Research on Temperature Prediction Algorithm for Water-Oil Interface Layer in Oil Tank Fire Based on Neural Network

The boilover and splashing of oil tank fires is one of the major dangers faced in fire rescue disposal. The temperature change trend of the water-oil interface layer during the combustion process of oil products is one of the important determinants of the occurrence and degree of boilover. Aiming at the threat of boilover and splashing to fire rescue personnel during storage tank fires and improving the boilover early warning capability of storage tank fires, a BP neural network and experimental data-driven temperature prediction method for the water-oil interface layer of oil tank fires is proposed. By carrying out multiple groups of small-scale boilover fire experiments, the bottom temperature of the tank, the wall temperature of the tank, and the temperature of the water-oil interface layer during the whole combustion process are measured. A neural network-based temperature prediction model for the water-oil interface layer of oil tank fires is built, and the prediction ability of the model is tested. The test results show that the R² reaches 0.9891, the RMSE reaches 3.635, and the relative error is within 10%, providing data support for the critical determination of boilover in the disposal of oil storage tank fires.