The article presents a comprehensive analysis of the evolution of neural network architectures for computer vision tasks, including an assessment of the prospects for increasing computational complexity and improving the accuracy of visual information processing algorithms. Modern deep learning methods demonstrate significant potential for application in unmanned maritime navigation systems, ensuring increased autonomy, navigation safety, and resilience of control in dynamically changing marine environments. The aim of the study is to develop a prototype of a Decision Support System (DSS) that provides automated detection, tracking, and trajectory prediction of maritime objects based on a hybrid combination of neural network computer vision methods and state-filtering algorithms. The synergy of computer vision and DSS represents an important step toward the creation of fully autonomous decision-making systems capable of automating visual monitoring processes, enhancing navigation safety, and reducing the impact of the human factor. The scientific novelty of the research lies in a hybrid approach that combines deep neural networks and state-filtering algorithms adapted to maritime monitoring. The most optimal neural network architecture was selected for integration into the DSS with an adaptive Kalman filter. The DSS prototype development applied data preprocessing techniques, augmentation methods, and optimal training regimes under limited computational resources. The software implementation is based on a modified YOLO v11 architecture integrated with an adaptive Kalman filter for the detection, tracking, and trajectory prediction of maritime objects. The developed system supports the processing of images, videos, and real-time video streams. The research results confirm the practical significance of the hybrid and the of its into for maritime and control.
To monitor vapor content and to detect leaks in the coolant circuit of a nuclear power reactor that uses heavy water, a laser photometric method is used, but as a result of H/D exchange in crystals, semi-heavy water molecules HDO can be formed, which as a result reduces transparency and reduces laser radiation power. The work examined the IR transmission and absorption spectra of hexagonal lithium iodate crystals, widely used both in nuclear reactors and in laser beacons. The transparency range of the lithium iodate crystal is quite wide (from 300 nm to 5000 nm), which is most favorable for the perception of the operator's eyes and for the optimal functioning of laser systems. The activation energy and wave numbers of vibrational centers caused by vibrations of the H3O+, OH-, H2O and HDO groups were determined from the IR spectra. It is shown that the band (1450 ... 1650) cm(-1) is a superposiion of two absorption bands centered at similar to 1550 and similar to 1600 cm(-1). It was experimentally established that the ratio of the absorption coefficients of these bands was similar to 1:1 (75:74 cm(-1)) for crystals grown in H2O, and similar to 2:1 (161:82 cm(-1)) for crystals grown in D2O. This allows us to conclude that the bands at 1550 cm(-1) and 1600 cm(-1) are associated with vibrations of H2O and HDO molecules. The conducted studies have proven that IR spectra can be successfully used to diagnose the quality of laser crystals and the presence of heavy and semi-heavy water molecules in them.
Liquefied natural gas (LNG) not only has the potential to optimize the global energy mix but has also become the most widely used alternative fuel for marine propulsion systems due to its advantages, such as low emissions, technical maturity, and high cost-effectiveness. Compared to conventional fuel oil, the use of LNG as a marine fuel can reduce NOx emissions by up to 90%, SOx and particulate matter emissions by nearly 100%, and CO2 emissions by nearly 30%, thereby meeting increasingly stringent IMO regulations on ship emissions and global sustainable development goals. A disadvantage of LNG is the emission of residual gas fuel, incompletely burned in engines. To monitor emissions of harmful substances, it is proposed to measure their concentration in the gas-air environment based on their absorption of infrared (IR) radiation in the mid-infrared range. The results of calibration of the infrared system for one homologue of saturated hydrocarbons are presented. An experimental verification of the admissibility of this approach was performed using the integrated absorption cross-section of the n-hexane molecule. It has been shown that the proposed method of calibrating an infrared system for one component is acceptable for measuring the residues of gas fuel not completely burned in engines, within the established error of 25%. The proposed infrared system, in contrast to the previously described system of laser monitoring of methane molecules using Raman lidar, is preferable because the infrared LEDs used in it are cheaper than lasers, have a small weight and dimensions, and can be integrated into existing equipment with a measurement error of the same order of magnitude.
This paper presents methodological aspects of creating a diagnostic platform for the digital transformation of the technical operation of marine propulsion systems, using the example of Azipod (R) V propulsion and steering systems. The development of diagnostic systems for marine propulsion systems is constantly evolving. Currently, intelligent and digital technologies are being used to improve the analysis of the current technical condition, as well as the detection and prediction of malfunctions in marine propulsion systems during operation. The aim of this paper is to expand the methodological foundations of technical diagnostics for marine propulsion and steering systems, based on modern intelligent data processing technologies, including machine learning methods, digital twins, and approaches to managing the lifecycle of machine learning models. The paper utilizes statistical processing and comparative analysis of data from the monitoring system for Azipod (R) V propulsion and steering systems. The developed criteria for evaluating research objects will help determine the level of technological development of propulsion plant components required for the design and construction of new vessels, as well as lay the foundation for updating the regulations and inspection methods used by the Russian Maritime Register of Shipping. The diagnostic platform being developed as part of the digital transformation of propulsion plant technical operation represents an important foundation and practical tool for the transition from traditional shipping to intelligent and autonomous maritime transport.
