An Event-Driven Scheme of Scenario Deduction for Railway
Emergency
Zhenhai Zhang
*
, Xiaozhen Yin and Zhilong Xu
School of Automation and Electrical Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China
* Zhenhai Zhang (1983—), Male, Linzhou city, Henan province, associate professor, Post-doctoral researcher, interested in
s
cenario deduction
f
or railwa
y
emer
g
enc
y
Keywords: Railway transportation, emergency information, scenario deduction, event-driven model.
Abstract: Railway emergency management is currently transforming from plan management to scenario management.
Firstly, an event-driven model of scenario deduction for railway emergency was established. Then, according
to the evolution process of scenarios for railway emergency, the emergency scenarios were divided into initial
scenario, middle scenario and final scenario, among which the mutual relationships were analyzed.
Furthermore, using Bayesian network inference algorithms, the specific process of scenario deduction for
railway emergency was elaborated. Finally, taking train derailment accidentals for example, the states of node
variables in scenario network were deduced. The scenario deduction results are in line with the real thing,
which prove the proposed scheme feasible and effective.
1 INTRODUCTION
Railway has played an important role to ensure the
smooth operation of economy and social stability.
Due to the extremely high requirements of high-speed,
high-density and heavy-load railway, the
management tasks of railway transportation are
increasingly arduous. Thus, the possibility of
unexpected events is also increasing. At present, all
kinds of railway emergency plans are improved and
published, such as, flood protection, fire accidents,
network and information security incidents and so on.
Combined with geography, physiognomy, climatic
characteristics in every railway administration,
emergency plans have been consummated, which
conforms to the characteristics and requirements of
the region to make the operation program specific.
Foreign railway companies also use the safety
management system to provide dangerous source
management, emergency investigation, information
of track quality and standard degree and so on. It
provides effective information for the timely
prevention and timely setting of emergency plans for
railway accidents. By comprehensively analyzing the
relevant data, it can provide auxiliary decisions for
the maintenance, supervision, accident prediction,
and emergency rescue of the railway line.
The evolution of emergencies itself has strong
dynamic, complexity and uncertainty, it determines
the problems that emergency decision of the
emergencies are also uncertain and complex(Wang, et
al, 2011). At present, the decision-making method
based on experience depends on the current
information and experience which has been analyzed.
Therefore, it has strong subjective randomness, and
the incomplete analysis of information may seriously
affect the decision-making results of decision-makers,
so it is difficult to ensure the correctness of decision-
making results. The emergency decision-making
method of current emergency has begun to change
from subjective decision-making based on
experience to scientific objective decision-making
based on comprehensive data analysis. It combines
the expert knowledge and mathematical model.
In the field of emergent evolution and emergency
decision-making research, She Lian put forward the
theory of traffic disaster early warning management,
Huo Ran had done a lot of careful research on the
construction of fire emergency scenario and the
analysis of incident evolution. Jiang Hui made a
preliminary analysis of the concept and connotation
of the scenario evolution in the real time decision
making of a rare emergency, divided the evolution of
the event stage. According to the limitation of the
traditional “prediction - response” and the
characteristics of various stages of unconventional
emergencies, he put forward corresponding measures
and countermeasures. The research of scenario
Zhang, Z., Yin, X. and Xu, Z.
An Event-Driven Scheme of Scenario Deduction for Railway Emergency.
In 3rd International Conference on Electromechanical Control Technology and Transportation (ICECTT 2018), pages 125-130
ISBN: 978-989-758-312-4
Copyright © 2018 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
125
inference process is an important means of
emergency decision support for railway emergency.
This paper will establish an event driven model of
railway emergency scenario inference, study the
function of event driven by analyzing of the process
of deduction.
2 SCENARIRESPONSE MODE OF
RAILWAYEMERGENCY
The scenario coping model is a solution, which is
made by the decision maker. They analyze the
existing information and data, integrate them,
complete the scenario construction, analyze the
situation by the experience information, evaluate
the evolution of scene situation and then obtain the
corresponding solution according to the evaluation
results(Li, et al, 2012). The starting point of the
scenario coping model is that the scenario is not static,
but dynamic. As the emergency process goes on, the
content of scenario element description also changes.
Emergency decision makers can predict the evolution
of events and trends based on scenarios. According to
the results of the prediction, the corresponding
measures are taken to control the development of the
events, and change the direction of a sudden event to
a good aspect.
