A Distributed Method for Predicting Building Fires Based on a Two-Layer Zone Model

Abstract

With the emergence of large scale buildings, how to improve the efficiency of the fire prediction model has urgently been a problem in the field of fire emergency. Based on an existing fire zone model, CFAST, and some other assumptions, this paper puts forward a distributed fire model capable to make a fast and acceptable prediction. The whole building structure is divided into some small basic zones in which a group of control equations run independently relying on the local information. With a small scale of the local structure, the computational complexity will not increase significantly as the building scale enlarges. In this paper, a typical building structure with multiple rooms and a long corridor is discussed. Two kinds of models, namely fire zone model and corridor zone model are set up and run locally. At a prescribed time, based on the limited communication between those two kinds of models, the properties of the corridor can be updated. By repeating these steps, the global state can be predicted. One typical floor of a real building is used to test this distributed fire model with a 3 MW steady fire and the comparison against the conventional CFAST model is carried out. The results show that the proposed distributed fire model can perform well in a short term prediction (about 150s after fire breaks out), but for the long term prediction, the simulations are affected by the “far away” wall and HVAC condition, which resulting in a diverging solution from the conventional CFAST model.