Evacuation Dynamics Simulation Using Discrete Models With Uncertainty Factors

With the development of the economy and building technology, there appear a wide variety of super high and large buildings, which open the way for the form of dense crowd. In case of emergency such as fire and earthquake, how to prevent and reduce the casualties and ensure the safety evacuation has become the most important thing of public safety in the world. The study of typical characters and fundamental rules of pedestrian evacuation can provide the scientific guiding principle for the design of buildings and the emergency evacuation plan.The study on pedestrian dynamics mainly adopts computer simulation approach because of the lack of experimental data and its imprecision. However, Most of the existing models focus on the study of evacuation time and rules of pedestrian movement, instead of the complexity and uncertainty factors. Aiming at this purpose, after analyzing the uncertainty factors that influence the evacuation, this paper proposes two kinds of discrete egress models considering the uncertainty of pedestrians'number and the pre-movement time respectively.An evacuation model considering the characteristic of pedestrian number is built based on the common lattice gas model and the mean-field model. With the concept of statistic probability, the interval estimation of evacuation time and pedestrian numbers under certain confidence level can be obtained, even if the exact pedestrian number in the building is unknown. Compared to the common lattice gas model, the computational efficiency is developed in the condition that the exact pedestrian number in the building is unknown.In order to study the evacuation from a multi-barriers building, a multi-grid model considering the pre-movement time of each pedestrian is proposed. The applicability of the model is validated experimentally. The results indicate that the model can reproduce some typical phenomena in evacuation dynamics such as variable velocity, dislocable queuing, monopolizing exit, distribution of pre-movement time and so on. The pre-movement time exhibits a truncated log-normal distribution and has great influence on the total evacuation time. The distribution of pre-movement time and the coordination among pedestrians may induce the uncertainty of the evacuation results. The two models proposed in this thesis are expected to be the framework model in the study of evacuation uncertainty, and the simulation results may be useful to enhance the reliability of the evacuation design and evaluation.