Condicionamientos del cambio semántico

In case of passenger ships, particularly ferries, passengers may sit on seats which can be arranged in lines. As a result, in order to evacuate, each passenger may have to wait for the passenger before him (see for example the geometry proposed in the SAFEGUARD validation data set 1 ([186])). Although this might seem a simple test condition, the comparison with experimental data presented in this section shows that this particular situation highlights some limitation of simulation tools.

The detailed experimental data used to analyse this situation, were not related to the maritime field, but, to the civil field, and have been presented by Santos et al. [188]. The experimental data were gathered during an evacuation exercise conducted at a teaching and research institute of Lisbon University, where 64 first year graduate students participated. During the evacuation exercises videos were recorded in order to analyse the students’ behaviour. Moreover the students answered a questionnaire at the end of the activity. The answers to the questionnaire showed that the 46% of the students was motivated to do the activity and the 93% of them felt safer in the building after the evacuation exercise. Video records showed that some that students, however, did not maintain a correct behaviour during the exercise (they stopped for no apparent reason or they moved slowly or they waited for friends). The simulations performed through the developed tool of one of the evacuation situations described by

Santos et al.[188] provided a further clarification of the effect of these behaviours on

the exit time of the students.

The considered situation is the evacuation from a classroom where the desks were aligned and the evacuation drill started with the subjects sitting at the desks. The condition can be considered, however, similar to the one of passengers starting an evacuation sitting at aligned seats. Specifically the situation reproduced with the mathematical model corresponds to the evacuation exercises performed in Class 2 as it is indicated by Santos et al.[188]. The population of subjects in the experimental test consisted in 14 students mostly aged between 18 and 20 years. The intervals for the unimpeded speed of the modelled agents have been defined starting from those provided by MSC.1/Circ.1533 [147] for ‘Males younger than 30’ and ‘Females younger than 30’, and rescaled according to the average unimpeded speed of 1.06 m/s which was measured during the experiments. The mass of agents has been set according to the data acquired during the experiment, whereas the dimensions of agents have been obtained by rescaling in accordance to equations (2.10). In the simulations, 100 independent Monte Carlo simulations have been performed where the unimpeded speed varies according to the agents gender and initial position, while the mass and dimensions have been left unchanged.

The initial positions of the people and the geometry used for the simulation are reported in Figure 4.40.

Figure 4.40: Experiment on exit from a classroom by Santos et al.[188]. Initial positions and geometry.

Figure 4.41 reports snapshots of one simulation. From the simulation a continuous flow towards the room exit can be observed, which is slowed down only by the bottleneck condition occurring at the door. However, from the recorded videos, it is clearly visible that the flow of students for this condition is often interrupted. These interruptions are either due to the presence of chairs in the room or by the fact that some students occasionally stopped, thereby blocking the flow.

This behaviour is confirmed by the analysis of exit times in Figure 4.42, which reports the exit time of each agent and compares them with the experimental exit time of the corresponding evacuee. The exit times resulting from the simulations match the experimental ones for more than half of the evacuees. In the other cases, instead, the evacuees exit time in the experiment is much longer than the one obtained from simulation. An explanation for the observed discrepancies can be given by looking at the recorded video. In fact, the video shows that evacuee 2 stops before the exit in order to wait for evacuees 3 and 4, since evacuees 3 and 4 were left behind because they stopped to take their own bags. Those kinds of behaviours might affect the overall time performance of the evacuation process, and they are not part of the modelling used in the simulations.

The phenomenon where, due to geometrical constraints, one evacuee slows down all the others in the row could, however, be observed also in the simulations outcomes. To this purpose those simulation where the evacuees 1, 5 and 12 obtain the largest exit time were highlighted in black, red and green respectively in Figure 4.42. The figure shows that in each desks line the exit time of the evacuees is affected by the reduced speed of the evacuee occupying the initial position of the line, as it was expectable.

Figure 4.41: Experiment on exit from a classroom by Santos et al.[188]. Snapshots of one simulation at different time instants.

Figure 4.42: Experiment on exit from a classroom by Santos et al.[188]. Exit times resulting from 100 realizations. Simulation results are reported for each agent through box plots (min, 25%, 50% and 75% percentiles, max), scatter data. Experimental data are also reported. Some simulations are highlighted. In particular are highlighted the simulation outcomes associated with simulations

having the maximum exit time for the evacuee 1 (green square), the evacuee 5 (red cross) and the evacuee 12 (black circle). Evacuees 1, 5 and 12 are the ones positioned initially at the beginning of

their respective desk line, thereby their performance, in terms of exit time, affects also the performance of all the agents in the same line.

The effect, on the overall exiting performance, of the lack of complete modelling of evacuees behaviours can be better appreciated by looking at the exit time curve reported in Figure 4.43. Apart from an initial delay, it can be observed that the flow rate of the evacuees in the experiment matches well the one obtained from the simulations.

Figure 4.43: Experiment on exit from a classroom by Santos et al.[188]. Simulated exit time and comparison with experimental data. Outcomes from all realizations are reported together with

ensemble median and 5% and 95% percentiles. Experimental results are also reported for comparison purposes.

As anticipated, it can be noted that, in Figure 4.43, early in the evacuation process, the experimental curve is shifted by some seconds with respect to the median of simulations. However, the exit rate in the subsequent part of the evacuation is very similar to that obtained from simulations. The initial shift is due to the initial slowdown of the evacuation process, which is present in the experiment but not in any of the simulations. This shift corresponds, in the video, to the situation of an evacuee stopping in front the exit while waiting for two others evacuees, as discussed before.

Behaviors such as stopping for apparently no reason or going in the opposite direction with respect to the exit are generally not considered when the assessment of evacuation time is performed through the use of simulators. For instance, there is no requirement by IMO concerning the modelling of these behaviours (MSC.1/Circ.1533 [147]). This is mainly due to the fact that models simulating these behaviours are very difficult to calibrate and validate. Moreover, such unpredictable behaviours might change drastically if considering a different population or different conditions (for example a real condition of danger).

Due to the relevance to the maritime field of a test condition with evacuees starting aligned behind some obstacles, an additional test could be suggested in the guidelines in

the exit times reasonably reflect the initial position of the agents, as it is highlighted in Figure 4.42. Such a test could eventually assess the capability of the tool to correctly reproduce the human behaviour in narrow spaces.