Analysis of evacuation procedures in high speed trains fires

Analysis of evacuation procedures in high speed trains fires

Fire Safety Journal 49 (2012) 35–46 Contents lists available at SciVerse ScienceDirect Fire Safety Journal journal homepage: www.elsevier.com/locate...

1MB Sizes 1 Downloads 90 Views

Fire Safety Journal 49 (2012) 35–46

Contents lists available at SciVerse ScienceDirect

Fire Safety Journal journal homepage: www.elsevier.com/locate/firesaf

Analysis of evacuation procedures in high speed trains fires J.A. Capote, D. Alvear n, O. Abreu, A. Cuesta GIDAI Group – Fire Safety – Research and Technology, University of Cantabria, Ave. Los Castros, s/n 39005 Santander, Spain

a r t i c l e i n f o

abstract

Article history: Received 4 October 2010 Received in revised form 21 September 2011 Accepted 8 December 2011 Available online 20 January 2012

This paper uses egress modelling to explore the impact that crew procedures have on evacuating two high-speed trains under different fire scenarios. The paper begins by analysing an evacuation drill performed by the Spanish Railroad Administration, RENFE Operadora. This analysis is used to obtain input data for the simulations. The second part of the paper analyses the effects of passenger preevacuation activities and train crew procedures (when the fire is detected and the train is still in motion). For each scenario, multiple simulations are performed to capture the stochastic variations in egress times. The results have important implications for rail safety and also show that there are qualitative and quantitative advantages for evacuating under conditions controlled by the train crew rather than using a self-rescue strategy, which allows us to make safety recommendations for managing emergencies. & 2011 Elsevier Ltd. All rights reserved.

Keywords: High speed trains Evacuation procedures Egress modelling

1. Introduction Fire incidents inside passenger trains can constitute a significant risk to life. Therefore, it is necessary to define effective passenger evacuation strategies, both when the fire is detected in the moving vehicle and when evacuating a vehicle that has stopped. The train crew are responsible for passenger safety during an on-board fire emergency. The first priority is to direct passengers away from fire in a controlled manner. The train driver needs to inform the train operations centre about the situation and reach an appropriate place for the evacuation (i.e., the closest station). If this is not possible, the second priority is to perform a controlled evacuation to the tracks. Furthermore, the train crew needs to establish the urgency of the situation and determine how to perform the evacuation. To do this, the crew needs to know the number of passengers and disabled persons on-board, the dangers present inside and outside the train, a safe area where passengers should be moved and which doors need to be opened. In most cases, however, the conditions dictate the specific actions and choice of the appropriate egress strategy. Therefore, it is not easy for the train crew to make the correct decisions. In addition, there may be fewer crew members for the number of passengers than for other modes of transportation, and PA (public address) systems are usually used to issue emergency instructions. In fact, there is less direct contact between crew members and passengers, and this may delay passenger responses. All these factors should be considered when analysing train evacuations.

n

Corresponding author. Tel.: þ34942201826; fax: þ 34942202276. E-mail address: [email protected] (D. Alvear).

0379-7112/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.firesaf.2011.12.008

To improve passenger safety, the train crew needs to be presented with a variety of scenarios for practicing and testing their decision-making skills. Many accident reports have described a lack of training and preparedness in emergency procedures [1]. To address this problem, rail companies perform full-scale drills. These tests have various problems, however, such as their lack of realism and their economic cost. It is well known that single trials produce little information on the variety of potential outcomes seen in evacuation process. In addition, there are standards describing the general requirements for ensuring passenger safety. For example, the ATOC (Association of Train Operating Companies) Vehicles Standard stipulates a passenger evacuation time of 90 s, a minimum flow rate of 30 per/min in evacuation to the track level and a minimum flow rate of 40 per/min to the adjacent coach when the vehicle in question is at the end of the train [2]. These values are called ‘‘magic numbers’’ by some fire engineers. If we consider them as optimal values, they may be achievable but not realistic. The standards do not consider passenger characteristics and behaviours, the train crew’s responsibilities and behaviour or the effects of different procedures in a variety of changing scenarios. In fact, passenger decision-making process can be particularly influential on evacuation procedures during emergency situations in compact spaces such as trains [3,4]. In summary, the properties of passenger train evacuation procedures have not been studied in depth under a variety of conditions. To the authors’ knowledge, there has been little research using computer modelling to analyse crew procedures in transportation systems [5,6], particularly in passenger trains. The current study uses egress modelling and reliable data to examine various evacuation strategies and assess passenger safety. Introducing computer modelling analyses

36

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

Once the train had stopped, the doors were automatically opened by the driver. There was a 35 s delay between stopping the train and opening the doors. It should be noted that the scenario in which the passengers manually open the doors and self-evacuate should only occur in the most extreme situation. Two video cameras were located in passenger coach 09. One camera was located inside the passenger compartment, and the other camera was in front of the exit (see Fig. 2). The footage collected from the evacuation drill was analysed frame by frame using software tools, such as Vegas Pro 9.0, to determine the passenger walking speeds, flow rates, response times and merging flows at the exit door.

could improve evacuation procedures under a variety of conditions. The selected model is STEPS (Simulation of Transient Evacuation Pedestrian movementS) [7], and the technique examines the effects of crew procedures in simulations to analyse their impact on the evacuation process.

