Methodology of Real-Time Forecast of Fire-Extinguishing Agent Jet Trajectory from Fire Robot Monitor

The study is dedicated to the development of a methodology of real-time forecasting of the movement trajectory of a fire-extinguishing agent from a fire gun carriage considering various external disturbances. Multiple problems of fire robot management, particularly targeting a jet to a fire source, protected equipment or building structures, precipitated clouds of poisonous or radioactive gases, vapors, dusts, etc., can be solved based on the methodology. One of the main problems in the applied field considered is the impact on the movement of a fire-extinguishing agent that can significantly change its trajectory such as wind, low temperatures, etc., by external factors. At the same time, it is not always possible to compensate for their impact by feedback control due to insufficient visibility in conditions of strong smoke. The goal of the study is to assess the efficiency of the solution for the problem using the methodology of forecasting the movement trajectory of a fire-extinguishing agent from a fire robot barrel. Forecast model machine learning is the basis of the methodology proposed. At the same time, the training sample is selected using methods of computational hydrodynamics whereas its adequacy is evaluated by comparison with results of full-scale tests. The article provides a functional scheme of the developed methodology as well as the algorithms to solve its basic problems, i.e., building a forecast model and initial data generating: training, verification (test), and validation (experimental) samples. To demonstrate the efficiency of the methodology developed within the frameworks of the study, the movement trajectories of fire-extinguishing agents from a fire barrel obtained by the trained model and based on the results of a full-scale experiment have been compared. Values of the absolute mean square error of the forecast and its calculation time have also been presented.