SYSTEM DYNAMICS MODEL OF MATERIaL Flow
Case of a Steel Plant
RAKESH KUMAR
NATIONAL PRODUCTIVITY COUNCIL
ODI ROAD
NEW _DEIHI - 110 003 (INDIA)
OLAF KIZINE
PROJECT MANAGEMENT (PVT)LTD
BAHNHOF STRASSE 148
7120 BIETIGHEIM - BISSINGEN (WES? GERMANY)
ABSTRA
The System Dynamics Method has been applied to
simulate the flow of production in a steel plant. This
model has been designed to be an aid in long term
plenning. The model is driven by a time varient input
i.e incoming orders of nine different types of
finished steel products, The internal dynamics is
generated by six negative feed back loops of a
production shop. The material flow takes place through
16 such shops each having its om dynamics which gets
induced to other shops as material flows from coke
ovens to finishing mills. The model makes explicit
the environmental influences, policy parameters and
their relationships with production. Zogether these
explain the dynamic behaviour of monthly production.
It can now be used to experiment with all that
can be thought of to influence the parameters and
improve upon the production performance of the steel
plant. The extended version of this model which
includes the financial aspects is a top management
laboratory for experimentation with different
scenarios of environmental influences and cownter-
acting strategies.
27
1 INTRODUCTION
A system dynamics model has been designed to
simulate behaviour of production and inventory in response
to changes in exogenous variables such as
~ demand
- raw material & power availability
- technical parameters
as well as changes in policy. such as
- stendards
- xeaction co-efficients
The steel plant materiel flow model is presented
here using a deductive approach. First of all the model
and the system boundary are described giving the reader
@ glimpse of the exogenous variables of the model and
their treatment for the purpose of simulation experiments.
Next an overall view of the model is given, outlining the
approach adopted in assembling the submodels of various
production shops constituting the whole steel plant.
Subsequently, feed back concepts underlying the model of
a Shop axe presented, Some of the built~in company policies
which can be tested using the model are briefly discussed.
How the model simulates production is then taken up and
substantiated with om example. Some insight is given into
validation of the model. Finally, some of the various
possible applications are presented.
2 MODEL AND SYSTEM BOUNDARY
The three components of the material flow model are
shown in Exhibit ~ 1. These are
- Environmental scenario Inputs
- Strategic Parameters
- DYNAMO programme.
ENVIRONMENTAL
INFLUENCES STRATEGIC
PARAMETERS
INPUTS
EXHIBIT-1
Environmental Scenario Inputs are the exogenous
variables of the model e.g share of demand, supply constraints
of critical inputs like cooking coal, power etc, The
variables grouped under this head are exogenous: based on the
system boundary decided by the model designer. Consequently
4
the user of the model has to make assumptions about
their behaviour, To carry out the task of making such
assumptions in a qualified manner, the model user is
required to unearth the various forces of change
influencing the exogenous variables and explicity
state his basis for arriving at a scenario. Cne scenario
consists of a set ofexogenous variables and their assumed
bahaviour over the simulation period, Each scenario,
results in a defined behaviour of production using the
feed back structure underlying the material flow model.
Strategic parameters are also exogenous as a@
result of the system boundary. When compared with
environmental scenario inputs these are considered to
be within the control of the management. In other words,
the values attained by these variables at any point of
time are a result of managerial decision process regarding
the choice of technology and organisation. These parameters
also have an impact on dynamic behaviour of production.
DYNAMO programme consists of nearly 1300 statements
and 96 negative feed back loops. These feed backs are the
thizd source of explanation of the dynamic behaviour of
production, Feed back structure, environmental scenario
and Strategic parameters together result in a behaviour
of production.
28
In all cases when the simulated production does
not match the expectations of the management the programme
provides the opportunity to simulate changes in strategic
parameters till the simulated production is identical
with the desired -production. This exercise provides the
management an insight into the extent of changes necessary
in strategic parameters, Now the question arises whether
these can be accomplished with the existing technology
or not. In case the answer is 'NO' the need for change
over to new technology becomes obvious. This programme
cen also be used for appraisal of projects which could
be contemplated for improving the production in the steel _
plant, Heuristic simulations and project appraisal carried
out using the package ave exhaustive, reliable, impartial,
quick and cheap.
3. MODEL OF STEEL PLANT
The dynamo peckage simulates flow of production of
9 varieties of finished steel products, These products are
~ Cold Kolled Sheets
- Hot Holled Sheets
- Spiral Welded Pipes
= Electric Welded Pipes
- Heavy Plates
- Dividing Plates
- Galvanised Sheets
- TMnned Plates
- Electric Steel Sheets.
6
The monthly incoming orders of finished steel products
are the prime movers of material from one shop to the other.
Using suitable conversion factors the incoming orders are
converted into requirements of various in process materials
like coke, hot metal, ingots etc. This computation results
in definition of monthly inflow of work orders for each
production shop. In case this inflow of orders drops dow
to zero, production also comes to a halt and as order
inflow picks up production responds, subject to the
constraints of capacity and materiel availability.