2. Data collection There is a wide range of available models for simulating the movement and behaviour of people during emergencies. Some of these models have been applied to evacuation analyses for passenger trains [8–11]. These models rely on a broad range of data for simulating evacuation processes in a given scenario. There is also research about human performance on trains [12,13]. However, there is a need for more data that will significantly increase the capacity to simulate emergency conditions in passenger trains. The input data for the STEPS model were obtained from an announced evacuation drill for a high-speed train that was performed by RENFE Operadora (the Spanish Railroad Administration). The University of Cantabria was not responsible for the organisation of the drill. The evacuation drill took place on 19th September 2009. The train was a high-speed S 130. A total of 218 participants took part in the evacuation drill (73.15% of the maximum load). The participants included staff members and some of their families. Table 1 shows the characteristics of the participants. The evacuation drill consisted of a simulated fire that had started in coach 04 (the lounge coach), and the relocation procedure was performed coach by coach along the length of the train before the train stopped inside the Guadarrama Tunnel. The passengers were divided into two distinct evacuation groups away from the fire (see Fig. 1). The drill procedure followed the preferred method of train evacuation in which the driver stops and then opens the doors onto the platform.

2.1. Flow rate during the relocation procedure Before the train had stopped, the flow rates of 66 passengers from coaches 05, 06, 07 and 08 during the relocation process were measured. Fig. 3 shows the number of passengers passing a certain point on coach 09. There were clearly four stages that corresponded to the different evacuation groups from coaches 05, 06, 07 and 08. The average flow rate calculated from Fig. 3 was 36 per/min. However, if we consider the evacuation rate averaged over the relocation period, the flow rate was 24 per/min. Both these values are lower than the 40 per/min established by ATOC standard. The density level of relocated passengers in the aisle of the coach 09 was 2.55 per/m2. 2.2. Walking speeds within the aisle The walking speeds of the passengers in the aisle during the relocation procedure were also measured.

No of passengers

70

Table 1 Characteristics of the participants. Gender/age groups

%

Male younger than 40 years old Male older than 40 years old Total

47 34 81

Female younger than 40 years old Female older than 40 years old Total

16 3 19

11

10

09

60 50 40 30 20 10 0 0

20

40

60

80 100 120 140 160 180 Time [s]

Fig. 3. No. of passengers passing a certain point in coach 09 during the relocation procedure.

08

07

06

05

Fire

03

02

Evacuated coaches Fig. 1. Layout of the relocation procedure during the evacuation drill.

Fig. 2. Location of video cameras during the evacuation drill.

01

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

37

and a tread of 25 cm) of 0.47 per/s. Furthermore, the flow rate observed during the drill was lower than values from other studies [11] due to the height of train steps (0.25 m) and the gap. This result suggests that the exit/door width, its design and how passengers respond to it have to be considered [15]. Fig. 7 shows how each passenger required different times to negotiate the train steps. It should be noted that passengers exhibited hesitation at the exit before negotiating the train steps and that

The data were obtained from passengers moving in a single queue without being constrained by others while the train was still in motion (see Fig. 4a). The sample of the walking speeds fit a normal distribution, as assessed by the K2 D’Agostino test, with a mean value of 0.99 m/s and a standard deviation of 0.20 m/s. Fig. 4b shows the histogram and the curve distribution. 2.3. Time to prepare for evacuation The passengers were informed about the emergency over the PA system before the train had stopped. In this case, the response times were dependent on the available space to move inside the coaches (i.e., space available to access the aisle). When the train stopped, the aisles were occupied by the relocated passengers. Once they decided to start the evacuation procedure, however, some of the seated passengers spent time performing other actions, such as preparing themselves for the evacuation (e.g., donning jackets, collecting belongings) or waiting for others. The distribution of the time spent by passengers to prepare for the evacuation is presented in Table 2.

Table 2 Distribution of time to prepare for the evacuation. Variable

l

r

Max.

Min.

Time to prepare (uniform dist.) (s)

12

8.0

26

1.5

No passengers

4

2.4. Merging flows at exit door of coach 09 Results of merging relations are displayed in Fig. 5. A total of 28 passengers from coach 10 used the exit door of coach 09, and 16 passengers from coach 09 used this same exit. Significant deference behaviours were observed, and male participants deferred to allow females to first proceed to the exit. It should be noted that during 15–20 s of the evacuation process, a young male from coach 09 deferred to allow 4 participants (including 2 females) to proceed to the exit. The same behaviour was observed in another male participant from the same coach (coach 09), who deferred to allow 2 females and 1 male participant to proceed to the exit between 30 and 35 s. The average merge ratio between coach 09 and coach 10 was 36.4:66.6. Between 35 s and 50 s, however, the merge ratio was 50:50. During this time, the density levels increased, and passengers alternated their access to the exit door.

3 2 1 0 Time intervals [s] From Coach 09

From Coach 10

Fig. 5. Merge ratios between coach 09 and coach 10 towards the exit door of coach 09.

