that the adjusting parameter related to the feed
forward element (Ta) does not exist in the second-order term, so it has no effect in
oscillatory behavior. The second-order term can be written in the standard form of
and gain the adjusting parameters. To gain stability, we set ∆<0, so:
(5)
  Ti  4TiTp  0  Ti  4Tp
2

Supposing that and placing in the second-order equation, a good design can be
achieved and the relation between Ti and Tp will be obtained as:
(6)
Ti  [ Tp 2Tp ]
Through applying the pipeline policy to the model and simulating behaviors of
AINV and ORATE signals for Tw=Ti=Tp=Ta=2 the difference between non-
using the pipeline policy (figure 3) and using the pipeline policy (figure 4) in the
mentioned charts can be seen.
As it can be seen in the figure 3, all the three policies of inventory, demand and
production have been used, so it has reduced rise time, settling time and
fluctuations compared with the figure 3 which uses inventory and demand
policies.

Figure 2: The APIOBPCS block diagram in MATLAB (SIMULINK) software with

Tw=Ti=Tp=Ta=2
826 A. Toloie Eshlaghy and M. Razavi

Figure 3

Figure 4
Modeling and simulating supply chain management 827

+ AINV
Completion Rate Consumption Rate
Orate
+
-

Delay Time
+
Desired INV
+ Time Constant (k)
AT (INV)
-
+
Error INV

Const. DINV

Switch

Figure 5: The schematic IOBPCS dynamic system in Vensim software

Average
Consumption
+

Average Time

- + + AINV
Orate
AT (INV) Completion Rate Consumption Rate
- ++

+
+ - Delay Time +
Desired INV
AT (WIP) AWIP
AWIP Rate + Time Constant (k)

Error WIP + +
+ Const. DINV
DWIP

- +
Switch
Error INV

Figure 6: The schematic APIOBPCS dynamic system in Vensim software

As it is seen, the problem has been simulated by Vensim software for two case of
only the inventory policy and all the three policies. Results for the two cases are
similar.
 828 A. Toloie Eshlaghy and M. Razavi

Figure 8

Conclusion

The article has used the linear control method and transfer functions analysis for
modeling the supply chain in time-invariant situation, so that it creates the
condition for exchanging the order rate and the best inventory level. Results have
been simulated using Vensim software.

References:

[1] Sarimveis H, Patrinos P, Tarantilis CD, Kiranoudis CT, Dynamic

modeling and control of supply chain systems: A review, Computers &
Operations Research, 2007.
[2] Zhou L, Naim MM, Tang O, Towill DR. Dynamic performance of a
hybrid inventory system with a Kanban policy in remanufacturing
process, Omega 2006; 34:585–98.
[3] Disney SM, Naim MM, Towill DR. Genetic algorithm optimization of a
class of inventory control systems. International Journal of Production
Economics, 2000; 68:259–78.
[4] Ogata.K., Modern Control Engineering, 2nd edition, Prentice- Hall
nternational Edition, 1990.
[5] Disney SM, Towill DR, On the bullwhip and inventory variance
produced by an ordering policy, Omega 2003; 31:157–67.
[6] Evans GN, Naim MM, Towill DR, Application of a simulation
methodology to the redesign of a logistical control system, International
Journal of Production Economics, 1998;56–57:157–68.

Received: October, 2010

View publication stats