fire science unit VIII PowerPoint presentation

Escape behavior in factory workshop fire emergencies: a multi-agent simulation

Kefan Xie • Jia Liu • Yun Chen • Yong Chen

Published online: 24 May 2014

� Springer Science+Business Media New York 2014

Abstract In this study, a multi-agent simulation is con-

ducted to explore the relationship between fire escape

survival rate and occupants’ risk preferences and stress

capacity. The results indicate that, the escape survival rates

for occupants with different risk preferences and stress

capacities can be significantly different. More specifically,

the simulation shows that the smaller the number of

occupants is in a fire, the higher the survival rate can be

expected. In addition, the simulation shows that the larger

the number of individuals with stronger stress capacities is

in a group, the higher the escape survival rate the group

has. Moreover, the simulation shows that the more disperse

the individuals’ risk preferences is in a group, the higher

the escape survival rate the group has. Based on the sim-

ulation results, the paper proposes a framework of

E-evacuation system to guide the rational escape and

evacuation when enterprise workshop fire occurs. Sugges-

tions for increasing escape survival rates during fires are

provided.

Keywords E-evacuation � Risk preference � Stress ability � Multi-agent system � Fire � Evacuation � Escape survival rate

1 Introduction

Fire is one of the most common emergencies. It may cause

huge property damages and casualties, especially in places

such as cinemas and factories. In such places, it is extre-

mely hard for occupants to escape. If occupants have not

received any escape training before or a well prepared

escape plan is not available, they can hardly survive. This

is simply because in such emergent conditions, occupants

have to make a correct decision right away and take proper

actions immediately. The correct escape decision and

proper actions in scene of a fire means the difference

between life and death.

In the past decades, scholars have explored occupants’

decision making and escape behaviors in emergencies. For

example, Kelley and Condry [13] find that the more the

losses are, the lower the escape survival rate is; and that the

bigger a group is, the fewer group numbers escape suc-

cessfully. They also find that if the behaviors of group

members are guided by each other, the escape results will

change; and that a positive expression of a group’s confi-

dence will substantially increase the group’s success

escape rate. Helbing et al. [10] explore the irrational

characteristics of the group panic escape behaviors in

emergencies. Saloma et al. [23] study the self-organized

queuing and scale-free behavior in real escape panic. Joo

et al. [11] conducted agent-based simulation of affordance-

based human behaviors in emergency evacuation. Lv et al.

[16] examine evacuation decision-making behaviors and

risk analysis under multiple uncertainties in emergencies.

Ozbay et al. [18] model the emergency evacuation in

Northern New Jersey based on a regional transportation

planning tool. Pereira et al. [19] employ a finite automata

approach to simulate the evacuation of a congested popu-

lation in emergency. Sun et al. [29] develop an emergency

K. Xie � J. Liu � Y. Chen (&) School of Management, Wuhan University of Technology,

Hubei 430070, People’s Republic of China

e-mail: [email protected]

Y. Chen

Old Dominion University, Norfolk, VA 23529, USA

123

Inf Technol Manag (2014) 15:141–149

DOI 10.1007/s10799-014-0185-1

evacuation information system in order to make emergency

decision-making more effective. Numerous other authors

have studied emergency situations [5, 6, 24, 25, 33, 34, 36].

All of these studies focus on observing occupants’

escape behavior characteristics directly. However, they did

not pay attention to how occupants’ risk preferences and

stress capacity impact their escape behaviors. More spe-

cifically, few researches explore how occupants’ risk

preferences and stress capacity impact escape survival rate

in fires. Risk preference represents an individual’s attitude

to risks. Engelmann and Tamir [4] prove that individuals’

risk preferences have influences on their decision-making

processes with neuroscience methods. Risk preferences can

be divided into risk loving and risk avoiding. According to

Abrahamsson and Johansson [1], the majority of social

emergencies resulting in death are related with risk seeding

behaviors. Their further study points out that individual

risk preferences are influenced by circumstances and that

group risk preferences exist. Stress capacity refers to an

individual’s ability to deal with stress when he/she

encounters critical, complex, and difficult situations [31].

Stress capacity can impact the quality of decision [3] and

decision-making process [14, 17]. As such, this paper

adopts a multi-agent simulation to simulate occupants’

emergency escape decision-making process and actions

during a fire in the context of labor-intensive factory

workshops. The goal of this paper is to explore the rela-

tionship between escape survival rate and occupants’ risk

preferences and stress capacity.