Flow [per/s]

1.5

2.5. Flow rate at the exit door of coach 09 The passenger flow rate at the exit door of coach 09 was measured. This door was 0.81 m wide (0.51 m of the effective width). Fig. 6 shows the flow rate of the evacuation drill compared to the SFPE maximum specific N&M (Nelson and Mowrer) flow rates [14]. An average flow rate of 0.57 per/s was obtained during the overall evacuation process. This value was between the SFPE maximum doorway flow rate of 0.66 per/s and the SFPE maximum flow rate for stairs (with a riser of 19.1 cm

1 0.5 0 0

10

20

30

40 50 Time [s]

Drill SFPE N&M Doorway

60

70

80

Average SFPE N&M Stair

Fig. 6. Comparative of the flow variations through the exit of coach 09 and the maximum constant flow rates N&M (Nelson and Mowrer).

Frequency

24 18 12 6 0 0

0.5 1 1.5 Walking speed [m/s]

Fig. 4. (a) Passengers walking during evacuation drill. (b) Walking speed distribution.

2

38

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

2 of the 46 passengers who negotiated the exit of coach 09 were persons with mobility problems who required more than 3 s to access the platform.

3.2. Selected trains Two high-speed trains are modelled: Train A (see Fig. 9 and Table 3) and Train B (see Fig. 10 and Table 4). The trains are

3. Evacuation modelling analysis 3.1. Required Safe Egress Time (RSET) calculation in trains

Time to negociate the exit [s]

Trains can be in motion when emergencies occur. The Required Safe Egress Time (RSET) calculations should consider the time necessary for the train to stop, the time spent opening the doors and the time spent by the train crew setting up the evacuation elements (i.e., portable ladders or ramps). Therefore, the RSET contains three main components: (1) the detection time (tdet), (2) the time required for the train to stop (tstop) and (3) the evacuation time from the train (tevac). The egress time model for fires on trains is presented in Fig. 8. When a fire is detected aboard a running train, it is necessary to move the passengers to a place of relative safety (other passenger coaches). The time required for a high-speed train to stop can be more than 15 min. For this reason, the passengers may need to be protected from the effects of the fire for several minutes. Therefore, the response strategy should include completing the relocation procedure before the train stops. This strategy leads to formulating the following questions. How much time does the crew take to complete the relocation procedure? How many passenger coaches should be evacuated? In other words, what is the most appropriate strategy? In the next sections, we define the pre-evacuation procedures and explore their impact on the evacuation process.

Fig. 9. Layout of Train A.

Table 3 Configuration of Train A. Coaches No. of seats Exit/side a

12 30 –

11 36 1

10 36 1

09 36 1

08 36 1

07 19 þ 2a 1

lounge – 1

05 24 1

04 26 1

03 26 1

02 21 1

01 24 1

Disabled people (wheelchair users).

5 4 3

Fig. 10. Layout of Train B. 2

Table 4 Configuration of Train B.

1 0

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46

Passengers Fig. 7. Time spent by each passenger to negotiate the train step (exit of coach 09).

Coaches No. of seats Exit/side a

11

10

09

08

07

06

05

lounge

03

02

01

20 –

36 1

36 1

36 1

36 1

36 1

36 1

– 1

22 þ1a 1

26 1

14 –

Disabled people (wheelchair user).

Fig. 8. Egress time model for fires in trains.

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

2.94 m wide. The exit doors are single-leaf sliding, 0.81 m wide structures, and the aisles have maximum widths of 0.69 m in the first-class coaches and 0.52 m in the second-class coaches. Both the trains have one train manager (guard), two employees (contracted by the railroad company) in the lounge coach and one train host in the first class coaches. From the viewpoint of fire safety, both the trains have smoke detectors in each passenger coach (in the electric cabinets but not in passenger compartments) and in the toilets. The on-board equipment in case of evacuation includes one emergency ladder stored in one locomotive and one ramp stored in the lounge coach of Train A and in coach 03 of Train B. It should be noted that the ramp is normally used to transfer passengers from one train to another in cases of train technical failure. Both the emergency ladder and the ramp are 3 m long and consist of two separate parts that have to be assembled. These evacuation elements hold only two passengers simultaneously.

3.3. Scenarios Figs. 11–14 show the evacuation scenarios considered for the simulations. It should be noted that we have used the definitions established by the Health & Safety Executive ‘‘Guide for providing equipment and arranging evacuation and escape from trains in an emergency’’ for evacuation (controlled) and escape (uncontrolled) [16]. These definitions do not consider whether the exit is a normal or abnormal route, as established by ATOC standard. In the fire scenarios, two consecutive dynamic processes were simulated (1) evacuating passengers to a place of relative safety along the train (pre-evacuation activities) and (2) evacuating from the train. They were simulated using the ‘‘exit events’’ feature of the STEPS model, which allows changing the availability of certain exits during the course of the simulation. Using this feature, the user can open, close or make exits unavailable. When an exit is set to closed, the agents will still consider the exit when choosing

