Analysis of a serious accident emergency drill with a simulation model
Angela Martin-Méndez, B.Sc.
amartin2013@ student.hult.edu
ABSTRACT
Studies of labor risks with simulation models can be categorized into two groups: those which analyze serious
accidents to obtain applicable rules for the future, and those which analyze the rules and mechanisms of control of
risks to simulate their functioning before potential accidents occur in order to identify and manage the necessary
guidelines of prevention and course of action. Here we have a) an ex-post approach where we study the chain of
events that have caused the accident to prevent it from repeating itself, and on the other hand we have b) an ex-ante
approach that analyzes the complexity of the relations between the parts of the system that can give rise to accidents
and simulate diverse strategies that enable reduction of the risk. Later the system of prevention of labor risks of a
generic company is described using a causal graph, and that same graph transforms itself in a model of simulation -
in spite of handling qualitative variables - to simulate the effects that are produced in a serious emergency exercise.
Keywords risk; labor; accident; safety; emergency; control; prevention; simulation model; system dynamics; causal loop diagram
Simulation models have shown a high level of
effectiveness in specific activities in companies, such
as, for example, the design of mechanical
components, where it is possible to carry out trials of
vibration resistance in a simulated model before a
prototype is built. Simulation models are also useful
in production management because they allow
foreseeing the impact of sudden changes on demand
levels or a shortage in the stock of one of the main
resources, thus allowing for forward-planning of
effective solutions for each situation.
Inside a company it is also possible to use
Dynamic Systems simulation models in the realm of
management (Sterman, 2000), and a extensive
bibliography with abundant examples that can serve
as an orientation is available. Thus, before using this
technique in our own situation we can check what
others have done while tackling a problem similar to
ours.
System Dynamics is a well-defined methodology
with the first publications dating from more than 50
years ago, and a sizable number of projects and
publications that capture the experiences of
researchers and professionals who have used the
simulation models created with this methodology to
analyze a wide spectrum of areas and problems.
In the business field we can differentiate between
models that deal with production issues and those
related to managerial ones. Management-focused
simulation models have an interesting distinguishing
factor that is not commonly found in other
methodologies and that is the need to explain them in
a clear and fast manner to people not familiar with
the methodology in order to justify conclusions or
suggestions in a convincing way. It is indeed a
formidable challenge, which sometimes is not
achieved and hurts the development of otherwise
valuable projects because the decision-making
Managers may not have enough time to go deeply
into the details or hypotheses used to inform the
design of the simulation model.
It is necessary to design a detailed simulation
model so as to analyze the problem and being able to
justify suggestions or conclusions. Even further, it is
necessary to design a very simplified version of such
model so we can explain it in a clear and concise way
how the conclusions are not derived from personal
#
opinions but from the rigorous evaluation of different
alternatives. This situation is not present in other
simulation models in companies, like production
management, where the person responsible for it will
need to know the model structure in detail, what
values were chosen, where the data came from, what
simulations justify the suggestions and where he even
may be concemed with ease of use and the option for
further updating of the model.
This work illustrates with an example the
complete process of using simulation models with
System Dynamics in a specific managerial topic, in
this case labor accidents prevention, paying special
attention to the need for presenting the management
team with clear conclusions based on a simple
simulated model. The process starts with the precise
definition of the problem (labor accident risks),
followed by displaying the expert’s opinion in a
causal diagram, which will later be translated into a
simulation model. The model’s variables will then be
assigned values, thus allowing for their observation
over time allowing us to reach convincing
conclusions or recommendations. It should be noted
that sometimes the causal diagram itself offers
enough information about the structure of a problem
to allow for solid conclusions (Senge, 1990, Kim,
1992, Marais, 2006).
In the field of labor security or labor risk
prevention there is a recent paper (Miang, 2010) with
a general review and a selection of the most
important contributions. Many of the published
works analyze accidents that have already occurred
and thus deal with particular details of those
incidents. For our purpose, (which is to illustrate how
to offer good recommendations based in a simple
simulation model) we have selected an interesting
work (Cooke, 2006) that offers a generic vision of
risk and labor accidents based on the opinions of
subject matter experts. Nevertheless, it would be very
difficult to present this work in a few minutes, since
the simulation model is quite complex and has close
to 40 variables. We will use these subject matter
experts’ concepts and opinions found in Cooke’s
paper as the basis for building a case that illustrates
the whole process of applying System Dynamics
simulation models to the field of management.
1. CASE DESCRIPTION
A transport company involved in moving
dangerous goods has never had any serious accidents,
even though it has had many small ones. They want
to analyze the effects of carrying out a serious work
accident exercise before taking it to actual practice
because of the high cost attached to it.
On the basis of several meetings with managers
and employees of the company, a causal graph is first
developed, gathering the elements that intervene and
the relations that exist between them.
Before constructing and executing the simulation
model it is useful to draw the behavior we expect
(figure 1) of the principal variables of the system
before and after the moment the serious work
accident exercise takes place (in month 6). Later we
will compare our intuitive expectations with the
results of the simulation model in order to analyze the
reasons for the difference between expectations and
results.