3 UNCERTAINTY OF SCENARIO
DEDUCTION
After the failure of the early warning of the railway
incident, the event scene will enter the evolutionary
stage. Aim at the dynamic characteristics of the event,
decision makers must make real time decisions on the
basis of the process of scene evolution. The current
state of the situation changes with uncertainty, so it
will inevitably make the scenario deduction and the
evolutionary path uncertain. The situation of railway
emergencies is a process of dynamic evolution. The
process of evolution is shown in Fig. 3.1, which may
change almost at any time. The key to influencing the
decision-making subject's important decision is the
state of the railway emergencies’ scenario factors.
According to the whole emergency management
process of railway emergencies(Rong, et al,2012),
this paper divides the situation of railway
emergencies into the initial scenario, intermediate
scenario and ending scenario. The initial scenario is a
description of the state of the scenario factors when
the railway incident occurs. For a specific railway
emergency, the initial situation is determined. The
middle situation is caused by the influence of the
initial situation through various scene elements or the
interference from the outside world. The impact of
different factors will lead to the evolution of the
scenario path. The ending scenario is the state of the
system after the emergency response of the railway
emergency. The evolution of railway emergency
scenarios may be affected by different factors, which
may lead to multiple evolution paths. Therefore, there
may be many situations when the scene ends.
()
k
qpqq
p
p
mnmm
n
n
tt
iii
iii
iii
xxx
xxx
xxx
,,
1
21
22221
112
21
22221
11211
11
L
L
MOMM
L
L
L
MOMM
L
L
×
×
Figure 3.1: Process of scenario evolution.
4 EVENT-DRIVEN SCENARIO
DEDUCTION MODEL
4.1 Elements of event-driven model
The emergency response was immediately activated
after the incident occurred. Emergency response
events belong to the driving elements of the scenario,
which is the emergency task that the emergency
subject needs to implement in order to meet the
emergency target, or the action of the emergency
object, which will inevitably lead to the change of the
state of the emergency situation. Therefore, we can
indicate an emergency response event by a group of
seven elements:
{
}
Event = ID, Name,Subject, Object, Action, De
s
cription, Previous
Among them,
ID
is the only sign of the event;
N
ame
is the name of the event;
Subject
and
Object
represent the subject and the object of the event
respectively;
A
ction
is the action of the subject for
the object;
escription
is the description of the
event;
P
revious
is the set of predecessor nodes for
the execution of the event.
Scenario inference is a discrete visualization
process of the emergency entity state and behavior in
the specified space and time, which is the main line
of the implementation process of the emergency task.
ICECTT 2018 - 3rd International Conference on Electromechanical Control Technology and Transportation
126
The emergency scenario inference process includes a
series of scenarios. The scenario inference uses an
event driven model. An event can be considered a
function that maps time to a Boolean value, that is:
{
}
:, Event t True False
other
at t happensevent if
)(
=
Flase
True
tE
4.2 Construction of the scenario model
In the scenario model of the railway emergencies, the
change of the state of the event scenario has a great
influence on the evolution of the situation, and there
is a strong causality between them. Therefore, the
deduction process can be expressed through the
Bayesian network. According to the analysis of the
evolution of railway emergencies, the sub-elements
of scenario
i
S
are used as network node variables to
construct Bayesian network for scenario deduction.
The node of Bayesian network in
i
S
is divided into
three categories according to its effect on the process
of inference, that is, the scenario input variable, the
state analysis variable and the scenario output
variable.
(1) Scenario input variables: Describes the input
elements for the situation analysis, which are the
determined variables in scenario
i
S
. Scenario input
variables mainly include the variety of factors that
cause the situation or the collection of various
measures in the process of situation response
{
}
1
i
Pp im=≤
, the set of state variables of the
event scenario itself
{
}
1
k
IS is k l=≤
, and the set of
response variables that control the events to change
{
}
,
R
CR JR=
,among them
{
}
1
j
CR cr j n=≤
,
{
}
1
j
J
Rjr ef=≤
.
(2) State analysis variables: Describes the state of
the elements require scenario analysis. The
development trend of the situation element state can
be obtained through the analysis of the state of the
event itself and the state of the disaster situation. In
this paper, the state analysis variables are divided into
two parts: the state variable set
{
}
1
k
IS is k l=≤
and
the event state of the disaster
{
}
1
q
ES es q t=≤
.
