Emergency Door Capacity: Influence of Door Width, Population Composition and Stress Level

The capacity of an emergency door depends on several aspects, among which the composition of the population using the door, the conditions under which the door is used and the door width. Before describing these experimental variables in more detail, some boundary conditions are set.

In the experiments, an opening represents the emergency door: subjects pass a free passage of a certain width. In this opening, no doorstep is present, to reduce hindrance and prevent possibly dangerous situations for participants. In addition, the pedestrian flow is one-directional, implying that no counter flows are present caused by fire fighters and people from emergency services. In reality, these people will rarely enter a building when the evacuation process is still going on.

The experiments performed by Peschl [1] have been based on a student population. However, in practice, the population will not consist of persons being in good shape, but the persons will have different physical conditions. In this research, we will use age as an indication for a person’s physical condition. Here, we distinguish four categories: children (under 18 years of age), adults (between 18 and 65 years of age), elderly (over 65 years of age) and disabled persons. With these age categories, we are able to compose populations corresponding to a variety of situations, see Table 1. The disabled people are represented by three persons in wheelchairs and three blindfolded persons.

The conditions under which an emergency door is used may vary considerably. In the experiments, both the stress level of the participants and the sight are varied. Not much is known on how to introduce stress in an experiment. In the past two methods have been considered favorable: enforcing participants to hurry e.g., by rewarding participants according to their performance and exposing participants to noise. Here, we have chosen to use for the latter option by sounding the slow-whoop signal. In addition, the stress level of the participants is raised by a combination of the slow-whoop signal and stroboscope light. In total, participants have been exposed to three stress levels: none, a slow-whoop signal and a combination of a slow-whoop signal and stroboscope light.

The sight is reduced by reducing illumination to a low level. Two alternative light situations are considered: full lighting (200 lux) and dimmed (1 lux, corresponding to emergency lighting).

In the experiments, the opening width is varied between 50 cm (the minimal free passageway of an escape route in the Building decree for existing buildings) and 275 cm. In addition to an opening of 85 cm wide (minimal free passageway of an escape route in the Building decree for new estates) openings are a multiple of 55 cm. Furthermore, an opening of 100 cm is tested to see the correspondence with the normative capacity expressed as the number of persons passing an opening of 1 m wide in 1 min.

The final experimental variable relates to whether or not the outflow of pedestrians after passing the doorway is free. In reality, doors cannot always open 180°, but may be restricted. To restrict the outflow of pedestrians after passing the doorway, we used a door being fixed at an opening of 90°. The total doorway width is not affected.

Ideally, all combinations of experimental variables should be investigated. Since this is not feasible due to time restrictions (the experiments should not last longer than a single day), for each experiment one variable is changed, while for the other variables the default value is maintained. By interpolation of the results of the various experiments, pronouncements can be made on the not performed experiments. The stress levels are varied for all experiments.

Each experiment will be performed multiple times to guarantee the reliability of the observations. To determine the number of repetitions, a total time of congestion of 3 min should be achieved. Since the time of congestion for wide doors is shorter than for narrow doors, more repetitions are performed for the wide doors.

An overview of the experiments is shown in Table 2.

A digital video camera and an infrared camera are used to observe the experiments. The infrared camera observes LEDs, attached on top of the caps of the participants. This technique guarantees good observations for the dimmed conditions. For the other experiments a digital camera is used, which is attached to the ceiling next to the infrared camera.

In total 75 children of 11 years old (blue caps), 90 adults (red caps) and 50 elderly persons (yellow caps) have participated in the experiments. This leads to populations of between 90 and 150 persons, which are large enough to cause congestion upstream of the door to observe capacities.

To represent an emergency door, a wall has been built in the middle of a large hallway, perpendicular to the sidewall. In this wall, an opening is made, whose width is easy to vary. At the side of the wall, some space is left to walk from one side of the wall to the other without using the opening. Above the centre of the opening an emergency exit sign has been hung up. An overview of the experimental site is shown in Fig. 1. To use the doorway more efficiently the participants use it in two directions: in the first experiment, they walk from one side of the wall to the other and in the next experiment they walk back again.

The images of the digital video camera form the basis to calculate the capacity of an emergency door. The movie of each repetition of an experiment is split into separate images with a frequency of 25 images per minute. Figure 2a shows such an image of the reference experiment with a doorway of 85 cm wide, an average population, 200 lux, no open door and no stress.

To calculate the capacity of the emergency door, the moments that the subjects pass a cross-section directly downstream of the door need to be known. Figure 2b shows the considered cross-sections: the dotted line indicates the cross-section to determine the capacity of the flow from right to left, while the striped line corresponds to the cross-section to determine the capacity of the flow from left to right.

