Promoteur : Prof. Sélim DATOUSSAID

Co-promoteur : Prof. Thierry DESCAMPS

Résumé :

Historical and recent building fire events have shown that ongoing efforts are still to be made to
enhance the fire safety of the built environment. Several evacuation models have been developed to
predict the level of safety of a design. These models became key tools used by practitioners to assess
the fire safety of complex buildings such as healthcare facilities. Their results rely on the input values
selected by users based on existing datasets. Several evacuation studies have been carried out in various
built environments to generate input values for evacuation models. However, those regarding healthcare
facilities evacuations are still rare. In this research work, we present a new evacuation dataset for
healthcare facilities. The data was collected from an outpatients’ area of a public hospital in Auckland
(New Zealand) during three unannounced fire drills which involved patients and staff. Pre-evacuation
times, evacuee horizontal travel speeds, exit selection and evacuation times were collected and analysed.
Moreover, evacuee behaviours were investigated to study staff and patients’ interaction during the
evacuation process. The results showed that pre-evacuation time of patients ranged from 5 to 63 s, while
pre-evacuation time of staff ranged from 8 to 141 s. In addition, during the evacuation movement phase,
staff who were not assisting patients, and patients with no mobility impairments, walked at a similar
average speed (i.e. 1.06 m/s for staff members and 0.91 m/s for patients with no mobility impairments).
The results indicated also that the average travel speed of patients with walking impairments and
caretaking staff was about half of the travel speed of the first two groups (i.e. 0.52 m/s).
The reliability of the results produced by evacuation models depends highly on the development
assumptions and features that are incorporated in these tools. For instance, many of existing evacuation
models do not incorporate assisted evacuation feature which is a key component in healthcare facility
evacuation simulation. Upon a preliminary fire risk analysis, this work attempts to fill this gap in the
existing evacuation models by developing an algorithmic formulation of an assisted evacuation submodel.
This formulation is then implemented in Unity game engine and a proof of concept, through a
case study, is presented.
In a healthcare facility context, the success of a fire safety procedure in a real-life emergency mainly
depends on staff decisions and actions. One of the parameters influencing staff decision-making is their
training. Typically, healthcare facilities’ safety educators use conventional training methods such as
slide-based lectures combined with live training on the operation steps of fire extinguishers, in order to
prepare the staff for fire emergency situations. Despite the advantages of these methods, various existing
studies have shown that they easily fail from a pedagogical point of view due to several limitations.
Virtual Reality (VR) is gaining fire safety community’s attention for being an interesting training tool.
There exist two types of VR-based applications: non-immersive and immersive applications. Little is
known about the influence of the degree of immersion of VR-based training applications on users’
perception of the content of the simulation. Moreover, the effectiveness of these tools is yet to be
investigated in a healthcare setting. The present research addresses the above-outlined limitations by
developing two versions (non-immersive and immersive) of a VR-based fire safety training simulator
and by assessing its usability and effectiveness compared with a conventional slideshow. A first study
is therefore carried out in order to assess the content of either version of the simulator, while a second
study is conducted to assess their effectiveness when compared with the slideshow. The three training
interventions are compared by assessing the increment and retention of fire safety recommendations as
well as the changes in users’ motivation and self-efficacy regarding fire safety training. 93 staff from
Charleroi hospital (Belgium) participated in this study. They were divided into three groups: Group A
was trained using the slideshow, Group B was trained using the non-immersive VR-based serious game
(SG) and Group C was trained using the immersive VR-based SG. On one hand, the results of study 1
indicated that the great majority of the participants had a positive perception of the VR-based SGs
content. On the other hand, the results of study 2 revealed that the VR-based training solutions were
more effective than slide-based training in terms of knowledge acquisition and retention of the acquired
knowledge. Whereas, using either intervention, the analysis revealed no statistically significant change
in participants’ motivation. Finally, the VR-based SGs showed a greater impact on self-efficacy in short
and long terms than the slideshow.