RUDN Journal of Engineering Research

The automation of agricultural tasks using mobile robots is becoming increasingly important for precision farming. Navigation in complex, dynamic, and unstructured environments, such as fields with uneven terrain, dense vegetation, and obstacles such as rocks or irrigation systems, presents a significant challenge. Therefore, efficient path planning and collision avoidance algorithms are of particular significance. The primary objective of this study is to implement and evaluate the RRT* algorithm for coordinated path planning. A novel contribution lies in improving the planning efficiency and path optimality compared to the standard RRT algorithm, particularly through enhanced generation of time-optimal trajectories and robust collision avoidance in complex terrain. The RRT*-based collision avoidance system was evaluated through MATLAB simulations, testing its performance in scenarios with a high obstacle density typical of agricultural environments. The simulation results demonstrated a 94% success rate for trajectory planning and collision avoidance, indicating high performance potential in complex agricultural landscapes. RRT* was shown to be a highly effective trajectory planning solution for multi-wheeled agricultural robots, outperforming standard RRT. It successfully delivers optimized, collision-free trajectories in unstructured environments, offering a robust foundation for autonomous navigation. The 94% success rate obtained in the simulation validates its potential and indicates the need for further research and field testing.

The main objective of this study is to develop an algorithm for the adaptive control of dynamic processes in an organizational and technical system within a neural-network computational framework. To solve the set linear programming problem, a dynamic-static network was used to provide clear interpretation of neural network solutions and to simplify the implementation of inequality constraints. The algorithm for solving the optimization problem of adaptive control of dynamic processes in an organizational and technical system includes the following stages: preparation of the initial problem data and their transformation into a form convenient for representation in the neural network framework; synthesis of the neural network; organizing accesses to the neural network with different initial states of the dynamic neurons within the allotted time and recording the obtained solutions; selecting the best neural network solution from all those obtained when accessing the network and interpreting the selected solution within the scope of the original problem. The proposed algorithm addresses the problem of adaptive control of complex organizational and technical systems that include a large number of elements and subsystems and are characterized by multiple structures defining various types of relationships between these elements and subsystems.

In the context of industrial digital transformation, the transition from scheduled preventive maintenance to condition-based maintenance is a key factor in increasing the reliability of electric drives. Industrial frequency converters (FCs) are critical components of process chains; however, existing maintenance procedures are often economically inefficient and do not prevent sudden failures of power electronics. The aim of this study is to develop a methodology for assessing the remaining useful life (RUL) of critical components of industrial frequency converters, namely IGBT modules and DC-link capacitors, based on a hybrid analysis of real-time operational data. The authors combine physical failure models with deep learning algorithms (CNN-LSTM). To overcome the limitations of the closed architecture of industrial controllers, a two-tier data collection system based on edge computing principles is proposed. Diagnostics are performed by indirectly assessing the saturation voltage drift (VCE(ON)) and equivalent series resistance (ESR) through an analysis of spectral distortions in the output current and DC link voltage ripple. A converter technical condition classification matrix with quantitative degradation thresholds has been developed . A numerical experiment based on a historical dataset from a chemical industry plant showed that the proposed hybrid model reduces the RUL prediction error to 12-15% compared to traditional extrapolation methods, enabling the identification of pre-failure conditions conditions 160-200 hours before failure. The implementation of the developed model will enable a full transition to condition-based maintenance, thereby improving the efficiency of maintenance and repair activities.

The precise synchronization of the main drive shaft and the driven shaft of a screw conveyor is a critical industrial task, upon which the overall operability of the screw conveyor depends, as well as the maintenance of the drive shafts, sprockets, and belt in a functional state. This study presents a comparative analysis of the experimental application and simulation results of a PID controller for synchronizing the rotation of two shafts using incremental encoders and frequency converters. Modern methods of automated control systems are adopted as the methodological basis of the research, including the adjustment of parameters for frequency converters and PID controllers, and the digital processing of encoder signals. The objective of the study is to compare the accuracy and dynamic characteristics obtained during the simulation phase with the results obtained from the real system. The research task is to develop a mathematical model of the system, tune optimal PID controller parameters for both the mathematical model and the real system, and analyze the resulting discrepancies between the data, identifying their causes. As a result of the study, a mathematical model of the synchronization system was developed, optimal PID controller parameters were determined, the main factors causing discrepancies between simulation and the real system operation were analyzed, and recommendations were formulated to improve its accuracy.

Fatigue failure under cyclic loading is a critical concern for design engineers because tensile residual stresses can significantly reduce the service life of a component. Additive manufacturing processes, such as three-dimensional (3D) printing, inherently generate such stresses owing to high thermal gradients, necessitating a thorough evaluation of the residual stresses in printed components. To address this issue, this study developed a finite element model that simulates the 3D printing process for a standard fatigue test specimen. The simulation was performed for Inconel 625 with the build orientation aligned with the specimen thickness direction (PDT). The calculation results provide detailed distributions of the full residual stress tensor and von Mises equivalent stress at the final stage of fabrication. The analysis shows that although significant tensile stresses develop in the longitudinal direction of the gauge section, the selected PDT orientation successfully localizes the highest stress concentrations in the regions of contact with the support structures, away from the critical testing region.