Interest from security managers
10 ;
1
1
1
1
1
1
y
Os
0,0
1
'
'
1
'
° 6 12 months
Interest from evaployees on the security
Os
0,0
12 months
The experts think that after the safety exercise (in
the 6th month), interest in safety topics will increase
on the part of the managers and executives of the
company. Furthermore, it is also going to increase the
interest in safety on the part of the employees, but
only for some time. As a result the level of risk of
accidents will decrease for a specific period of time,
retuming slowly to the current level of risk.
2. CAUSAL LOOP DIAGRAM
A causal loop diagram is an important tool used in
the initial stages of a system analysis which
qualitatively and explicitly expresses our ideas about
the basic causal feedback relationships — their
linkages and whether positive (increasing an effect)
or negative (decreasing an effect) - between major,
key components of the system under study. In this
example, (figure 2) a major increase in interest in
safety topics by the managers prompts an increase in
resources dedicated to safety (for example
improvements in systems maintenance, active safety
policies, risk analysis, etc.), that thereby — causally -
diminishes the risk of accidents.
risk of
accidents
ye
£+— \
costs +
es \
interest from
securit
managers 7
7 security Figure 2
resources
In addition, dedication of more resources to safety
increases general safety awareness because there are
more measurements of all kinds (i.c,“positive
feedback loop”). This increase in safety awareness
leads to an increase in the interest of managers in
safety. On the other hand, we need to remember that
as more resources are dedicated to safety, we incur
higher costs, acting as brake or stabilizer (i.e,
“negative feedback loop”), diminishing the interest of
managers in safety. In addition to effects on
management, a key element in considering safety
topics are the employees themselves who experience
the risk directly. In this way (figure 3) a major risk of
accidents is translated into more reported minor
d by the empl ly, a large
increase in the number of minor accidents increases
the employees’ interest in safety topics, increasing
their knowledge of the topic and therefore, on the
basis of the information or the experience,
i the risk of accid R ing these
elements and relationships with | a causal graph, we
are able to observe that there exists a link of a
negative sign, stabilizing the system. This way, an
increase of risk provokes an increase of accidents,
which motivates the employees to improve their
knowledge, which in turn diminishes the risk.
risk of
- accidents
v Ss
serious
accident \
practice
+
employee's \ reported minor
knowledge accidents
about risks
+
information to
\ employees
\ ‘ ff
interest from” *
gues TEE
Do bear in mind this additional point relating to
the employees: they have to report and communicate
information about accidents, especially the minor
ones. In this way, (figure 4) an increase in reported
minor accidents makes it necessary for management
to allocate some additional resources in order to
investigate the causes of each accident, to evaluate its
importance, to propose preventive measures, etc.
reported minor
accidents _ _— resources
i 7 7
li Ka a \
/ 5
| resourves needed }
for security |
\ \ /
\ — ve
i 4 r
importance
interest from
perceived by the
‘employees on the 4 .
security} enyproyere Figure 4
A greater need of resources dedicated to safety
translates into having more resources allocated,
which, in tum, sends a message to the employees on
the importance that the company gives to safety
topics. Nevertheless the employees are going to
perceive the importance of safety topics only if the
resources dedicated to safety topics are equal to or
more than those which are considered necessary. For
this reason there is a link with a the negative sign,
which indicates that” a greater need of resources
produces a perception in the employees of a lesser
importance in any measure that does not devote an
equal quantity of resources in safety.
We can integrate the elements and relationships
we have described above into one causal graph, such
as the one that appears below (figure 5). Most of the
elements are qualitative variables, to which there is
assigned the initial value of 0.5 to gather the current
situation. We are going to use this model to run a
simulation throughout 12 months and to reproduce an
exercise of serious work accidents in the 6th period in
which this variable will take the value 1.
serious
accident
practi
employee's
knowledge
about risks
3. MODEL STRUCTURE
The causal graph of the model is formed by twelve
elements and the relationships between them. All the
equations are simple arithmetical expressions,
without delays of information and linear relations,
which will allow us to easily understand and explain
the inner working and the results.
In general all the variables of the model have
arithmetical simple equations that allow and support
the initial values (0.5) throughout the time.
To simulate the accident it uses the function
PULSE that is activated in the 6th period. To prevent
the level of risk of accidents becoming zero or
negative in any moment, the model uses the function
IF THEN ELSE so if the risk is less than 0.2 and the
impact of the knowledge and the resources dedicated
to safety tend to make it go down and down, the
variation is null.
security
reported minor resources
accidents 05
a 0s
N
0.5
information to resources needed
employees for security
0.5 0. wa
\
5 perceived by the
employees
interest from 05
employees on the - 4
security 9.5
4, RESULTS OF THE MODEL
The model enables the analysis of the reasons for
the changes that are perceived in the variables after
the exercise of a serious work accident. In particular,
it shows that the number of reported non-serious
accidents increases and why the safety alerts
diminish.
The number of reported non-serious accidents
increases after the exercise. This happens because the
interest of the employees in safety topics increases,
and they are much more careful before any accident
occurs - even if the accident would be negligible.