The rest of this paper is organized as follows: In Sect. 2,

background of fires occurring in factory workshops is

provided. Occupants’ escape behaviors and the main

challenges for their decision-making during a fire are dis-

cussed as well. Section 3 provides an overview of the

multi-agent programming language and modeling envi-

ronment that this paper adopts to simulate factory work-

shop fires. Parameters settings are introduced in this

section. Section 4 lists the behavior rules for agents in this

simulation. In Sect. 5, the operation process of the per-

formed simulation is introduced. Results of the simulation

are discussed in Sect. 6. Based on the simulation, a

framework of E-evacuation system for enterprise workshop

fire emergency is proposed in Sect. 7. At the end, Sect. 8

concludes the whole paper and provides suggestions for

increasing escape survival rates during fires.

2 Background

Occupants in factory workshops must have certain risk

preferences and stress capacities in order to make the

correct decision and take proper actions to escape because

fires occurring in factory workshops are very dangerous

and might cause unexpected losses. The texture and

quantity of materials stored in workshops will affect the

speed at which a fire spreads. Specifically, whether the

materials in a workshop are flammable and the degree of

flammability play important roles in fire spreading speeds:

the higher the degree of flammability is, the quicker a fire

spreads. For example, textile workshops are usually full of

combustible materials. Once a textile workshop is on fire,

these materials will cause the fire to spread quickly. As a

result, workers have very little time to escape and suffer

great pressure. Fires occurring in chemical workshops are

more dangerous because other than flammable materials,

toxic substances and explosives are usually stored in these

places. The opportunity for workers to escape from such

fires is pretty slim. In addition, the amount of materials in a

workshop can directly affect the spreading speed of a fire

and occupants escaping routes. Large amount of materials

means many flammable things, long burning time, and

large scale of fire. Meanwhile, the stacked material can

block occupants escaping routes. This will slow down the

escaping speed. Furthermore, the logistic model chosen by

a company will affect the inventory in a workshop. Spe-

cifically, the options of first-party, second-party or third-

party logistics by a company will significantly influence the

changing speed of workshop inventory per unit time, which

thereby affects the patency of occupants escaping routes.

Finally, workshop building design is an important factor

that impact workers’ escape. The lack of exits will cause

congestion during the evacuation in a fire.

According to Chu and Sun [2], how long occupant

evacuation last depends on ‘‘fire detection and alarm,

occupant characteristics (such as age, sex, physical and

mental ability, sleeping or waking, and population density),

human behavior in fire (such as seeking information,

informing others, collecting belongings, and choosing an

exit) and building characteristics (such as corridor width,

exit numbers and widths), etc.’’ (p. 1126). Occupants’

escape behaviors vary in factory workshop fires. For

example, on August 27, 2011, Bao Dongsheng Plastic

Products Factory in Longgang District, Shenzhen, China

was on fire. Disconcerted workers looked for ways to

escape. In the chaos, two flustered workers jumped from

the workshop building, one dead and another seriously

injured. On December 14, 2009, a workshop at a third floor

with 45 workers in Runsen Shoe Factory in FuZhou, China

caught fire. Because the fire blocked the emergency exit,

workers had to go upstairs and jumped from the 4th floor,

14 seriously injured. In contrast, workers in a workshop fire

breaking out in Ningbo China escaped successfully by

hitting a hole in the wall when the emergency exits was

blocked by the fire.

Fire escape behaviors in factory workshops are the

results of occupants’ decision making that are influenced

142 Inf Technol Manag (2014) 15:141–149

123

by many variables, such as interactions between the

occupants, the building, and the developing fire [22]. Dif-

ferent people have different levels of awareness of what is

happening in their surrounding environment [20]. Once

they realize that a fire occurs, occupants make their deci-

sions based on the result of their risk assessment. For

example, they can put out the fire first and then escape, or

escape first and come back later with help to put out the

fire, or take an active but risky way to rush out of the fire

scene, or just wait for rescue. Table 1 shows the main

problems in occupants’ decision making process in a fac-

tory workshop fire.

The key point is that occupants must make their deci-

sions immediately. Sime [28] points out that delays in

occupants starting to move and movement other than

escape could be a major feature of human behavior in fires.

Shields and Boyce [26] indicate that a primary factor

contributing to fire deaths is not travel distance to exits, but

delays in warning occupants and extended times before

movement commenced. According to Proulx and Sime [21]

and Sime [28], the delay in starting positive evacuation

actions can be much longer than the time to travel the

distances to and through exits. Occupants do not have

much time to think over the options listed in Table 1. The

condition of workshop can easily boost quick spread fires.

