Search Results (280)

This survey provides an in-depth review of large language models (LLMs), highlighting the significant paradigm shift they represent in artificial intelligence. Our purpose is to consolidate state-of-the-art advances in LLM design, training, adaptation, evaluation, and application for both researchers and practitioners. To accomplish this, we trace the evolution of language models and describe core approaches, including parameter-efficient fine-tuning (PEFT). The methodology involves a thorough survey of real-world LLM applications across the scientific, engineering, healthcare, and creative sectors, coupled with a review of current benchmarks. Our findings indicate that high training and inference costs are shaping market structures, raising economic and labor concerns, while also underscoring a persistent need for human oversight in assessment. Key trends include the development of unified multimodal architectures capable of processing varied data inputs and the emergence of agentic systems that exhibit complex behaviors such as tool use and planning. We identify critical open problems, such as detectability, data contamination, generalization, and benchmark diversity. Ultimately, we conclude that overcoming these complex technical, economic, and social challenges necessitates collaborative advancements in adaptation, evaluation, infrastructure, and governance. Full article

With the rapid evolution of 5G wireless communications, millimeter-wave (mmWave) technology has become a crucial enabler for high-speed, low-latency, and large-scale connectivity. As the critical interface for signal transmission, mmWave antennas directly affect system performance, reliability, and application scope. This paper reviews the current state of mmWave antenna technologies in 5G systems, focusing on antenna types, design considerations, and integration strategies. We discuss how the multiple-input multiple-output (MIMO) architectures and advanced beamforming techniques enhance system capacity and link robustness. State-of-the-art integration methods, such as antenna-in-package (AiP) and chip-level integration, are examined for their importance in achieving compact and high-performance mmWave systems. Material selection and fabrication technologies—including low-loss substrates like polytetrafluoroethylene (PTFE), hydrocarbon-based materials, liquid crystal polymer (LCP), and microwave dielectric ceramics, as well as emerging processes such as low-temperature co-fired ceramics (LTCC), 3D printing, and micro-electro-mechanical systems (MEMS)—are also analyzed. Key challenges include propagation path limitations, power consumption and thermal management in highly integrated systems, cost–performance trade-offs for mass production, and interoperability standardization across vendors. Finally, we outline future research directions, including intelligent beam management, reconfigurable antennas, AI-driven designs, and hybrid mmWave–sub-6 GHz systems, highlighting the vital role of mmWave antennas in shaping next-generation wireless networks. Full article

Proportional symbol maps are a frequently used method of thematic cartography. Using an intuitive principle—the larger, the more—provides a simple and precise way of visualizing quantity in maps using geographic information systems (GIS). However, none of the current GIS software provides a proper map legend that could be used to interpret exact phenomenon quantity values from the map in reverse. Cartographers have been designing value scales manually for such a possibility of interpretation. Eventually, they preferred to resign to the accuracy of the interpretation and use the legend offered by the software. The paper describes the development of an easy-to-use online value scale generator for static maps, aiming to eliminate the time-consuming process to make map design more efficient while preserving the precision of cartographic visualization and its subsequent interpretation. The tool consists of a free web platform performing all necessary calculations and rendering an appropriate value scale based on user-defined input parameters. This functionality is performed for most typically used symbol shapes as well as for custom-design shapes provided by the user in SVG vector graphics. The output is then returned in a vector SVG and PDF file format to be used directly in a map legend or possibly edited in graphic software before such a step. The presented tool is therefore independent of which software was used for map design. Within the research, two user experiments were performed to compare generated value scales with simple legends generated in GIS and to gather insights from cartography experts. Full article