39

their target and form a queue in front of it. When the exit is unavailable, it is considered to be no longer usable, and nobody moves towards that exit. In most cases, the conditions dictate the specific actions and the choice of the appropriate egress strategy. Two adaptive procedures are considered. The first procedure is called partial relocation (PR). It consists of evacuating the fireexposed coach and the two immediately adjacent coaches. The second procedure is called full relocation (FR) and consists of evacuating as many coaches as possible away from the fire. Due to the train immediately stopping in scenarios 1.3, 1.4, 2.3 and 2.4, the passengers have to evacuate to the track level and are forced to use the emergency ladder. This scenario is the evacuation procedure usually applied by the rail operator (RENFE Operadora). In these cases, the ramp is not used as an additional evacuation element. Note that the ramp is mainly used to transfer passengers from one train to another. During an evacuation to the track level when a fire occurs in an intermediate coach, however, two evacuation elements are required to ensure that all passengers can safely leave the vehicle. In reality, the devices (emergency ladder and ramp) can only be installed in the following fixed locations: coaches 05 and 07 in Train A and coaches 03 and 05 in Train B. Other scenarios require considering other locations for these evacuation elements. The hypothetical location of evacuation elements is considered in Scenarios 3.3 and 3.4 (Figs. 13 and 14). In these scenarios, the evacuation elements are located far from the fire, which facilitates an adequate distribution of the number of passengers per exit. An evacuation strategy that gives priority to the passengers that are closer to the fire is simulated in Scenarios 2.4 and 3.4 (controlled) [17].

3.4. Input data in the model The pre-evacuation activity is defined as the time from the detection of the fire to the time the train stops. The fire is

Scenario 1.1 (uncontrolled) and 1.2 (controlled) 12

11

10

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

Scenario 2.1 (uncontrolled) 12

11

10

Scenario 2.2 PR (controlled) 12

11

10

Scenario 2.2 FR (controlled) 12

11

10

Scenario 3.1 (uncontrolled) 12

11

10

Scenario 3.2 PR (controlled) 12

11

10

Scenario 3.2 FR (controlled) 12

11

10

Fire exposed coach Adjacent coach 1 Adjacent coaches 2, 3...n

Controlled = Evacuation supervised by crew members Uncontrolled = Self-rescue (escape)

Fig. 11. Evacuation scenarios to a platform in Train A.

40

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

Scenario 1.1 (uncontrolled) and 1.2 (controlled) 11

10

09

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

Scenario 2.1 (uncontrolled) 11

10

09

Scenario 2.2 PR (controlled) 11

10

09

Scenario 2.2 FR (controlled) 11

10

09

Scenario 3.1 (uncontrolled) 11

10

09

Scenario 3.2 PR (controlled) 11

10

09

Scenario 3.2 FR (controlled) 11

10

09

Fire exposed coach

Controlled = Evacuation supervised by crew members

Adjacent coach 1

Uncontrolled = Self-rescue (escape)

Adjacent coaches 2, 3...n Fig. 12. Evacuation scenarios to a platform in Train B.

Scenario 1.3 (uncontrolled) and 1.4 (controlled) 12

11

10

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

09

08

07

06

05

04

03

02

01

Scenario 2.3 (uncontrolled) 12

11

10

Scenario 2.4 PR (controlled) 12

11

10

Scenario 2.4 FR (controlled) 12

11

10

Scenario 3.3 (uncontrolled) 12

11

10

Scenario 3.4 PR (controlled) 12

11

10

Scenario 3.4 FR (controlled) 12

11

10

Fire exposed coach Adjacent coach 1 Adjacent coaches 2, 3...n Priority flow Secondary flow

Controlled = Evacuation supervised by crew members Uncontrolled = Self-rescue (escape) Evacuation elements

Fig. 13. Evacuation scenarios to the track level in Train A.

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

41

Scenario 1.3 (uncontrolled) and 1.4 (controlled) 11

10

09

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

08

07

06

05

04

03

02

01

Scenario 2.3 (uncontrolled) 11

10

09

Scenario 2.4 PR (controlled) 11

10

09

Scenario 2.4 FR (controlled) 11

10

09

Scenario 3.3 (uncontrolled) 11

10

09

Scenario 3.4 PR (controlled) 11

10

09

Scenario 3.4 FR (controlled) 11

10

09

Fire exposed coach Adjacent coach 1 Adjacent coaches 2, 3...n Priority flow Secondary flow

Controlled = Evacuation supervised by crew members Uncontrolled = Self-rescue (escape) Evacuation elements

Fig. 14. Evacuation scenarios to the track level in Train B.

considered to be manually detected in the passenger coach where it starts. The detection time is set to 60 s. Therefore, the passenger response time is assumed to follow a log-normal distribution with a minimum value of 60 s, a mean of 75 s and a standard deviation of 15 s. The passenger response times in adjacent coaches are summarised in the following equation: t rf ire o t radj 1 ot radj 2    o t radj i    o t radj n where trfire is the average response time of fire exposed coach; tradj the average response time of i-th adjacent (at both sides of exposed) coaches. The uncertainties about passenger response times under these conditions lead us to consider the following simulation hypotheses.