Panic occupants will worsen the chaos at fire scene. Even

worse, emergency exits might be blocked by fires. In such

emergent situations, occupants’ individual risk preferences

and stress capacities play important roles in their decision-

making processes.

Prior research has proved that training, more specifically

risk preference and stress capacity trainings, impact indi-

viduals’ decision making and behaviors in a fire. Fire drill

can help occupants make correct decisions and take proper

actions during a fire. For example, on April 1, 2010, a

2000 m 2

workshop with 1,078 workers in YierKang Shoe

Company in Zhejiang, China was on fire. Thanks to the

regular fire evacuation drills, these workers covered wet

towels on their mouths, followed the scheduled escape

route, and fled the fire in an orderly manner. Fortunately,

the fire did not cause any casualty.

3 Simulation and parameters setting

This study conducts a simulation based on Netlogo to

explore the relationship between escape survival rate and

occupants’ risk preferences and stress capacities. NetLogo

is a multi-agent programming language and modeling

environment for simulating natural and social phenomena

[32]. It allows researchers to give instructions to huge

number of independent ‘‘agents’’ who are operating con-

currently. In this regard, NetLogo can help researchers

‘‘explore connections between micro-level behaviors of

individuals and macro-level patterns that emerge from their

interactions’’ [32]. Figure 1 shows the interface of the

established multi-agent simulation applied in this study.

The simulation consists of machinery, equipment, walk-

ways, workshop facades, workshop exits, and safety zones.

Suppose that workshop exterior walls, machinery, and

equipment are taller than an ordinary person’s height and

that no one can stand or walk on the top of the machinery,

equipment or workshop exterior walls.

The simulated event is a fire that occurred at KPT Fit-

ness Equipment Co., Ltd. in Dongguan China. Established

in 1993, KPT has solid experiences in indoor fitness

equipment R&D and manufacturing. It specializes in pro-

ducing electric treadmills, exercise bikes, and elliptical

machines. On March 6, 2012, welding sparks ignited a fire

in one of its workshop on a second floor. Very soon the fire

burned down all windows. Raw materials, such as foam,

burned quickly and stuck together. This caused the work-

shop full of a pungent odor. Although 256 occupants were

trapped, they were well organized and escaped in order.

Fortunately, they all survived in this terrible fire.

The multi-agent simulation accepts different parameters

within a certain range. Therefore, various scenarios in

factory workshop fires can be simulated. For example, the

position of the start point of a fire is controlled by hori-

zontal axis (Fire-Start-Point-X [ [-16, 16]) and vertical axis (Fire-Start-Point-Y [ [-16, 16]). The speed at which a fire spreads is controlled by fire spread speed (Fire-Spread-

Speed [ [1, 20]). People count (People-count [ [1, 300])

Table 1 Main problem in occupants’ decision making process in a factory workshop fire

Decision making

problems

Initial decision Subsequent decision

General decision 1. Fight against

the fire

2. Escape and

evacuation

1. Choose fire fight first, when

fire cannot be weakened,

even bigger and more

dangerous, choose to escape

2. Escape successful and

come back to fight against

the fire

Self-save and

waiting for

rescue decision

1. Be active to

escape

2. Wait for the

rescue

1. Not successful in escape,

then wait for the rescue

2. No result in the external

rescue, then seek for self-

rescue

Channel (exit)

selection

decision

1. Follow other

occupants to

escape

2. Choose a

nearby exit to

flee

When exit is blocked, choose

some means, such as

smashing windows, hitting

hole in walls, to flee through

the broken exits

Inf Technol Manag (2014) 15:141–149 143

123

represents that the number of occupants that are evenly

distributed in the walkways. Stress capacity proportion

(Stress-Capacity-Proportion [ [1, 10]) is the ratio between the number of occupants with strong stress capacities and

the reciprocal of the value of Stress-Capacity-Proportion.

Risk preference proportion (Risk-Preference-Proportion [ [1, 10]) is the ratio between the number of occupants who

are risk loving and the reciprocal of the value of Risk-

Preference-Proportion. Results vary if different parameters

are set.

4 Behavior regulations for agents

Each unit area in a factory workshop, such as safety zone,

exterior wall, exit, walkway, machine, equipment, and other

fixed facilities, is set as Patch, whereas each occupant is set as

Turtle in this simulation. A Turtle has four attributes, namely

life value, risk preference, stress capacity, and escape direc-

tion. Risk preference has two values: ‘‘0’’ and ‘‘1’’. The former

represents risk loving while the latter means risk avoiding.

