Human Behaviour in road tunnel fires: comparison between egress models (FDS+Evac, STEPS, Pathfinder)

Performance based design is becoming a common method to achieve aesthetic, cost and functional goals while maintaining safety levels for every type of environment, including tunnels. In this context, computational models can be considered as useful tools to ensure the safety conditions in road tunnels due to emergency scenarios. However, the actual state-of-the-art presents a multiplicity of models for emergency evacuations, each with their own strengths and weaknesses. Therefore, there is a need to investigate the differences in terms of egress times between different models to provide the safety conditions in this specific environment. In fact, the study of the evacuation process – for the case of road tunnel fires - requires the analysis of many factors and processes related to human behavior, such as pre-movement times (e.g. reluctance to leave the vehicle), interactions between different kinds of occupants, interactions between occupants and smoke, smoke influence on walking speed, herding behaviors, way-finding, etc. [1-5]. Computer modeling permits to simulate psychological processes that otherwise hand calculations do not allow. For this reason they provide a more realistic approach about the evacuation process. Each model has its own specific features and often practitioners do not have a thorough understanding of the variables that could be input in each model and how they will affect the results. This study begins with a short description and explanation regarding the models and their classification (i.e. flow based, cellular automata, agent-based, etc.) [6, 7]. Consequently, an analysis of the inputs of a selection of computational models well-known in the field of evacuation modeling (FDS+Evac, STEPS and PathFinder) is performed. In addition, the evacuation times are calculated using the SFPE capacity method. This approach calculates the total evacuation times from the means of the evacuation capacity. The attention is then focused on specific issues related to human behavior in road tunnel fires [8]. Different models can enable the simulation of human behavior and movements of each person individually or collectively. In addition, there are different approaches regarding the input step: the deterministic approach or the stochastic approach using distribution laws. Each model is then analyzed, highlighting which factors (fire scenario, distribution laws, etc.) can be implemented directly or “artificially” (collecting data from other models, e.g. using FDS results to model the fire and the smoke). In the second part of the paper, there is an analysis of several evacuation simulations for the Lantueno tunnel in Spain (a two-bore road tunnel with an emergency link tunnel in-between the two bores). Different road tunnel fire scenarios are considered. Response times, body diameters, unimpeded walking speed, number and position of the occupants, fire position and fire spread, etc. are some of the variables considered in these scenarios. Simulations are run in different conditions in order to obtain a consistent number of results. The different models are tested using the same input data in different scenarios (taking into account that it is necessary to make assumptions due to differences between the models). An analysis is made in order to evaluate the effect each individual factor has on the evacuation process and on the reliability of the results, for each model. The analyzed results are the evacuation times and the position of the arising bottlenecks (if the model is capable of performing this analysis) in different scenarios. The variability of the results between each model is also investigated, as well as an analysis of the conditions in which the evacuation results are similar. Consequently, a thorough description of the factors causing differences between the results is made. Finally, a novel methodology is defined. For the same scenario under the same specific conditions, each model can produce different evacuation times. It is likely that some models will provide a more reliable analysis of the evacuation process depending upon the possible variables that can be input into the model. Thus, a methodology using “comparison coefficients” will be defined by making a statistical analysis of the results. For a specific case, if only one model is used to calculate an egress time (as is most often the case in non-research applications), then a previously defined “comparison coefficient” for this specific condition can be applied in order to ensure safe and reliable final evacuation times.