Numerical models provide detailed information of situation in buildings during a fire. But the models can have unsatisfactory practical performance for emergency response, due to model defects and deviations of pre-set input parameters. Data assimilation methods for error mitigation have shown great performance improvement by combining real-time measurements with fire models. However, these methods cannot revise the simulation effectively when there are obvious systematic errors between fire model and sensor readings. Ensemble Kalman filter (EnKF) is one of classical data assimilation methods. In this paper, a decentralized EnKF-based fire prediction method is proposed for the dynamic situation awareness in building fires. Individual EnKF-based fire model is established for each substructure of a building. The proposed method is less affected by systematic error of the fire model. Based on the EnKF predictions, important information of smoke and fire hazard are extracted and a fire hazard mapping is created for first responders. The multi-compartment case study validates the effectiveness of the EnKF-based fire prediction and dynamic situation awareness for emergency response.

This content is only available via PDF.
You do not currently have access to this content.