 Fast response times. The passengers from adjacent coaches

 

start evacuating before the previous coach has been completely evacuated. High levels of interaction between the passengers and leadership behaviour are assumed. Medium response time. The same conditions as above apply, but the response times of the passengers are slower. The train crew is assumed to have a high level of assertiveness. Slow response time. The passengers from the adjacent coach remain in their seats waiting for the passengers from the previous coach to cross the coach or to receive instructions from the train crew.

Once the train has stopped, different response times for the remaining passengers (the non-relocated passengers) are considered. Case1 assumes emergency notification after the train has stopped for uncontrolled evacuation, it has m ¼53 s and s ¼47 s based on data from [18]. Case 2 assumes emergency notification

before the train has stopped (controlled evacuation). In this case, the response time of the data described above (m ¼12 s and s ¼8 s) applies. 3.5. Assumptions The following assumptions were used in the simulations. 1. As explained above, the applied technique consists of determining and imposing the simulated outcomes of crew procedures. 2. The passengers involved in the pre-evacuation procedures are ready to start evacuation once the train stops. 3. The passengers are compliant with the train crew’s commands. This assumption is a basic requirement for seeing the effects of the evacuation procedures that are implemented. 4. The impeding effect of the internal sliding doors is ignored. 5. The train crew does not use the PA system during the preevacuation stage. Only the passengers directly involved in the fire are warned, while the passengers distant from the fire remain in their seats. 6. For the evacuation to the track level, no alternative escape routes, such as other exit doors where the passengers have to climb higher than 1.1 m, are considered. 7. In Train A, the ramp is stored in the lounge coach. In Scenarios 3.3 and 3.4 (fire in the lounge coach and evacuation to track level), it is assumed that the ramp is available for the evacuation. 8. For the evacuation to the track level, the time spent by the train crew setting up the evacuation elements (the portable ladder and ramp) is ignored.

42

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

4. Results Each evacuation scenario described in Figs. 11–14 was run 50 times. Examples of the dynamic simulations can be viewed at http://www.gidai.unican.es/. 4.1. Relocation procedures The results in Table 5 show that the predicted average evacuation times of the fire exposed coach under maximum loading conditions are longer than 90 s in both trains, as established by Table 5 Evacuation time distributions of the fire exposed coach. Scenario/strategy

2.1 2.2/PR 2.2/FR fast response 2.2/FR med. response 2.2/FR slow response

Train A

Train B

Mean (s)

S.D. (s)

Mean (s)

S.D. (s)

119 119 117 119 118

10 10 11 10 8

112 117 114 116 115

9 11 10 13 8

the ATOC standard. These times are seen because a detection time with a minimum value of 60 s is assumed. In fact, the first passengers start to move at 60 s. Therefore, the time required by the passengers to leave the coach is approximately 58 s in Train A and 54 s in Train B. If a safety factor of 2, which is suggested in [19] as a design margin, is applied to these times, the total evacuation time obtained is close to the values displayed in Table 5. The results suggest that the detection and passenger response times need to be added to increase confidence in the egress calculations. The results summarised in Tables 6 and 7 suggest that in the event of a fire in rear most coach, the predicted average evacuation times from the fire-exposed and adjacent coaches can be lower than 180 s for both trains. In this case, a self-rescue strategy is assumed. Furthermore, by applying the PR procedure, the train crew can relocate 96 passengers in Train A and 86 passengers in Train B in less than 210 s. In this case, only the fast response time is considered in the simulations. In case of a fire in the rearmost coach of Train A, it is possible to evacuate 6 passenger coaches with 195 passengers between 4 min 55 s and 7 min 23 s when the FR procedure is applied. By applying the FR procedure to Train B in Scenario 2.2, the train crew can relocate a maximum of 164 passengers from 5 coaches between 4 min 28 s and 6 min 22 s. In these cases, there is available space to relocate passengers along the train.

Table 6 Predicted time to evacuate the passenger coaches inside Train A (s).

Evacuated coaches. PR ¼Partial Relocation, x¼ Cannot move forward (no more room in passenger coaches) and FR ¼ Full Relocation.

Table 7 Predicted time to evacuate the passenger coaches inside Train B (s).

Evacuated coaches. PR ¼Partial Relocation, x ¼ Cannot move forward (no more room in passenger coaches) and FR ¼Full Relocation.

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

In the scenarios where a fire in the lounge coach is considered (Tables 6 and 7), the adjacent passenger coaches were found not to be completely relocated, and the passengers were caught in congestion until the doors were opened for evacuation. It is noteworthy that due to the specific configuration of Train B, the passengers relocated towards the front end have only one coach between them and the fire-exposed coach (see Table 7). In this situation, the passengers cannot be considered to have relocated to a safe place in the moving train, and the evacuation has to proceed as quickly as possible. 4.2. Evacuation to a platform

