Application of Machine Learning in Identifying the Burning Stage of Oil Tank Fire

The oil tank fires have the characteristics of long combustion time, and large temperature change range, which makes it prone to developing into violent reactions such as boilover and splashing. Accurate judgment of the combustion stage is very important for making disposal measures. In this paper, multiple sets of small scale oil tank fire experiments were conducted to monitor the temperature changes at different positions of the oil tank. A method that combines machine learning algorithms with experimental data was applied to identify the combustion stages of oil tank fires. A classification model based on the combination of convolutional neural network (CNN) and long short-term memory (LSTM) network has been proposed, and its performance was compared and analyzed with four classic machine learning models. The results show that the average recognition accuracy of the proposed method for each stage reaches 94.5%. It is higher than other traditional machine learning models. It can be used to provide a reference for identifying the combustion stages of oil storage tank fires.