(3) Scenario output variables: Describe the next
possible occurrence of Scenario
1i
S
+
, including the
affected external environment state and the loss state
of the event scenario. Let
{
}
1
w
OS os w h=≤
be the
set of affected external environment state variables,
and
{
}
1
z
FS fs z x=≤
be the loss variable set caused
by event scenario, among them
{
}
1
,
i
SOSFS
+
=
.
The key to determining the impact of the event's
scenario input variables, state analysis variables, and
scenario output variables is how to combine specific
railway emergencies to collect event information and
domain expert knowledge. Based on the analysis of
the evolution of railway emergencies, the three-tier
Bayesian network structure of "scenario input layer -
state analysis layer - scenario output layer" is further
constructed according to the causal relationship
between the nodes, as shown in Figure 4.1.
Figure 4.1: Scenario deduction model.
5 CASE ANALYSIS
Taking the derailment accident of a train as an
example to make empirical analysis .When the freight
train reached about 200m before the exit of a tunnel,
the vehicle pulled behind the locomotive suddenly
derailed and caused the leakage of the liquefied
petroleum gas tanker. So train drivers take braking
measures. The locomotive and 1-37 vehicles parked
in the tunnel and 38-55 vehicles parked out of the
tunnel. The leakage of liquefied petroleum gas spread
rapidly and burn, that caused casualties in the tunnel,
and the line interrupted.
(1) Scenario profile
The event is simplified as follows: the freight
trains in the tunnel are derailed for some reasons and
cause fire and explosion
{
}
,
I
SPISSIS=
. The drive
event is
PIS
and the current state of the event is
SIS
.
At present, the emergency management department
should take a variety of emergency plans and
different emergency plans lead to different evolution
paths of event scenario. Thus, the Bayesian network
structure of the event scenario evolved is shown in
Figure 5.1. In view of the current situation,
An Event-Driven Scheme of Scenario Deduction for Railway Emergency
127
emergency management departments may take a
variety of contingency measures, and different
emergency measures will lead to the evolution of
event scenarios in different directions.
Figure 5.1: Instance of scenario deduction.
(2) Node probability assignment
First, we need to determine the probability of each
situation and the probability of not happening. In this
case, the expert specifies the conditional probability
of the node according to the knowledge experience
and the changing process of the event scenario, and
constructs the conditional probability table of the
event scenario. The conditional probability of each
scenario is shown in Table5.1.
(3) Scenario deduction
Using Bayesian reasoning algorithm, we start
from the top scenario node in Figure 5.1, and
gradually calculate the state probability of the follow-
up critical scenario to track the change process. The
state probability for the event
1es
is calculated as
follows:
() ( )( ) ( )( )
11/ 1/P es P es S True P IS True P es IS false P IS false== =+ = =
Similarly, the probability of the state of the other
critical scenarios is calculated from the conditional
probability table, and the results are shown in Table
5.1.In the process of reasoning, we simplify the
process of case scenario evolution, and the path of
scenario evolution only takes into account two
cases(Shu, 2012).Considering the actual situation of
the evolution of the event scenario, the above
reasoning results are basically consistent with the
actual situation, which proves that the event driven
railway scene scenario deduction method is effective
and feasible. However, there are a variety of
responses to the same scenario, so there are many
ways of the evolution of the situation. Therefore, in
practical applications, we must conduct a
comprehensive and m-path reasoning analysis for
specific events, and make the evolution of event
scenarios more realistic and systematic.
Table 5.1: Probability of critical scenario state of events.
IS
1
f
s
True
1
True
0.1*1+0*0=0.1
False
0
False
0.9*1+1*0=0.9
1es
2
f
s
True
0.9*1+0*0=0.9
True
0.3*0.9+0*0.1=0.27
False
0.1*1+1*0=0.1
False
0.7*0.9+1*0.1=0.73
2es
3
f
s
True
0.95*0.9+0.2*0.1=0.875
True
0.5*0.875+0*0.125=0.4375
False
0.05*0.9+0.8*0.1=0.125 False 0.5*0.875+1*0.125=0.5625
3es
4
f
s
True
0.8*0.875+0.1*0.125=0.7125
True
0.9*0.7125+0.3*0.2875=0.7275
False
0.2*0.875+0.9*0.125=0.2875
False
0.1*0.7125+0.7*0.2875=0.2725
ICECTT 2018 - 3rd International Conference on Electromechanical Control Technology and Transportation
128
6 COMPARISON OF RELEVANT
STUDIES
At present, the methods applied to scenario deduction
are mainly based on contingency plans and case-
based reasoning. These two methods can help to
realize the scientific decision of emergency response
of railway emergency, and can assist decision makers
to make emergency decisions to a certain
extent(Cheng, et al, 2012). They can use the existing
knowledge to form a more scientific and auxiliary
decision-making knowledge and improve the
scientific nature of emergency decision-making.
