Xi, Xiuqi; (2023) Data Driven Forecast for Fire. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Data_Driven_Forecast_for_Fire.pdf - Accepted Version Download (9MB) | Preview

Abstract

Being able to forecast the evolution of a fire is essential for fire safety design and fire response strategies. Despite advances in understanding fire dynamics and improvements in computational capability, the ability to predict the evolution of a fire remains limited due to large uncertainties associated to multiple scales and the non-linearities. The data-driven approach provides a viable technique from models corrected by observations. However, the complicated coupling between gaseous and condensed phases has, in the past, limited proper prediction with a positive leading time. This work proposes and investigates a series of approaches to data-driven hybrid modelling that integrate analytical and numerical descriptions to address the coupling effects. The data-driven hybrid model is developed for different scenarios covering various complexity and scales. Different approaches are evaluated to reflect the dominant physics; nevertheless, they are structured by differentiating the condensed and gas phases. The initial scenario corresponds to one-dimensional convective-diffusive droplet combustion in micro and normal gravity. Then, concurrent flame spread in micro and normal gravity where a two-dimensional boundary layer combustion approach is implemented. Finally, the Malveira fire test represents a large-scale, three-dimensional, travelling fire. Coefficients assimilated with their experimental observations are used to alter analytical formulations describing the gas and condensed phases. By separating the phases, the data-driven hybrid model can forecast various types of variables while reducing processing resources. Convergence of the assimilated coefficients is used as an indicator for an appropriate representation of the model and therefore is suitable for predictions. The proposed methodology still requires ongoing research, however. This work provides evidence for specific approaches and of areas where additional attention is necessary. It has become apparent that to adequately predict real-scale fire, it is necessary for more sophisticated explanations of heat and mass transfer and descriptions of the interactions between fire and its environment.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item