The model distinguishes 25 levels of different
inventories,including major raw materials like ore, coal,
limestone ete, and semi-finished products like coke, hot-
metal etc. and the nine varieties of finished goods. The
material flow beginning with raw material, pesses through
sixteen shops which are arranged in six Stages of production,
before it becomes finished steel as show in Exhibit 2. A
stage comprises of one or a group of production shops. During
a simulation period the quantum of flow from one stage to
another depends upon availability of capacity, workload and
material. Model also takes into account the process loss
(yields), handling loss, wastage & serap. When the production
of a stage is further processed in more than one shop in the
next stage the distribution key is the ratio of the workloads
of various shops comprising the next stage.
29
7 a
COKE OVENS
SINTER PLANT
BLAST FURNACES
|
OPEN LD
HEARTH CONVERTORS
L_
SLABBING MILL
! i
HOT STRIP PLATE
MILL MILL
I
1 T T 1
GALVANISING ELECTRIC ERW. DIVIDING
LINE SHEET PIPES: LINES
Tinning = MILL COLD ROLLING sw
LINE MILL PIPES
EXHIBIT-2
4. MODEL OF A PRODUCTION SHOP
The feed back structure of any shop of the steel
plant has been designed based on six different phenomenon
acting simultaneously on production.
SHORTAGE OF INVENTORY (INPUT OF A SHOP)
The production rate during any simulation period will
be throttled in case the raw material inventory sinks
beyond an alarming level. The alarming level of inventory
and the extent of throttling are the policy parameters
built in this model. This phenomenon can also be represented
by a feed back loop as shown in Exhibit - 3.
30
RAW MATERIAL PRODUCTION
SHORTAGE, _ RATE
+
EXHIBIT-3
EXCESS INVENTORY (OUTPUT OF A SHOP)
The production rate during any simulation period
will be throttled in caSe the level of inventory holding
approaches the alarming level. The alarming level for
throttling and the extent of throttling constitute the
policy parameters in this model. This phenomenon cam also
be represented by a feed back loop shown in Exhibit ~- 4.
9
RIVENTO
PRODUCTION _ AVE ee
RATE
EXHIBIT-4
DOWNTIME
The production rate during any simulation period is
limited by the installed capacity computed after giving due
allowance for downtime. The dowmtime phenomenon has been
modeled as function of usage of equipment. This can also
represented by a negative feed back loop as shown in
Exhibit - 5. The model has been designed to simulate
three types of events resulting in dowmtime. These
are random failures, planned minor repairs and planed
capital repairs,
31
+
PRODUCTION EQUIPMENT
RATE ~ ‘ USAGE
+
BREAKDOWNS:
EXHIBIT-5
CAPACITY SHRTAGE
The production rate is also effected by the level
of capacity, When workload level exceeds certain defined
standards (policy parameter) capacity expansion as well
as recruitment tekes place to push the production rate up.
On the other hand, vhen the workload level sinks below
certain defined standards, divestment as well as retrenchment
takes place. This phenomenon is represented by a negative
feed back loop as shown in Exhibit - 6.
a
WORK
— LOAD
PRODUCTION
RATE -
+
+
CAPACITY
EXHIBIT-6
WORKLOAD
The work load growth rate influences capacity growth
in the long run as discussed in para above. Over short
periods of time the utilisation of capacity is effected.
As work load inrate goes up production in rate picks up
to bring down pending work load level, The phenomenon
can also be expressed by a negative feed back loop as
shown in Exhibit - 7.
32
+
WORK PRODUCTION
LOAD om RATE
EXHIBIT- 7
INCENTIVE
The production rate in this model defines the amount
of incentive earned by the mmployees of a shop. The incentive
so earned influences the production rate. When incentive
earned during a period falls short of the expected incentive
level, the labour productivity is understood to push the
production rate. This phenomenon can also be expressed
by a negative feed back loop a8 shown in Exhibit - 8.
13
INCENTIVE
EARNING
+
PRODUCTION _
RA j
va = INCENTIVE
SHORTFALL
DESIRED
INCENTIVE
EXHIBIT-8
5. BUILT IN POLICIES
Some of the feed back phenomenon discussed in para 4
have built in policies of the management. These policies are
expressed by parameters which represent the conditions when
action should be taken i.e inrates or outrates have to moderated
either upwards or downwards, so that levels are back to
their acceptable limits. The choice of parameters represents
management's philosphy/attitude and reflects the degree of
risk the management is prepared to accept during decision making.
For example, consider raw material inventory level, the policy
33
14
parameter when compared with the Simulated inventory level
gives a ratio, which indicates whether control action is
desized or not. When this ratio lies between O and 1 it
implies that the inventory level is below the limit set by
the policy parameter end therefore ection needs to be
initiated by reducing the consumption rate, This is
shown in Exhibit - 9. Value of ratio greater than 1 does
not warrant any action. The degree of action is identified
in a coordination system as shown in Exhibit ~ 10.
