A predictive model for centerline temperature in electrical cabinet fires

Nuclear power plants are critical components of sustainable urban energy systems. However, the electrical cabinet, as a critical power distribution component in nuclear power plants, poses a significant threat to reactor safety during fire events. Accurately predicting the obstructed plume centerline temperature induced by electrical cabinet fires is crucial for evaluating plant safety, protecting critical equipment, and managing reactor risk. The Fire Dynamics Simulator is employed to model cabinet fires, with heat release rate (HRR) and top opening size identified as varying parameters. Buoyant jet theory is applied to model obstructed plumes, demonstrating that the dimensionless centerline temperature rise is related to Froude number, dimensionless height, and dimensionless initial temperature. A comprehensive framework integrating physics-based, mathematics-based, and machine learning methodologies is established to derive a predictive model for centerline temperature. Specifically, conservation laws and dimensional analysis are utilized to establish the correlation of obstructed plume temperature at top opening. Levenberg-Marquardt algorithm is then used to develop a correlation for the obstructed plume velocity at the top opening. Thus, algebraic models link the three independent variables with HRR and opening size. Subsequently, a backpropagation neural network is employed to model the nonlinear relationship between the dimensionless centerline temperature rise and three independent parameters. Compared to traditional methods, predictive results indicate a reduction in error ranging from 36.1 % to 39.2 % under various HRR conditions. Given that HRR and opening size are known parameters, the resulting model proves highly applicable for risk assessment in nuclear industry engineering practices.