Similarly, stress capacity has two values: ‘‘0’’ and ‘‘1’’ as well.

The former means weak stress capacity and the latter repre-

sents strong stress capacity. Escape direction have four values,

namely ‘‘1’’, ‘‘2’’, ‘‘3’’, ‘‘4’’. These four values indicate the

directions of the exits that occupants head for and they rep-

resent northwest, northeast, southeast, and southwest respec-

tively. During a factory workshop fire, occupants’ eyesight

might be blocked by smoke or equipment easily. In such an

urgent situation, it is very hard for them to make correct

judgments about what is the shortest way to the nearest exit. In

Netlogo simulation, one Patch usually accommodates one

Turtle. But in factory workshop fires crowded occupants are

trapped in limited spaces. Therefore, in this simulation, a

Patch is set to accommodate two Turtles.

Gwynne et al. [9] and Sime [27] point out that instead of

heading towards the nearest exit, occupants prefer to move

toward other more distant exits with which they have had

previous experience and with which they feel more confi-

dent. However, in this simulation, if occupants are familiar

with all formal exits, those who have strong stress capacity

are more likely to calmly observe their surroundings and to

find the shortest way to the nearest exit, whereas those who

have weak stress capacity tend to run about aimlessly

because they randomly choose one exit to escape. In such a

case, stress capacity impact occupants’ decision processes.

However, risk preferences do not have any impact. There-

fore, the study first sets the following rules for Turtles’

decision making in the simulation:

Rule A1: If Turtles are familiar with all formal exits,

those with strong stress abilities choose to escape from

the nearest one;

Rule A2: If Turtles are familiar with all formal exits,

those with weak stress abilities randomly choose one to

escape.

Fig. 1 The interface of system simulation

144 Inf Technol Manag (2014) 15:141–149

123

In some cases, alternative exits exist but occupants do

not know where they are. Since those with strong stress

abilities prefer to find a nearest exit and those with weak

stress abilities randomly select an exit, all of them have to

face a challenge if their preference is an alternative exit.

Alternative exits are not used often and few occupants

know how to access them and how they work, so it is very

risky to pick an alternative exit during a factory workshop

fire whether this exit is a nearest one or a randomly selected

one. Due to the risks, alternative exits are not crowded as

formal ones. In such a case, risk loving occupants tend to

choose alternative exits, whereas risk avoiding occupants

prefer formal ones. Combining the effects of stress abilities

and risk preferences, the paper continues to set the fol-

lowing rules for Turtles’ decision making in the simulation:

Rule B1: If Turtles only are familiar with formal exits,

those risk loving Turtles with high stress capacities

choose to escape from the nearest alternative exit;

Rule B2: If Turtles only are familiar with formal exits,

those risk avoiding Turtles with high stress capacities

choose to escape from the nearest formal exit;

Rule B3: If Turtles only are familiar with formal exits,

those risk loving Turtles with low stress capacities

randomly choose an alternative exit to escape.

Rule B4: If Turtles only are familiar with formal exits,

those risk avoiding Turtles with low stress capacities

randomly choose a formal exit to escape.

5 Simulating operation process

Once the parameters are set, the initialization of the sim-

ulation is done. In the initialization stage, Patches are built

up, including machinery, equipment, walkways, workshop

facades, workshop exits, and scenes of safety zones. Preset

Turtles are randomly distributed to the walkways. In

addition, the attributes of each Turtle are configured based

on preset risk preference proportion and stress ability

proportion values.

Next, the simulation moves to the running stage. Turtles

follow the established rules and select an exit first. On one

hand, they need to calculate whether the space in front of a

Patch is big enough to accommodate them. If yes, they can

move on to the next step. If no, their next movement

depends on their positions. For example, if a Turtle is at a

crossing, it can choose a different direction. Otherwise, it

has to take a step backward. When a Turtle arrives at a

target exit, the system will automatically set its location as

a safe area. Accordingly, observation on this Turtle is done.

On the other hand, Turtles need to figure out the intensity

of the fire. The fire will spread at a preset speed, so in a

given period of time a Patch will catch fire. When the fire

reaches a Turtle’s Patch, the turtle’s life value will minus 1.

If a Turtle’s life value is less than 0, the Turtle is dead. The

fire can reach any place inside the workshop other than safe

areas. When all areas inside the workshop are covered by

fire, the simulation ends (Fig. 2).