ACTION ON
INVENTORY CONSUMPTION
LEVEL RATE
Ir
INVENTORY
RATIO 41
POLICY
PARAMETER
EXHIBIT-9
15
(INVENTORY RATIO —>
1
FAMILY OF
MULTIPLIER SECURVES
COEFFICIENT
(DEGREE
OF oS
ACTION)
EXHIBIT-10
Os 4
On the x-axis is the inventory ratio ranging from
© to 1 and on the Y-axis the degree of action read as
multiplier co-efficient. In case ratio is 1 the degree
of action is zero and the mltiplier co-efficient is 1,
implying that there is no constraint on consumption rate.
If ratio becomes zero the multiplier co-efficient is also
zero thereby making the consumption rate also zero. In
between the two extremes there is infinite choice of action
represented by a family of S-curves. As the curvé move
closer to Y-axis they represent less and less risky policy
as far a8 inventory shortages are concerned, The choice of
34
16
curve represents the risky posture of the management.
This methodology of identification of policy parameters,
computation of ratio and reading the degree of action
from S-eurve has been applied in most of the cases for
modelling production constraints a5 well as pressures
pushing the production to go up.
6. PROWCTIGN FLOW SIMULATION
The monthly simulation of production of a shop
begins with calculation of the installed capacity. Installed
capacity is represented by the level of machines which is
also expressed in terms of maximum machine hours. These
are then adjusted for domtime to arrive at available
machine hours, Similarly the level of men employed is
translated in term of available man-hours after taking
into accomt absenteeism. To compare the two, machine hours
are translated into equivalent manhours. Minimum of the two
represents the available capacity. The available capacity
hours are then multiplied with various constraint factors.
These constraint factors are computed in the manner discussed
in para 5. the constraint factors have been designed to take
care of conflicting situations also. Consider for example
that the input material of a shop sinks below alarming level,
the policy parameter indicates throttling of production. at
the same point of time it may also be the case that output
material of this shop drops down to an elerming level thereby
17
indicating pressure to push the production up. The
model takes cognizance of the conflicting policies
and adopt a moderate approach.
Te MODEL VALIDATION
To validate the models of shops end finally the
assemblage of these models i.e the steel plant model,
ex-post simulations have been mate, During these
simulation the environmental scenario inputs were the
actual values and so were the strategic parameters. As
far a8 strategic parameters are concemed, only average
values were used, Now it was expected that in case
model formulation has captured the relevant causes
of dynamic behaviour of production the model behaviour,
should be in close proximity with the actual behaviour
of production in the past. By close proximity it was meant
that the average simulated value should be with in an
accuracy of + 1% of actual monthly average of past 48
months. Also, on comparison at any point of time the
monthly simulated production should not exceed + 10% of
the actual production value, It was further expected that
changes in trends of simulated production should be equal
to or above 90% of such changes in reality, All these
conditions have been finally satisfied before the model
has been used for ex-ante Simulations.
35
18
8. APPLICATIONS
Model has been used to make simulations with two
scenarios of environmental inputs based on optimistic
assumptions and pessimistic assumptions, keeping the
strategic parameter values seme gs had been used in
final validation run, This has been done to study the
necessity of divestment, expansion investment, retrenchment
and vecruitment under the extreme behaviour of business
environment. A simulation run based on a scenario input
provides information about
- where to invest/divest
- when to invest/divest
- how mich to invest/divest
Yhe ex-post simulations have revealed that investments
as Suggested by model in some caSe have not been identical
with those made in reality, In the course of discussion
with management it has been appreciated that the investment
proposals made using the model are taking into account an
integrated view of the whole steel plant, Vide the
conventional approach this integrated view is invariably
overlooked and at times distorted to provide adequate
justification for certain preferred, subjective choices
of investment.
The steel plant model differentiates 6 stages in
production arranged in series. During any Simulation run
19
it highlights the stage that has produced the minimum
quantity thereby suggesting that efforts to change
strategic parameters should be concentrated here in
the near future for the purpose of debottlenecking.
The various projects that have been contemplated
to improve the production of a shop over the simulated
period can be evaluated for their priority as well as
utility in the context of the whole steel plant in,an
integrated manner by making few Simulation runs, one
with the project implemented and the other in the absence
of eny such project. For the purpose of understanding and .
evaluating the financial implications of such projects, the
material flow model has been extended to compute financial
results but has not been taken up for discussion in
this paper.
As explained earlier in para 6, incorporated in
this model ere a few policy parameters, Choice of these
policy parameters can be tested against the objective of
profit to understand the advantage of one choice against
the other.
Moderator:
Section 2; Public Sector Applications
Richard H. Day, University of Southern California
37
