A DYNAMIC NETWORK MODEL FOR CIRCULATION AND EVACUATION

摘要

The transition from prescriptive codes to performance based codes for fire safety management will need efficient tools that fulfil the dual requirements of quick and reliable simulation of people movement. Specifically, along with the commonly reported escape times, one needs an insight into the sensitivity of various factors impacting on the escape times and the associated bottlenecks. This paper introduces a way to fill this gap through our development of a new agent-based dynamic network model. The model is used for simulating the circulation and evacuation of people on a fast 3D network comprising "1.5D" links. Unique features of the model include an assignment of multiple, hierarchical choices of destination to individuals, stochastic distributions of individual/arrival/exit profiles, and competition for resources based on geographic location, attractiveness of destinations, queue lengths, and the final goal. The model monitors run-time queue lengths, performs congestion checks on links and carries out dynamic re-routing if required. The model has several advantages over the existing simulation tools. The 3D network environment gives the software an ability to perform thousands of quick runs to highlight the problems, while the built-in behaviour model provides a realistic determination of the likelihood of congestions and associated problems during movements. The model has been used for analysis of movements during critical scenarios in the design of a seven-storey racecourse building, a state-of-the-art school building, and is being further advanced as an alternative for building evacuation modelling. Details of the modelling work and results from the analysis are presented in the paper. It is argued that the model provides an efficient alternative for evacuation analysis using simple simulation parameters, making it suitable for the fast optimisation of vertical circulation provisions and sensitivity checks on such as the deference behaviour at stairs and junctions.