6 Analysis of simulation results

The simulation has three phases. In phase 1, the goal is to

explore how fast the spread speed of fire (FSS) is and the total

number of occupants (PA) that impacts the final escape sur-

vival rate (APA). So risk preference proportion (RPP) and

stress ability proportion (SAP) are both set as 2. Figure 3

shows results of four typical cases, in which FSS are set as 2, 5,

20, and 5 respectively and PA are set as 300, 300, 300, and 150

respectively. All four diagrams indicate that as time goes by

during fires, the number of survival occupants declines. This

means that the longer fires last, the fewer occupants can sur-

vive. According to diagram (1), diagram (2), and diagram (3),

lower speeds of fire cause fewer numbers of deaths. Diagram

(2) and diagram (4) show that the smaller the total number of

occupants is, the higher the survival rate is. Fewer occupants

will cause less congestion, so occupants have less troubles and

can escape faster. This result proves the finding reported by

Kelley and Condry [13].

In phase 2, the goal is to explore how stress capacity

ratio impacts the final escape survival rate (APA). Risk

preference ratio is the proportion of the risk preferences of

individuals in a group. Stress capacity ratio refers to the

proportion of the stress capacities of individuals in a group.

The two ratios can be calculated after risk preference

proportion (RPP), stress ability proportion (SAP), and total

number of occupants (PA) are set. In this phase, fire-caught

point is set as the start point (0, 0) and spread speed of fire

(FSS) is set as 5.

Tables 2, 3 and 4 show three cases, in which SAP are set

as 0.1, 0.5, and 1.0 respectively and PA are set as 100, 200,

and 300 respectively. Suppose that occupants are familiar

with all four formal exits, and that they show the analog

value of the escape survival ratio in the group with dif-

ferent individual stress capacity. In each case, the simula-

tion runs ten times and generates ten different values of

escape survival rate. The final escape survival rate is the

arithmetic average of these ten values. The three tables

show that the more individuals with strong stress capacities

are in a group, the higher escape survival rate this group

has. This is because individuals with strong stress capaci-

ties are capable of making independent judgments and

choosing a nearest exit, while those with weak stress

capacities are not able do so.

When high stress ability proportion of individuals is

close to 100 %, the highest ratio of escape survival can be

Inf Technol Manag (2014) 15:141–149 145

123

achieved. This is because in such a case everyone escapes

from the nearest exit. When all occupants are evenly dis-

tributed to exits, the selection of each exit is approximately

same. This is the way to minimize congestion and to

achieve the best escape survival rate. Meanwhile, the tables

also show that the larger the number of occupants is, the

smaller the escape survival ratio will be eventually.

In phase 3, the goal is to explore how risk preference

ratio and stress ability ratio interactively impact the final

escape survival rate (APA). Most settings are the same as

those in phase 2. The only difference is that phase 3 adds

risk preference proportion (RPP) to the setting list. In three

cases showed in Tables 5, 6, and 7, RPP are set as 0.1, 0.5,

and 1.0 respectively. Suppose that occupants are familiar

with both two formal exits, and that they show the analog

value of the escape survival ratio in the group with

different individual stress ability. In each case, the simu-

lation runs ten times and generates ten different values of

escape survival rate. The final escape survival rate is the

arithmetic average of these ten values.

Tables 5, 6, and 7 demonstrate that when alternative

exits exist, the finding in phase 2 is still valid. The more

individuals with strong stress capacities are in a group, the

higher escape survival rate this group has. In addition,

when the group risk preference ratio is set as 0.5, the

highest escape survival rate is achieved. This is because in

this setting the number of risk loving occupants is the same

as the number of risk avoiding ones. Risk loving occupants

choose alternative exits while risk avoiding ones chooses

formal exits. Occupants’ selections drive them to different

exits and congestions are minimized, so the final escape

survival rate rises. This result proves the findings reported

Establish Circumstances

Patch

Randomly Distributed Turtle

Setting Turtle’s Atributes

Low stress ability Turtle choose exits

at random

T urtle

fam il iar

w ith

all exi ts?

High stress ability Turtle choose the

nearest exit

Turtle Choose Escape Exit

High stress ability and risk loving

Turtle choose the nearest spare exit

Low stress ability and risk averse

Turtle choose the familiar exit at

random

Low stress ability and risk loving

Turtle choose the spare exit at random

High stress ability and risk averse

Turtle choose the nearest familiar exit

The fire spread to an additional area

The fire not spread

The intensity of fire gets more serious?