However, due to the differences and application
characteristics between the different methods, there
still have some limitations in practical applications,
as shown in Table 6.1. Among them, is preferred,
is medium, and is poor.
Through the comparison of the three kinds of
deduction models, studying the applicability of them
in the emergency response of railway emergencies,
and theoretically analyzing the differences between
them, we can conclude that: The event-driven
deduction model is superior to the other two methods
in terms of scope, flexibility and the level that need
expert participation. It has great development
potential.
7 CONCLUSIONS
Starting from the time law of scenario evolution of
emergency events, the emergency scenarios were
divided into initial scenario, middle scenario and final
scenario, among which the mutual relationships were
analyzed. The paper explicates how the event-driven
model can play a role in the scenario deduction for
railway emergency, and makes assumptions about the
process of scenario deduction of events. Then, an
event-driven model of scenario deduction for
railway emergency was established. In addition,
taking train derailment accidentals for example, the
Bayesian network of scenario was set up and
analyzed to further deduce the states of node
variables in scenario network. The deduction results
accord with the practical situation. The proposed
scheme has been proved to be feasible and effective
for railway emergency.
Table 6.1: Comparison of three models.
Project
Proposition - based
deduction model
Derivation Model
Based on Case - based
Reasoning
Event - driven
deduction model
Applicable scope
Flexibility
Processing speed
Targeted
Field expert Participation
ACKNOWLEDGEMENTS
This research was supported by the National Natural
Science Foundation of China(Grant No.61763025),
China Postdoctoral Science Foundation funded
project (Grant No.167306) and the Post doctoral
Special Foundation of Lanzhou Jiaotong
University(Grant No.2017003).
REFERENCES
Xu J G. China'srailway emergency response management
system researh. Chengdu: Southwest Jiaotong
University, 2010.
Wang Y H, Jia L M, Q Y, etc. Railway emergency
management ideas and methods. Beijing: China
Railway Press, 2011.
Jiang H, H J S. Unconventional emergency disposal scheme
based on scenario reconstruction method for generating
rapid research. China Emergency Management, 2012,
(1): 14-20.
An Event-Driven Scheme of Scenario Deduction for Railway Emergency
129
Li D, Chen J G, Chen T, Yuan H Y. Event chain
probabilistic model of emergencies. Journal of
Tsinghua University (Science and Technology),
2010,50 (8): 1173-1177.
Rong L L, Zhang J Y. Study on different evolution modes
of emergencies. Journal of Natural Disasters, 2012,21
(3): 1-6.
Hu X X. Bayesian network modeling technology and its
application in decision-making. Hefei: Hefei
University of Technology, 2006.
Fang Z G, Yang B H, Lu Z P, etc. Study on GERT network
model of disaster evolution based on Bayes reasoning.
China Management Science, 2009,17 (2): 102-107.
Xu L, Li X Y, Huang X Y. Based on bayesian network of
unconventional emergency assessment. Journal of
Shanghai jiaotong university (natural science edition),
2013,47 (5) : 846-850.
Shu Q L. Study on scenario evolvement and alternative
generation of scenaio-response decision-making
in unconventional emergencies. Journal of university of
science of technology of China, 2012, 42(11): 936-941.
Chen X Q, Jia L M, Qin Y, etc. Research on railway
emergency management. Journal of the China railway
society, 2012, 34(3): 7-13.]
Zheng Y J, Zhang M X, Ling H F, et al. Emergency
Railway Transportation Planning Using a Hyper-
Heuristic Approach. Intelligent Transportation
Systems,2014:1-9.
ICECTT 2018 - 3rd International Conference on Electromechanical Control Technology and Transportation
130