195 180 165 150 135 120 105 90

Scenario 1.2 satisfies the 90 s criteria for Trains A and B, and the 95th percentile from the simulations results is below the broken line. In this scenario, the passengers are ready to start evacuation once the train stops, and all the exits are available. In Train B (see Fig. 15b), the FR in Scenario 2.2 produced the longest evacuation time. In this scenario, the average predicted evacuation time increased by 18% compared to the PR in Scenario 2.2. It should be noted that when a lounge coach fire in Train B is considered, using the PR and PF procedures have no impact on the predicted egress times. In these scenarios, the relation between the passengers and the available exits is similar to that in Scenario 3.1. Passenger management during the pre-evacuation stage had a positive impact on evacuation efficiency when the PR procedure was used. It can be seen from Fig. 16a–c that using the PR procedure for Train A in Scenario 2.2 allowed an adequate distribution of passengers to each exit, thus reducing the evacuation times compared to Scenario 2.1, in which an unsupervised escape from the train is considered. In Scenario 2.2 (Train A), the average evacuation time for the FR procedure is 27% greater than that for the PR procedure. This increase occurred because 316

Evacuation time [s]

Evacuation Time [s]

In the fire scenarios, some exits are unavailable when the train stops, which increases the evacuation times. Fig. 15a and b show the minimums, means and 95th percentiles of the total evacuation times to the platform for Train A and Train B. The horizontal broken line in Fig. 15a and b represents the evacuation time established by the ATOC (90 s) in addition to the assumed delay time in opening the doors once the train has stopped (35 s). As expected, only

43

195 180 165 150 135 120 105 90

1.1 1.2 2.1 2.2 2.2 3.1 3.2 3.2 PR FR PR FR Scenarios

1.1 1.2 2.1 2.2 2.2 3.1 3.2 3.2 PR FR PR FR Scenarios Min. Mean 95th Perc. ATOC(90s+35s to open the doors)

Min. Mean 95th Perc. ATOC(90s+35s to open the doors)

Fig. 15. (a) Evacuation times of Train A. (b) Evacuation times of Train B.

80

N° of passengers

N° of passengers

Scenario 2.1 Scenario 2.2 Partial Relocation Scenario 2.2 Full Relocation

60 40 20 0 12 11 10 9 8 7 6 5 4 3 2 1 Coaches/exit

80

Scenario 3.1 Scenario 3.2 Partial Relocation Scenario 3.2 Full Relocation

60 40 20 0 12 11 10 9 8 7 6 5 4 3 2 1 Coaches/exit

1

Probability

0.8 Sce1.1 Sce1.2 Sce2.1 Sce2.2PR Sce2.2FR Sce3.1 Sce3.2PR Sce3.2FR

0.6 0.4 0.2 0 90

110 130 150 170 190 Time from the stop of the train [s]

Fig. 16. (a) Exit usage in Train A. (b) Exit usage in Train A. (c) Cumulative distributions of evacuation times to a platform in Train A.

44

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

passengers were compressed into 5 coaches, and the evacuation was performed using 5 exits. For the PR procedure in Scenario 2.2, a total of 8 exits were available, and 316 passengers were distributed between 9 coaches. Using the FR procedure in Scenarios 2.2 and 3.2 (fire in the rearmost coach) produces longer evacuation times. However, the passengers leave the train through exits further from the fire. 4.3. Evacuation to the track level In this section, we examined the impact of crew procedures on evacuations to the track level. The predicted evacuation times in the present analysis do not consider the time required to stop the train and the time expended by train crew in setting up the evacuation elements. These processes could produce a significant increase in the RSET. The discharge capacity of the evacuation elements (emergency ladder and ramp) determines the evacuation times. For this reason, the predicted evacuation times do not differ much when comparing controlled and uncontrolled evacuation scenarios, as shown in Fig. 17. A larger difference is found in Scenario 1.4, in which an emergency notification before the train stops has a positive impact on total evacuation times compared to Scenario 1.3. As mentioned above, the evacuation elements hold only two passengers simultaneously. The average flow rate obtained from simulations was 27 per/min. This value was significantly lower than the 30 per/min suggested by the ATOC standard. We found that a controlled evacuation that gives priority to the passengers closer to the fire has a positive impact. While this procedure has no effect on the total evacuation times, there is an improvement in the evacuation conditions of passengers inside the train. Passengers quickly move away from the incident area when the priority procedure is applied. Fig. 18 shows the layout of the passenger coaches analysed. Fig. 19a shows how the time to clear coaches 09 and 08 toward the exit coach (coach 07) had been considerably reduced in Scenario 2.4 when compared with Scenario 2.3. In Scenario 2.3, the mean time to clear coach 09 was 480 s, with a standard deviation of 31 s. In Scenario 2.4, this time had been reduced by 63% (the mean time to clear coach 09 was 176 s, with a standard

800

Time [s]

700 600 500 400 300 1.3

1.4

2.3 2.4 Scenarios

3.3

3.4

Mean 95th Perc. ATOC Flow time (30 pass/min) Fig. 17. Mean and 95th percentile of total evacuation times to the track level in Train B.