Turtle’s action is according to the

rules

Full of fire everywhere

End of Simulation

Turtle is in walkway?

Turtle, no action

Life - 1

Turtle’s walkway caught fire?

Life, no change

Turtle’s life is less than 0?

Turtle is dead Previous Patch has

Two Turtle?

Go one step further

Change to another direction

Turtle is at the crossing?

Draw one step back

Turtle is at the exit?

Configure Turtle to a safe area at random

Y N

Y

N

Y N

Y N

N

Y N

Y N

Y

Y

Y

N

T urtle’s

A cti o n

R ule

T urtle

E s cape

D ecision

R ul es

S ystem

Ini tial S

etti ng

F ro m

the second

tim e

p hrase to

the N

tim e

phr a se

T h e

firs t tim

e- step in

S ystem

O p erat io n

Purchase amount

Logistics category

Cargo flammability

Fires preading speed

Logistics category

Fig. 2 Simulation operation process

146 Inf Technol Manag (2014) 15:141–149

123

by Abrahamsson and Johansson [1]. When stress capacity

ratio is set as 1, and risk preference proportion is set as 0.5,

a comparatively high escape survival rate can be achieved.

But it is still lower than the rate when all occupants know

all formal exits. This is because when risk loving/avoiding

occupants with high stress capacities choose a formal/

alternative exit that is nearest to them, it is better for them

to know all exits. If they do not, the one they choose might

not be the nearest one.

7 A framework of E-evacuation system for enterprise

workshop fire emergency

Risk management of enterprise workshop fire emergency

covers each stage of fire management, namely beforehand,

concurrent, and afterwards. The key part is how to prepare,

organize, and conduct an orderly evacuation. Information

technology plays an important role in evacuation prepara-

tion, organization, and implementation. As such, this paper

proposes an enterprise workshop fire E-evacuation system

Fig. 3 Some typical simulation results

Table 2 The 100 occupants case (all occupants are familiar

with four exits)

SAP 0.1 0.5 1.0

APA 78.1 87.1 99.5

Table 3 The 200 occupants case (all occupants are familiar

with four exits)

SAP 0.1 0.5 1.0

APA 153.0 174.2 197.5

Table 4 The 300 occupants case (all occupants are familiar

with four exits)

SAP 0.1 0.5 1.0

APA 181.4 221.4 271.3

Table 5 The 100 occupants case (all occupants are familiar

with two exits)

RPP SAP

0.1 0.5 1.0

0.1 77.2 84.8 88.7

0.5 80.6 87.1 94.8

1.0 79.9 81.8 89.9

Table 6 The 200 occupants case (all occupants are familiar

with two exits)

RPP SAP

0.1 0.5 1.0

0.1 106.5 124.5 139.5

0.5 144.8 150.9 156.7

1.0 102.3 112.1 126.7

Table 7 The 300 occupants case (all occupants are familiar

with two exits)

RPP SAP

0.1 0.5 1.0

0.1 135.6 135.9 164.7

0.5 164.8 183.7 187.4

1.0 119.2 130.1 145.9

Inf Technol Manag (2014) 15:141–149 147

123

shown in Fig. 4. First, enterprises should conduct regular

workshop fire risk evaluation and identify potential risks. A

workshop fire risk early-warning system should be estab-

lished. Once a workshop fire potential risk is found, an in-

time warning should be available. Second, when a work-

shop fire occurs, the E-evacuation system should send out a

fire alarm immediately, start automatic spraying systems,

start electronic contingency plans, and mobilize timely the

fire brigade to put out the fire. Meanwhile, workshop

occupants should make correct decisions and take proper

actions to evacuate. When a fire alarm rings, safe exit signs

should be on.

8 Conclusions

Modeling and simulation has many application domains [7,

8, 12, 15, 30, 35]. This paper conducts a multi-agent simu-

lation to explore the relationship between escape survival

rate and occupants’ risk preferences and stress capacities.

The results verify that the escape survival rates of occupants

with different risk preferences and stress capacities are sig-

nificantly different. More specifically, the simulation proves

the finding in Kelley and Condry [13] that the smaller the

total number of occupants is in a fire, the higher the survival

rate is. In addition, the simulation shows that the more

individuals with strong stress abilities are in a group, the

higher escape survival rate this group has. Moreover, the

simulation shows that the more disperse the individuals’ risk

preferences is in a group, the higher the escape survival rate

this group has. These results prove the findings in Abra-

hamsson and Johansson [1] and Kelley and Condry [13].

The following suggestions can be made from the simu-

lation results to increase escape survival rates during fires.