This is in spite of the fact that the risk has diminished
because after the exercise the resources allocated for
safety have increased. It can be seen the behavior of
these variables in figure number 6, where the risk (A)
diminishes because of major resources in safety, but
the interest of the employees in safety (B) increases
due to the major information about labor risks..
2
08
B
06
04
02 A
i)
1 z 3 4.5 6 j 8 9 10 11 12
Figure 7
The interest of the managers in safety (A) has a
similar behavior (figure 8) to the risk perceived by
them across the safety alerts (C). Thus, it has a
maximum value coinciding with the completion of
the serious work accident exercise, and later it
diminishes to lower values than the initial one.
Figure 6
We see that the number of reported minor
accidents increases is linked to the increased
information about labor risks that the employees
receive and their increased interest in safety topics,
which translates to better knowledge of the risks, and
thereby a reduction in the level of risk.
The safety alerts diminish after the work accident
exercise (figure 7) basically due to the fact that the
risk (A) diminishes. This occurs as a consequence of
the significant safety resources that are allocated to
safety topics (B), or because of better organization
and efficient use of the resources, on the basis of the
conclusions that the experts obtain in carrying-out the
exercise.
Figure 8
Nevertheless (figure 9) the interest in safety topics
of the employees (A) is remains elevated after the
exercise and we can even observe that it is increasing
progressively. This slightly intuitive behavior owes to
the different dynamics of these variables as we will
see later.
The interest of the employees depends on the
information that they receive from the accidents they
report. A serious isolated accident increases the
employees’ level of information, which increases
their interest for safety (A), and this increases the
number of minor informed accidents (B), which
increases the quantity of information that they receive
(C) in aloop.
a
This process only diminishes because of the
negative behavior of the parameter reflecting the
declining importance perceived by the employees,
and that is because the resources allocated to safety
are not the necessary ones in the measure that the
interest of the managers is diminishing.
Figure 9
In this point, even though it is a theoretical
exercise, it is important to compare (figure 10) the
previous expectations of the company’s safety
experts (A) with the results that the model shows (B).
Sometimes the models confirm the previous
expectations, but in complex systems counterintuitive
behaviors often take place, so the model shows a
different behavior from what was previously
expected. In this case, after checking that the
functionality of the model is correct, we can follow
the model step by step from where the behavior of the
model is has originated and compare it with the
expectations in order to achieve a deeper
understanding of the reality.
It is always possible to closely analyze the results
of the model because the hypotheses used are those
contributed by the actual experts. Therefore, the
experts who have taken part in its design will not see
the model simply as a black box.
This model is an example of the power that the
causal graphs and the models of simulation of system
dynamics have to analyze the dynamics related to
labor risks, prevention of accidents, constant
improvements in safety culture, safety costs analysis,
etc.
1,0
os
Os
0,0
interest from security managers
T
'
'
'
1
'
1
y
S
12 months
Interest from enaployees ow the security
1
1
1
1
é
risk of accidents
12 months
5. CONCLUSIONS
The simulation model elaborated by the experts
enables us to observe that after a safety exercise of a
serious work accident risk will decrease about 50%
during a long time after the work accident exercise.
Additionally the model allows us to observe (figure
11) how after a security exercise the interest of the
managers (A) will diminish because they perceive a
minor risk as a consequence of the existing additional
measures of safety, hile the employ are
going to increase their interest in safety topics largely
(B) due to the feedback loop across the information
that they receive about minor accidents.
Figure 11
Acknowledgements to
Miguel Velasco and Michael Frenchman.
6 MODEL EQUATIONS
(01) costs=security resources
(02) employees knowledge about risks=
interest of the employees on the security
(03) importance perceived by the employees=
0.5+(-resources needed for security +security
resources)
(04) information to employees=serious accident
practicetreported minor accidents
(05) interest from security managers= INTEG (
var interest managers,
initial value=0.5
(06) interest of the employees on the security=
INTEG (var interest employees,
initial value=0.5)
(07) reported minor accidents= (risk of accidents
+tinterest of the employees on the security)/2
(08) resources needed for security=reported
minor accidents
(09) risk of accidents= INTEG (-var risk,
initial value=0.5
(10) security alerts=(security resources+serious
accident practice+risk of accidents)/2
(11) security resources=(resources needed for
security +interest from security managers)/2
(12) serious accident practice= PULSE(6, 1)
(13) varinterest employees= (information to
employees +importance perceived by the
employees)/2-0.5
(14) var interest managers=(security alerts-
costs)/2
7. BIBLIOGRAFY
Cooke, D. et al [1] (2006). Leaming from
incidents: from normal accidents to high reliability.
System Dynamics Review 22, 213-239.
Ghaffarzadegan, N. et al [2] (2009). Why and
How Small System Dynamics Models Can Help
Policymakers. System Dynamics Society Conference
Proceedings.
Miang, Y. et al [2] (2009) System Dynamics
Analysis of Organizational Accidents: A Review of
Current Approaches. System Dynamics Society
Conference Proceedings.
Sterman, J. (2000). Business Dynamics: Systems
Thinking and Modeling for a Complex World,
McGraw-Hill.
