Read in №3 of 2020 year "Prediction of Dangerous Fire Factors in a Room by Machine Learning Methods"

10 мар 2020


Existing possibilities of the artificial intelligence technologies, in particular the machine learning methods, in relation to the prediction of dangerous fire factors in a room are considered in the article. The analysis is presented concerning the known solutions to the scientific problem, based on which it is established that the methods for predicting dangerous fire factors in a room have advantages and disadvantages. Therefore, the issue of developing new approaches to solving the assigned task remains relevant. The libraries of the high-level programming language Python that are most widely used for machine learning are described. It is established that due to large computational labor intensity, which increases significantly with increasing the mathematical model inputs, the exact formulas that determine the relationship between its parameters are unknown, this is with the exception of the simplest cases. Mathematical predicting model is developed in the form of a system of linear polynomial equations that describes the existing functional relationship between its parameters characterizing the properties of combustible materials. Data processing results presented in the form of the correlation matrix are analyzed. The received results can be used to simulate the processes that describe the thermogasdynamic picture of a fire in a room. Visualization of the obtained results was carried out using the developed multidimensional interactive model, which greatly simplifies the processing and analysis of data, and, also provides for a better understanding of the properties and structure of the relationships of the analyzed object of study. Machine learning methods are an effective tool for predicting dangerous fire factors in a room in addition to already existing predicting methods.

Authors:
  • Kalachin S.V.
    Kalachin S.V.
    Dr. Sci. (Eng.), Prof., s.v.kalachin@mail.ru National Research Ogarev Mordovia State University, Saransk, Russia