First, make sure that every exit works properly and that

occupants are familiar all available exits. Second, provide

occupants fire escape training so that their stress abilities can

be improved. Third, if alternative exits are available, move

risk-loving workers to work close to these exits. Fourth,

conduct fire drills and let occupants be familiar with evac-

uation process. These suggestions can be applied not only to

enterprise workshop fire prevention and evacuation, but also

to emergency management in other densely populated

places.

In addition, enterprises can take the following measures to

manage their fire evacuation process: (1) Help employees

identify safe exit signs. Safe exit signs usually are in green

lights indicating directions in walkways. By following the

green lights, employees can find the exits; (2) Maintain

smooth communication. When fire occurs, emergency notice

should be immediately broadcasted to every corner of the

workshop. The start point of fire should be identified

immediately and reported to the manager. If a fire is out of

workers’ control, emergency call should be made right away

for outside help. The manager(s) should assign workers

immediately to shut down equipment and fight against fire.

Meanwhile, the manager(s) should move important docu-

ments and data to safe places; (3) Organize workers to escape

orderly. When a workshop is on fire, the team leaders should

keep calm, count team members, and let them stay where

they are. Then each team should find the nearest safe exit sign

and follow the directions to escape. Workers on the first floor

can jump out of windows directly if the way to exit is

blocked. Workers on the second floor need to hang on the

windowsill first to minimize the height to ground and then

jump. Workers on the third floor and above should close

doors and irrigate them so that the fire can be separated. They

should use a damp cloth mask to prevent inhalation of toxic

gas, and make noises let firefighters know where they are.

They also need to lower their bodies close to the floor

because air in higher position is not good for breath. Occu-

pants should not consider jumping to the ground because it is

easy to get serious hurt or even die. If the condition allows,

occupants can help themselves escape. Otherwise, they

should remain in a safe place and wait for help. When a fire

occurs, elevators should not be taken because elevators are

very easy to get stuck due to power outage. Furthermore,

elevator shafts often become chimneys during fires. It is very

risky and dangerous to use elevators during fires. If there are

ladders, occupants can use them and climb to the roof sur-

faces waiting help there.

Acknowledgments This study was supported by National Natural Science Foundation of China (90924010) and Independent Innovation

Research Fund of Wuhan University of Technology (2013-lv-002).

References

1. Abrahamsson M, Johansson H (2006) Risk preferences regarding

multiple fatalities and some implications for societal risk decision

making—an empirical study. J Risk Res 9(7):703–715

Fire Emergency Risk Evaluation

Fire Emergency Risk Early-warning

Fire Emergency Alarm

Start Electronic Predominated Plans

Start Automatic Spraying Systems

Start the Electronic Display of the Evacuation Paths and Exits

Start the Fire Brigade Start Internal Workshop

Fire Self-help Carry out an Orderly

Evacuation

Fig. 4 Framework of E-evacuation system for enterprise workshop fire emergency

148 Inf Technol Manag (2014) 15:141–149

123

2. Chu G, Sun J (2008) Decision analysis on fire safety design based

on evaluating building fire risk to life. Saf Sci 46(7):1125–1136

3. D’Zurilla TJ, Sheedy CF (1991) Relation between social prob-

lem-solving ability and subsequent level of psychological stress

in college students. J Pers Soc Psychol 61(5):841–846

4. Engelmann JB, Tamir D (2009) Individual differences in risk

preference predict neural responses during financial decision-

making. Brain Res 1290:28–51

5. Fang S, Xu L, Pei H, Liu Y (2014) An integrated approach to

snowmelt flood forecasting in water resource management. IEEE

Trans Ind Inform 10(1):548–558

6. Fang S, Xu L, Zhu Y, Ahati J (2014) An integrated system for

regional environmental monitoring and management based on

internet of things. IEEE Trans Ind Inform. doi:10.1109/TII.2014.

2302638

7. Feng S, Xu L (1996) Integrating knowledge-based simulation

with aspiration-directed model-based decision support system.