Coach 12

Coach 11

Coach 10

deviation of 18 s). In Scenario 2.3, the average time required for the passengers to evacuate coach 08 was 598 s, with a standard deviation of 17 s. By applying the priority procedure in Scenario 2.4, the average time had been reduced by 47% (a mean of 317 s and a standard deviation of 45 s). Furthermore, the density levels at the exit coach (coach 07) in Scenario 2.4 were lower than those in Scenario 2.3. As the results of Scenario 2.4 show (see Fig. 19b), a density over 3.5 per/m2 was observed in Scenarios 2.3 and 2.4. In Scenario 2.4, however, the density was greatly reduced after 100 s, with values lower than 2.5 per/m2, and the passengers from the incident area can leave the train under optimal conditions. Based on the simulation results, it can be argued that applying this procedure is more appropriate in terms of safety during the evacuation process.

5. Discussion Several passenger performance and evacuation procedure issues for high-speed trains were identified when performing this analysis. Although in real emergencies passengers may be more motivated to escape than they are in an evacuation drill and although there were limitations inherent in the drill scenario analysed (a lack of realism and cooperative behaviour), it is necessary to employ conservative settings to prevent an unrealistic evacuation performance [20]. The main parameters obtained from the evacuation drill observations were used as inputs for the evacuation modelling anlysis. The data collection results suggest that the default values in current egress models (most of which are designed for buildings) may not be adequate for trains. Furthermore, the ‘‘magic numbers’’ established by the standards may not be realistic for a passenger train evacuation analysis. In fact, the ATOC standard establishes the minimum required values for design validation. Therefore, it is necessary to perform a more detailed evacuation analysis by considering different emergency scenarios and procedures. The 90 s recommended by the ATOC has been used as a point of comparison for different evacuation scenarios. The simulation results suggest that adding the detection and passenger response times to the minimum necessary time to exit the coach where the fire originates will increase confidence in the egress calculations. In these cases, we suggest that it is good practice to double the minimum recommended egress times in the ATOC standard. During the relocation procedures inside the train, the passengers are likely to stop in front of an exit and wait until the train stops. This behaviour has been observed in evacuation drills [18]. The perception of threat also plays a key role in causing passengers to stop. The passengers that are further ahead cannot be aware of the conditions further back and closer to the fire. In fact, information slowly propagates among the passengers in a long queue [21]. Management of the passenger flow by the crew is required during the relocation procedure to prevent passengers from stopping. At least one crew member should be at the front of the queue leading passengers and another should be at the end of the queue directing passengers away from the incident area. However, we found that in a real situation, disabled passengers (wheelchair users) are unable to cross aisles that are 0.52 m and 0.69 m wide by themselves and require assistance.

Coach 09

Coach 08

Track level Fig. 18. Layout of passenger coaches analysed.

Coach 07

70 60 50 40 30 20 10 0

45

4 3 per/m2

N° of passengers

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

2 1 0 0 100 200 300 400 500 600 700 Time from the stop of the train [s]

0 100 200 300 400 500 600 700 Time from the stop of the train [s] Sce 2.3 coach 08 Sce 2.4 coach 08

Sce 2.3 coach 09 Sce 2.4 coach 09

Scenario 2.3

Scenario 2.4

Fig. 19. (a) Time to clear coaches 08 and 09 in Train A. (b) Density at exit coach 07 during evacuation process: Train A.

This consideration, which was not simulated, can significantly delay the relocation procedure. Some exits may be unavailable in cases of evacuation to a platform. Therefore, the required safe egress times are higher than the 90 s established by the ATOC standard. Applying the PR procedure during the pre-evacuation stage had a positive impact for fires in the rearmost coach and allowed for an adequate distribution of the number of passengers at each exit. When it takes longer than 10 min for the train to stop, however, applying the FR procedure is recommended to ensure that those passengers in immediate danger will be relocated to safe coaches. In this case, the average evacuation time increased by approximately 30% over that of the PR procedure. In Train B, however, applying the PR procedure to fires in the lounge coach did not have the expected effect on evacuation efficiency due to the specific configuration. The use of emergency ladders was also considered in detail for evacuations to the track level. After the train immediately stops, the passengers need to evacuate down the railway embankment. For the passengers close to the fire, however, evacuation in this manner represents a significant risk due to the limited number of available exits. A controlled evacuation that gives priority to the passengers that are closer to the fire minimises the exposure time and reduces congestion. It therefore allows quick and safe movement of the passengers inside the train.

6. Conclusions In this paper, different evacuation procedures for two highspeed trains have been analysed through computer modelling and simulation. The simulations used the impact of different preevacuation strategies on the evacuation process as a relevant and informative benchmark of egress safety. The following conclusions can be summarised.

 Single trials produce little information on the variety of



potential outcomes during the evacuation process. Evacuation modelling based on reliable data can be used to predict the impact and benefits of different crew procedures in case of fire. This approach can be incorporated by rail operators when defining emergency plans and evacuation strategies. When there is a fire onboard, the number of passengers to be relocated inside the train is a key factor in selecting an appropriate egress strategy. Relocation procedures inside the train can have a great impact on egress calculations. In fact, the evacuation time to the platform is higher than 90 s in most of evacuation scenarios due to the limited number of exits. Under



these conditions, an adequate distribution of passengers to each exit is recommended. When the passengers have to evacuate to the track level due to the train immediately stopping, the evacuation elements (portable ladders and ramps) should be located far from the fire and should facilitate an adequate distribution of passengers to the exits. Controlled evacuations that give priority to the passengers closer to the fire were found to be a good practice.

