Instruments and Systems: Monitoring, Control, and Diagnostics Annotation << Back Generation of Initial Data for Machine Learning of Models for Evaluation of Parameters of Gas-Liquid flows in Fire-Fighting Robotic Systems I.N. POZHARKOVA The article focuses on the development of a methodology for generating input data (training, validation, and test datasets) that meet specified quality criteria for solving the problem of estimating parameters in gas-liquid flows within firefighting robotic systems using machine learning. The study describes the technology for preparing an experimental test site and conducting field tests, the results of which are used to construct a validation dataset based on photogrammetry. This dataset accurately captures the motion of a fire-extinguishing agent jet from a firefighting robot nozzle under various control and disturbance conditions, enabling the evaluation of the adequacy of numerical models employed in the proposed approach for generating training and verification datasets. A methodology for selecting appropriate computational fluid dynamics methods and their parameters is presented, ensuring high accuracy and computational efficiency. The results of testing the developed technology for generating input data for machine learning models estimating the parameters of a fire-extinguishing agent discharged from an LS-P20U fire monitor under crosswind conditions are provided. Keywords: machine learning, computational fluid dynamics, fire robots, training dataset, verification, validation. DOI: 10.25791/pribor.3.2026.1656 Pp. 14-20.