Syst Eng Electron 7(2):25–33

8. Gao Q, Xu L, Liang N (2001) Dynamic modeling with an inte-

grated ecological knowledge-based system. Knowl-Based Syst

14(5–6):281–287

9. Gwynne S, Galea ER, Owen M, Lawrence PJ (1999) Escape as a

social response. Society of Fire Protection Engineers, Boston

10. Helbing D, Farkas I, Vicsek T (2000) Simulating dynamical

features of escape panic. Nature 407(6803):487–490

11. Joo J, Kim N, Wysk RA, Rothrock L, Son YJ, Oh YG, Lee S

(2013) Agent-based simulation of affordance-based human

behaviors in emergency evacuation. Simul Model Pract Theory

32:99–115

12. Kataev M, Bulysheva L, Emelyanenko A, Emelyanenko V (2013)

Enterprise systems in Russia: 1992–2012. Enterp Inf Syst

7(2):169–186

13. Kelley HH, Condry JC Jr (1965) Collective behavior in a simu-

lated panic situation. J Exp Soc Psychol 1(1):20–54

14. Kobasa SC (1979) Stressful life events, personality, and health:

an inquiry into hardiness. J Pers Soc Psychol 37(1):1–11

15. Li N, Yi W, Bi Z, Kong H, Gong G (2013) An optimization

method for complex product design. Enterp Inf Syst

7(4):470–489

16. Lv Y, Huang GH, Guo L, Li YP, Dai C, Wang XW, Sun W

(2013) A scenario-based modeling approach for emergency

evacuation management and risk analysis under multiple uncer-

tainties. J Hazard Mater 246:234–244

17. Maddi SR, Khoshaba DM (1994) Hardiness and mental health.

J Pers Assess 63(2):265–274

18. Ozbay K, Yazici MA, Iyer S, Li J, Ozguven EE, Carnegie JA

(2012) Use of regional transportation planning tool for modeling

emergency evacuation. Transp Res Rec J Transp Res Board

2312(1):89–97

19. Pereira LA, Duczmal LH, Cruz FRB (2013) Congested emer-

gency evacuation of a population using a finite automata

approach. Saf Sci 51(1):267–272

20. Pires TT (2005) An approach for modeling human cognitive

behavior in evacuation models. Fire Saf J 40(2):177–189

21. Proulx G, Sime JD (1991) To prevent ‘panic’in an underground

emergency: why not tell people the truth. Fire Saf Sci 3:843–852

22. Purser DA, Bensilum M (2001) Quantification of behaviour for

engineering design standards and escape time calculations. Saf

Sci 38(2):157–182

23. Saloma C, Perez GJ, Tapang G, Lim M, Palmes-Saloma C (2003)

Self-organized queuing and scale-free behavior in real escape

panic. Proc Natl Acad Sci 100(21):11947–11952

24. Shan S, Wang L, Li L (2012) Modeling of emergency response

decision-making process using stochastic Petri net: an e-service

perspective. Inf Technol Manag 13(4):363–376

25. Shan S, Wang L, Li L, Chen Y (2012) An emergency response

decision support system framework for application in e-govern-

ment. Inf Technol Manag 13(4):411–427

26. Shields TJ, Boyce KE (2000) A study of evacuation from large

retail stores. Fire Saf J 35(1):25–49

27. Sime J (1992) Human behaviour in fire summary report. Central

Fire Brigades Advisory Council for England and Wales, London

28. Sime JD (1994) Escape behaviour in fires and evacuations, design

against fire: an introduction to fire safety engineering design.

Chapman & Hall, London

29. Sun QF, Kong FS, Zhang L, Dang XW (2011) Construction of

emergency evacuation information system based on the internet

of things. In: Proceedings of 2011 international conference on

mechatronic science, electric engineering and computer, Jilin,

China, pp 342–345

30. Tan W, Xu W, Yang F, Xu L, Jiang C (2013) A framework for

service enterprise workflow simulation with multi-agents coop-

eration. Enterp Inf Syst 7(4):523–542

31. Taylor SE (1999) Health psychology. McGraw-Hill, New York

32. Tisue S, Wilensky U (2004) NetLogo: design and implementation

of a multi-agent modeling environment. In: Proceedings of agent

conference on social dynamics and interaction, reflexivity

emergence, Chicago, IL, pp 7–9

33. Wang L, Xu L, Bi Z, Xu Y (2014) Data cleaning for RFID and

WSN integration. IEEE Trans Ind Inform 10(1):408–418

34. Xie K, Liu J, Chen G, Wang P, Chaudhry S (2012) Group

decision-making in an unconventional emergency situation using

agile Delphi approach. Inf Technol Manag 13(4):351–361

35. Xu L (1992) Simulating societal systems. IEEE Potentials

11:18–21

36. Xu S, Xu L, Basl J (2012) Introduction: advances in e-business

engineering. Inf Technol Manag 13(4):201–204

Inf Technol Manag (2014) 15:141–149 149

123

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.