Acknowledgements The authors would like to thank the Ministry of Science and Innovation of Spanish government for the Grant for the Projects ‘‘Automated Decision Support System for High Speed Passenger Trains in case of Emergency’’ Ref: P65/08 and the ‘‘EvacTrains Project: An Evacuation Model for High Speed Passenger Trains’’ Ref: TRA2011-26738. References [1] Independent Transport Safety & Reliability Regulator, Train Door Emergency Egress and Access and Emergency Evacuation Procedures, Independent Transport Safety & Reliability Regulator, Transport Safety Regulation Division, Document reference 02468, November 2004. [2] ATOC Vehicles Standard AV/ST9002, Vehicle Interiors Design for Evacuation and Fire Safety, Issue one, published on behalf of Association of Train Operating Companies (ATOC) by Railway Safety, London, December 2002. [3] J.A. Capote, et al., Evacuation modelling of fire scenarios in passenger trains, in: Proceedings of the Fourth International Conference on Pedestrian and Evacuation Dynamics, 2008, pp. 705–712. [4] J.A. Capote, et al., An evacuation simulation method for a high speed passenger train, in: Proceedings of the Fourth International Symposium on Human Behaviour in Fire, 2009, pp. 613–618. [5] E.R. Galea, S. Blake, S. Gwynne, P. Lawrence, Simulating the interaction of cabin crew with passengers during aircraft emergency evacuation conditions, in: Proceedings of the International Aircraft Fire & Cabin Safety Conference, /http://www.fire.tc.faa.gov/2004Conference/html/evac.htmS, 2004. [6] S. Gwynne, E.R. Galea, C. Lyster, I. Glen, Analysing the evacuation procedures employed on a Thames Passenger Boat using the maritimeEXODUS evacuation model, Fire Technol. 39 (2003) 225–246. [7] STEPS Simulation of Transient and Pedestrian movementS: User Manual, unpublished, available with egress model from Mott MacDonald, /http:// www.mottmac.comS. [8] P. Kagendal, D. Nilsson, Evacuation study, Fire Safety on Intercity and Interregional Multiple Unit Trains, Report 5117, Department of Fire Safety Engineering, Lund University, 2002, pp. 39–60. ¨ ¨ [9] F. Klugl, G. Klubertanz, G. Rindsfuser, Agent-based pedestrian simulation of train evacuation integrating environmental data, Lect. Notes Comput. Sci. 5803 (2009) 631–638. [10] N. Zarbotius, et al., Design of formative evacuation plans using agent-based simulation, Saf. Sci. 45 (2007) 920–940. [11] Passenger Rail Car Egress Time Prediction, Research Results, RR06-04, U.S Department of Transports, Federal Railroad Administration, March 2006. [12] M. Oswald, et al. Full scale evacuation experiments in a smoke filled rail carriage—a detailed study of passengers behaviour under reduced visibility,

46

[13] [14]

[15]

[16]

[17]

J.A. Capote et al. / Fire Safety Journal 49 (2012) 35–46

in: Proceedings of the Third International Conference on Pedestrian and Evacuation Dynamics, 2007, pp. 41–55. E.R. Galea, S. Gwynne, Estimating the flow rate capacity of an overturned rail carriage end exit in the presence of smoke, Fire Mater. 24 (2000) 291–302. H.E. Nelson, F.W. Mowrer, Emergency movement, in: P.J. DiNenno, et al. (Eds.), The SFPE Handbook of Fire Protection Engineering, third ed., Society of Fire Protection Engineers, Bethesda, MD, 2002, pp. 3-367–3-380. S. Gwynne, E. Kuligowski, J. Kratchman, J.A. Milke, Questioning the linear relationship between doorway width and achievable flow rate, Fire Saf. J. 44 (2009) 80–87. Guidance on the Provision of Equipment and Arrangements for Evacuation and Escape from Trains in an Emergency, HSE Health and Safety Executive, March 2002. J. Tubbs, Selecting appropriate evacuation strategies for supper tall buildings: current changes and needs, in: Proceedings of the Fourth International Symposium on Human Behaviour in Fire, 2009, pp. 41–50.

[18] J.A. Capote, et al., An evacuation model for high speed trains, in: Proceedings of the Fifth International Conference on Pedestrian and Evacuation Dynamics, 2011, pp. 421–431. [19] R.D. Peacock, P.A. Reneke, J.A. Averill, J.H. Klote, Fire Safety of Passenger Trains; Phase II: Application of Fire Hazard Analysis Techniques, NISTIR 6525, Fire Research Division Building and Fire Research Laboratory, December 2002. [20] S. Gwynne, E. Kuligowski, The fault with default, in: Proceedings of 12th International Fire Science & Engineering Conference, Interflam 2010, 2010, pp. 1473–1478. [21] P.J. Harding, Prediction and mitigation of crush conditions in emergency evacuations, in: Proceedings of the Fourth International Conference on Pedestrian and Evacuation Dynamics, 2008, pp. 233–246.