A cost‐effective building fire smoke spread prediction approach for risk mitigation based on data assimilation using Ensemble Kalman Filter

Summary Prediction of building fire smoke spread behaviors plays an important role in the evacuation in an emergency to mitigate fire risk. Based on data assimilation, a cost‐effective approach was proposed to predict fire smoke spread behaviors under unknown mixed disturbances by combining a zone model (CFAST) and Ensemble Kalman Filter (EnKF). CFAST is used to predict fire smoke spread behaviors as a deterministic fire model. Sensor data of smoke temperature were assimilated into CFAST simulation by leveraging EnKF to estimate unknown heat release rate (HRR) and thus improving the prediction accuracy. The performance of the proposed approach was tested via a series of Observing Systems Simulation Experiments. Two series of cases were conducted for unknown single HRR change disturbance and unknown single‐window breakage disturbance, one series of cases was conducted for unknown mixed HRR change and window breakage disturbances. Result comparisons have been presented in figures to assess the approach performance qualitatively, and root mean square errors (RMSEs) have been calculated to assess the approach performance quantitatively. The RMSEs are within the range of 1.78–29.88 K. The results showed that it was a viable solution to set a larger perturbation range for unknown disturbances. The proposed approach could provide more accurate and reliable predictive information about fire smoke spread behaviors for risk mitigation, as well as other safety‐related applications.