This paper proposes a concept of miniature multi-bore coaxial reactors with potential applications in microwave-assisted combinatorial chemistry. The reactor consists of a shielded cylindrical dielectric core with multiple bores. Current-carrying wires placed several or only one bore, while others support the flowing liquids that are heated in parallel by the wire’s electrical field. Initially, the design concept is confirmed through COMSOL electromagnetic numerical simulations of modes propagating in multi-bore coaxial waveguides. Experimental proofs are obtained by investigating the microwave heating of flowing ethanol, methanol, and reverse-osmosis water in a two-bore alumina eccentric coaxial waveguide, where the second channel contains an open-ended current-carrying wire. Joint analyses of temperature profiles, measured by an optical fiber sensor at three points of a PTFE liquid-filled tube installed inside a bore, and of input and reflected power curves measured by microwave meters, are performed. The results demonstrate adequate parallel heating of liquids using an open-ended wire heater, a configuration that can be extended to other multiple-bore coaxial reactors. It is expected that the proposed concept, confirmed by experiments and simulations, will find applications in multi-tube liquid heating and MW combinatorial chemistry.
The development of reliable control systems for unmanned aerial vehicles (UAVs) requires accurate modeling of dynamics and a control architecture capable of handling nonlinearity, external disturbances, and parameter uncertainty. This study addressed that challenge through a comprehensive survey of UAV control strategies, including classical linear controllers (such as PID, LQR, and MPC), advanced nonlinear methods [such as backstepping, sliding mode control (SMC), H-infinity, and active disturbance rejection control (ADRC)], and intelligent control algorithms like fuzzy logic and artificial neural networks (ANN). Based on this, the paper proposed a hybrid ANN-PID controller, where the neural network dynamically adjusts the PID coefficients to enhance adaptability and robustness. The method was evaluated through simulations on the UAV model using Euler–Lagrange dynamics. The quantitative results showed that ANN-PID achieves the lowest steady-state error (0.0229 m), nearly equivalent to PID (0.0230 m) but significantly superior to LQR (0.0807 m, an improvement of about 72%). Regarding rise time, ANN-PID responds quickly (~2.01 s), similar to PID and much faster than LQR (~7.4 s). Although the overshoot of all controllers is high, ANN-PID still achieves a lower value than PID and significantly reduces it compared to LQR under noisy conditions. These results affirm the superiority of the ANN-PID hybrid structure in UAV control, especially in nonlinear and complex dynamic environments.
Appearance defect detection is essential for ensuring cigarette quality during production. Reaching high-precision and lightweight automated cigarette appearance defect detection has long been manufacturers' key focus. However, existing methods struggle to balance detection accuracy and speed effectively. This paper proposes a high-performance detection model for cigarette defects, named cigarette defect YOLO (CD-YOLO), which builds upon the YOLOv10 network with three major improvements. First, an intra-scale feature interaction (ISFI) module is designed to enhance the model's ability to distinguish different defects. Subsequently, a multi-scale feature fusion (MSFF) network is developed to improve the model's performance in recognizing small-scale and subtle defects. Finally, a lightweight group convolution detection head (LGCDH) is implemented to substantially reduce the model's computational complexity and parameter count, accelerating detection speed. The experimental results demonstrate that the CD-YOLO model achieves a favorable trade-off between accuracy and speed, maintaining a detection speed exceeding 500 FPS, with a mAP@0.5 of 96.2%. Additionally, a novel data augmentation strategy is introduced in this paper, employing low-rank adaptation (LoRA) to fine-tune a pretrained stable diffusion model, which generates synthetic defect samples to alleviate data scarcity.
In recent years, the use of drones for cargo transport has experienced rapid development, driven by the need to reduce the environmental impact of traditional delivery systems and improve access to essential services, particularly in complex urban environments. In this study, we aimed to present the first experimental results for the development of a drone prototype characterized by low energy consumption and the integration of advanced technologies, with attention to the electronic components and embedded algorithms. The combination of cutting-edge technologies with such advanced algorithms enabled us to witness a reduction in the carbon footprint in terms of resource optimization, due especially to the neural network algorithms suitable for path analysis and obstacle detection, and the cutting-edge processor that enables us to manage and analyze the data derived from the different sensors. The design of this drone focuses on minimizing environmental impact through optimized energy resources by using high-efficiency components, including brushless DC motors and lithium polymer (LiPo) batteries integrated with perovskite photovoltaic cells that extend flight autonomy while reducing dependence on traditional energy sources. The drone's body, made of lightweight and durable materials, improves aerodynamic efficiency without compromising structural strength. Advanced artificial intelligence algorithms optimize flight paths, prevent collisions, and adapt dynamically to environmental conditions. Thanks to computing platforms such as the NVIDIA Jetson Nano and equipped with advanced sensors like LiDAR, accelerometers, and gyroscopes, the drone processes data in real-time, enhancing its autonomous operation and ensuring precise and stable navigation. The proposed prototype represents a significant step toward a future where technology supports sustainability, providing innovative solutions for modern society's needs aiming to reduce urban traffic and pollutant emissions, offering an efficient and ecofriendly solution for urban logistics.
A wireless power transfer system with a single transmitter fed by a voltage source and a single receiver with a load is widely studied to increase load power. A system with two transmitters is also studied, especially by controlling currents through transmitter coils to maximize load power, which means current sources to drive transmitter coils are assumed instead of the voltage source. This paper presents a method for adjusting voltage amplitude and phase of each voltage source embedded in two transmitters to maximize load power. The maximized load power is also presented in an explicit form based on circuit constants of the system. As a result, advantages of the two transmitters system over the single transmitter system are revealed from the perspective of efficient power delivery.
An Energy Management System (EMS) in a direct-current (DC) microgrid system is essential to manage renewable energy sources (RES), stored energy units, and demand load. However, the conventional load-following (LF)-based EMS strategy presents several issues due to its integration with proportional-integral (PI) controllers. These controllers have weak performance under sudden irradiation or demand load changes, which result in high overshoot, low convergence speed, and unwanted energy losses in DC microgrid responses. Therefore, this article introduces an improved LF strategy using the terminal-slide mode control (TSMC) method. Moreover, the studied power system consists of a photovoltaic (PV) system, a hybrid energy storage system (HESS) using lithium-ion batteries, and supercapacitors (SCs). The suggested EMS strategy aims to reduce the fluctuation of the grid voltage and enhance the reliability of the system under different irradiance and demand variations. It employs voltage regulation for the DC bus using a robust TSMC instead of using the classical PI controllers. The simulation was conducted using MATLAB/Simulink software. The obtained results indicate that the proposed LF-TSMC strategy can cancel the voltage overshoot, offer better settling time, and provide higher efficiency. Finally, the presented EMS introduces superior dynamics with a settling time of < 0.1 s and overshoot percentage of 1%, compared with a settling time of 0.45 s and overshoot percentage of 2.5% for the classical LF-PI method.
In response to issues such as high-dimensional sparsity, missing semantic information, and ambiguous topic boundaries in traditional methods, in this paper, we investigated a document expert information extraction method based on Word2Vec and Transformer, aiming to enhance the semantic accuracy and clustering effectiveness of document expert information extraction. First, semantic embedding vectors were generated through the Word2Vec model, effectively reducing high-dimensional sparsity and enhancing the semantic representation capability of documents. Second, the Transformer algorithm was used to extract expert information from document vectors, achieving effective differentiation between topics. Experiments showed that document semantic embedding based on Word2Vec can significantly improve the performance of Transformer in expert information extraction. Compared to the traditional TF-IDF + Transformer method, this approach demonstrates superior performance in topic consistency and semantic capture.