Study On Dynamic Spread And Ventilation Control Of Fire In Multi-Branch Tunnel

With the rapid development of China’s modernization process,the rational use of underground space has led to the continuous emergence of multi branch tunnels.Although the development of multi branch tunnels has brought convenience to people’s lives,once a fire accident occurs,the smoke is difficult to control,and its complex structure also poses a huge challenge for emergency rescue.In order to reduce the difficulty of smoke control,emergency evacuation,and rescue in multi branch tunnel fires,a typical multi-branch tunnel-three branch tunnel is designed based on the actual tunnel size,and a small-scale experimental platform with a1:20 scale model is independently developed.By combining theoretical analysis,experimental testing,and numerical simulation,the temperature characteristics of fire smoke in multi-branch tunnels,the characteristics of smoke countercurrent,and the critical wind speed and dynamic conditions for suppressing smoke countercurrent are studied.Conduct fire experiments using a self-developed 1:20 small-sized three branch tunnel to study the effects of ventilation,bifurcation angle,and branch slope on the maximum temperature rise and longitudinal temperature rise distribution of the ceiling inside the multi-branch tunnel,and establish corresponding prediction models.Through experimental observation and measurement of flue gas temperature in tunnels with different bifurcation angles and branch slopes,under the conditions of heat release rate and longitudinal wind speed of each fire source,the influence of bifurcation angles and branch slopes on temperature characteristics is analyzed.Based on theoretical analysis and experimental data,a nonlinear regression model is established to predict the maximum temperature rise and longitudinal distribution of temperature rise in the ceiling of a multi-branch tunnel.The model considers the effects of tunnel longitudinal ventilation,bifurcation angle,and branch slope,respectively.Using the self-developed 1:20 small-sized three branch tunnel,the study investigates the length of the main flue gas countercurrent at different bifurcation angles and branch slopes.Based on the conservation of energy,the force acting on the upstream countercurrent flue gas of the fire source is analyzed,and an expression for the length of the countercurrent flue gas is obtained.Through experimental observation,it was found that changing the tunnel structure has a significant impact on the length of flue gas countercurrent,which is nonlinear,with the influence of bifurcation angle being more significant.Changing the wind speed and heat release rate of the ignition source can also cause changes in the distance of the flue gas countercurrent.Based on theoretical analysis and experimental measurement results,a prediction model for the length of flue gas countercurrent considering the complex structure inside the tunnel is constructed.Using the self-developed 1:20 small-sized three branch tunnel,study the critical wind speed for longitudinal ventilation and the power required to suppress flue gas countercurrent under different bifurcation angles and branch slopes.According to the bernoulli equation,analyze the pressure situation inside the tunnel during longitudinal ventilation and obtain a prediction model for the critical wind speed.Based on experimental observations of wind speeds under conditions where the heat release rate of different fire sources approaches a critical state,determine the dynamic conditions required to consider the longitudinal ventilation of complex structures in the tunnel to suppress smoke countercurrent.Through numerical simulation,it was verified that the calculated dynamic conditions meet the requirements for suppressing flue gas countercurrent.The LSTM neural network is used to study the prediction model of fire smoke flow and temperature distribution in multi branch tunnels.The temperature data was collected through the self-developed 1:20 small-scale fire experiment,and the deep neural network LSTM was used to learn and train from the experimental fire database,and different fire source types were tested.It was found that the model could predict the temperature distribution in the tunnel well.Finally,the model’s early prediction ability(10s,20 s,and 30 s in advance)was tested,and the prediction results showed high accuracy with relative errors within ± 10%.This dissertation has 107 figures,22 tables,and 155 references.