To determine the cumulative curves on the specified cross-sections, the picture lines of the cross-section are placed next to each other for the duration of a repetition of an experiment using Matlab ^®. This gives a similar effect as a finish photo, see Fig. 3a. Figure 3b shows a zoom of the picture lines in Fig. 3a. This figure clearly shows that the first participants have passed the cross-section at high speed: their image is very narrow, while the images of the following persons are wider. A similar approach has been used to identify vehicle trajectories from video images made from a helicopter [9].

The next step is to recognize the caps to automatically derive the passage moments of the individuals using the cap colors. In the ‘photo finish’ image areas are looked for within a pre-specified color range. By not only looking for the colors bright red, bright blue and bright yellow, but by allowing a color range, the shadow effects, the cap edges and the stroboscope effects are compensated for. The detected pixels are then combined to caps. For this, two built-in methods in Matlab^® have been used: a method to identify areas within a specific color range and clustering methods. More information on this cap recognition methodology can be found in [10]. Figure 3c shows the tracked participants for a repetition of the reference experiment.

For each cluster (participant) its heart is calculated. The x-pixel of this heart indicates the time moment when the participant passes the cross-section, while the y-pixel indicates the lateral position of this event. A cumulative curve can be derived based on these passage moments, where the number of persons having passed the cross-section is plot against the passing moment, see Fig. 3d.

Assuming that the capacity of the door does not change during a repetition of the experiment, a straight line is fit through the cumulative curve. The derivative of this line corresponds to the average capacity of this door during this repetition. The average capacity of the experiment is then the average of the capacities of all repetitions. If the average capacity of each experiment is known, the relations between the capacity and the various experimental variables (door width, population, stress level, etc.) can be determined.

Based on the methodology described in the previous section capacities have been calculated for all repetitions of all experiments. In this section, the influence of the experimental variables, doorway width, population composition and the influence of an open door are discussed.

In the previous section, the influence of each independent variable on the capacity has been discussed separately. In this section we will show the influence of all variables together, performing a multivariate regression. Here, we take into account stress-level S (0 = no stress; 1 = slow whoop signal; 2 = slow whoop signal combined with stroboscope), door width W in meter, population (percentage of children P _C , percentage of elderly P _E and percentage of disabled persons P _D ), the presence of an open door D (0/1), light intensity L (0.05/1) and time T in hours from the starting moment. For the population, we have left out the percentage of adults P _A , since this is redundant with the other percentages (P _A  = 1 − P _C  − P _E  − P _D ). The capacity C can then be calculated as follows:Equation 1 shows that the reference capacity is 2.6685 P/m/s. The stress level hardly has an effect on the capacity. The door width also has a negative effect, implying that wider doors are less efficiently used. Obviously, the door width for which Eq. 1 can be applied is limited, otherwise the capacity becomes negative. Since the widest door handled in the experiments had a width of 2.75 m, we suggest to take 3 m as boundary width. With respect to the population, we can see that the children have the largest, positive, effect on capacity: the more children are part of the population, the higher the capacity. A higher percentage of elderly also has a slightly positive effect, whereas more disabled persons in the population have a strong negative effect on the capacity. The presence of an open door also has a negative effect on the capacity, whereas the lighting conditions have a positive effect (more light indicates a higher capacity). Finally, as we saw in earlier analyses, the time moment that the experiment took place has a slightly negative effect on capacity.

The next question is whether these parameters are significant. For this, we applied the t-test for each parameter. Table 3 shows that only the parameter for the stress level is not significant.

Finally, we checked the correlation between the parameters. Table 4 shows that the variables are not correlated.

The main conclusion that can be drawn is that the capacity of thirteen out of sixteen experiments is higher than 2.25 P/m/s. The experiments with a lower capacity have a population with disabled persons, a very wide doorway (275 cm) with less children in the population and an open door.

Another conclusion is that in the performed experiments more pushing does not lead to the ‘faster-is-slower’ effect. In the experiments a higher urgency (higher stress level) leads to higher speeds and to a higher capacity.

During the experiments, the conditions were optimal. This implies among other things that the inflow of pedestrians towards the door is unhindered, which is also the case for the outflow (except for the experiment with the presence of an open door). Also, factors such as people wearing winter clothing or carrying bags have not been taken into account. In order to determine the design value of the capacity of an evacuation door, users of the building code and designers of buildings should realize themselves in which conditions their doors will be used, and adapt the capacity value accordingly.

Many differences between the observed capacities can be explained by the different experimental variables. The images of the experiments indicate that an explanation can also be found in the individual behavior of the participants. When this microscopic behavior can be predicted, also the capacities can be predicted for a larger variety of conditions. This will be subject of future research.

The research described here has explicitly been focused on the capacity of emergency doors. This is only part of the total evacuation process. The previous process (pre-evacuation, route choice, walking towards the exit) has a direct influence on the arrival pattern of pedestrians at the emergency door, and thus whether or not capacity of the door will be reached. This is also subject of future research.