Mescape - A probabilistic mesoscopic model for fire evacuation in large gatherings

Abstract

This study presents a mesoscopic model that predicts pedestrian evacuation in large gatherings with probabilistic methods. The model is based on a coarse network representation of buildings, where occupants are treated at a different scale according to their location.
A microscopic approach is used to represent people in the initial enclosure: here individual characteristics and behaviours are attributed to each occupant through random sampling from user-defined probabilistic parameter distributions, and individual timelines are used to determine the initial flow of evacuees. A macroscopic representation is implemented in the following parts of egress routes, where occupants are treated as a homogeneous crowd with uniform characteristics. The SFPE hydraulic model is used to predict movement until a place of safety.
Two automatic iterative processes are implemented within the evacuation model to address variability and uncertainty. In the first step, behavioural uncertainty is investigated by running the same evacuation scenario multiple times until results meet user-defined converge criteria based on inferential statistics. In the second step, input parameters/distributions are modified automatically to predict evacuation in different scenarios for quantitative risk assessment.
Multiple tests show good agreement between the results of the proposed mesoscopic model and a microscopic model employed for benchmarking (Pathfinder used in SFPE mode). For horizontal egress routes, the proposed tool provides detailed and accurate outputs (evacuation curves, flow rates, queue size, etc.) with short computational time. For vertical egress routes, mesoscopic predictions capture well the qualitative evolution of the egress process, but quantitative results are underestimated and necessitate additional investigation.
Lastly, a quantitative risk assessment is performed for a case study, in which the proposed model is used for the calculation of the evacuation curve in several scenarios. The efficiency of modelling and calculation processes highlights that the proposed model represents a valuable contribution for the probabilistic analysis of fire evacuation in large gatherings.

Popular Abstract

Fire safety engineers often use computer models to predict how people will evacuate from a building in case of emergency. This is essential to make sure that they do not get injured by fire or smoke. In order to do so, modellers need to make several assumptions about the conditions at the beginning of the emergency (e.g., number of people that are in the building) and about the behaviour of the persons during evacuation (e.g., how quickly they recognise a fire alarm, which door they choose, how fast they walk, etc.). However, many of these parameters are often uncertain.
Consequently, to analyse complex design situations characterised by large uncertainties, engineers can work in a probabilistic manner. This means that instead of running just one evacuation simulation with a single set of input parameters, they perform hundreds or thousands of simulations with different combinations of inputs. In this way, they can check if the occupants of a building are safe in many different alternative scenarios.
However, most of the software that have been developed to perform evacuation simulations are not suitable for the probabilistic approach. On one hand, running many simulations with advanced computer models may take too long. On the other hand, simpler and fast evacuation models may not be sufficiently accurate. Therefore, how can engineers get refined results in an efficient way?
This study presents an evacuation model that has been developed precisely for this purpose. The model is based on a hybrid approach that combines the best features of both advanced and simple computer tools. This is achieved using a different representation of the persons according to their location in the building (mesoscopic approach). In particular, evacuees are modelled as individual people at the beginning of evacuation (microscopic approach), and as a uniform crowd in later stages (macroscopic approach). This allows engineers to get precise results very quickly, hence to use the model for the analysis of many evacuation scenarios. A set of algorithms has also been implemented in order to make the modelling process fully automated and stop as soon as the results are conservative.
In the last part of the study, the proposed mesoscopic model has been tested in a number of simple and complex buildings, and the results have been compared with the ones produced by the most popular advanced microscopic model (Pathfinder). The tests show good agreement between the outputs of the two models when they are used to simulate the evacuation from single storeys. The mesoscopic tool requires further development to refine the prediction of evacuation through stairs.
Overall, the proposed mesoscopic approach provides detailed and accurate outputs in short computational time. Therefore, it can be used in combination with probabilistic methods to investigate a large number of emergency scenarios in complex facilities.

• author: Contini, Lorenzo
• supervisor: Enrico Ronchi
• organization: Division of Fire Safety Engineering
• course: VBRM05 20221
• year: 2022
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: Human behaviour, crowd evacuation, mesoscopic modelling, hydraulic model, large gatherings, fire safety, performance-based design, quantitative risk assessment, consequence analysis, uncertainty analysis, probabilistic modelling
• publication/series: LUTVDG/TVBB
• report number: 5669
• other publication id: LUTVDG/TVBB-5669-SE
• language: English
• id: 9084392
• date added to LUP: 2022-06-07 10:19:48
• date last changed: 2022-06-07 12:24:58