This study investigates the influence of electrical actuation parameters on the performance of a morphing composite aerodynamic profile actuated by Shape Memory Alloy (SMA) wires. A fully coupled electro-thermo-mechanical finite element model has been developed to simulate the transient response of NiTi SMA, capturing the nonlinear interplay between temperature evolution, phase transformation, and mechanical deformation under Joule heating. The model incorporates phase-dependent material properties, heat effects, and geometric constraints, enabling accurate prediction of actuation dynamics. To validate the model, a morphing spoiler prototype has been fabricated using high-performance additive manufacturing with a carbon fibre-reinforced polymer. The SMA wires have been pretensioned and electrically actuated at different current levels (3 A and 6 A), and the resulting deformation has been recorded through video analysis with embedded timers. Experimental measurements confirmed the model’s ability to predict both actuation time and displacement, with maximum deflections of 33 mm and 40 mm corresponding to different current inputs. This integrated approach demonstrates an efficient and compact solution for active aerodynamic surfaces without the need for mechanical linkages, enabling future developments in adaptive structures for automotive and aerospace applications. Full article

This study introduces novel semi-active vibration controllers for magnetorheological (MR) fluid-based vibration control systems, specifically a band-pass frequency-shaped semi-active control (FSSC) and a narrow-band FSSC. These algorithms are designed without requiring an accurate damper model or system identification for control current input. Unlike active controllers, the FSSC algorithms treat the MR damper as a semi-active dissipative device, and their control signal is a control current, not a control force. The performance of both FSSC algorithms is evaluated through simulation using a single-degree-of-freedom (SDOF) MR fluid-based engine mount system. A comparative analysis with the classical semi-active skyhook control demonstrates the advantages of the proposed FSSC algorithms. Full article

The precise identification, classification, sorting, and rapid and accurate quality assessment of soybean seeds are extremely important in terms of the continuity of agricultural production, varietal purity, seed processing, protein extraction, and food safety. Currently, commonly used methods for the identification and quality assessment of soybean seeds include morphological analysis, chemical analysis, protein electrophoresis, liquid chromatography, spectral analysis, and image analysis. The use of image analysis and artificial intelligence is the aim of the presented research, in which a method for the automatic classification of soybean varieties, the assessment of the degree of damage, and the identification of geometric features of soybean seeds based on numerical models obtained using a 3D scanner has been proposed. Unlike traditional two-dimensional images, which only represent height and width, 3D imaging adds a third dimension, allowing for a more realistic representation of the shape of the seeds. The research was conducted on soybean seeds with a moisture content of 13%, and the seeds were stored in a room with a temperature of 20–23 °C and air humidity of 60%. Individual soybean seeds were scanned to create 3D models, allowing for the measurement of their geometric parameters, assessment of texture, evaluation of damage, and identification of characteristic varietal features. The developed 3D-CNN network model comprised an architecture consisting of an input layer, three hidden layers, and one output layer with a single neuron. The aim of the conducted research is to design a new, three-dimensional 3D-CNN architecture, the main task of which is the classification of soybean seeds. For the purposes of network analysis and testing, 22 input criteria were defined, with a hierarchy of their importance. The training, testing, and validation database of the SB3D-NET network consisted of 3D models obtained as a result of scanning individual soybean seeds, 100 for each variety. The accuracy of the training process of the proposed SB3D-NET model for the qualitative classification of 3D models of soybean seeds, based on the adopted criteria, was 95.54%, and the accuracy of its validation was 90.74%. The relative loss value during the training process of the SB3D-NET model was 18.53%, and during its validation process, it was 37.76%. The proposed SB3D-NET neural network model for all twenty-two criteria achieves values of global error (GE) of prediction and classification of seeds at the level of 0.0992. Full article

Scene text understanding, serving as a cornerstone technology for autonomous navigation, document digitization, and accessibility tools, has witnessed a paradigm shift from traditional methods relying on handcrafted features and multi-stage processing pipelines to contemporary deep learning frameworks capable of learning hierarchical representations directly from raw image inputs. This survey distinctly categorizes modern scene text recognition (STR) methodologies into three principal paradigms: two-stage detection frameworks that employ region proposal networks for precise text localization, single-stage detectors designed to optimize computational efficiency, and specialized architectures tailored to handle arbitrarily shaped text through geometric-aware modeling techniques. Concurrently, an in-depth analysis of text recognition paradigms elucidates the evolutionary trajectory from connectionist temporal classification (CTC) and sequence-to-sequence models to transformer-based architectures, which excel in contextual modeling and demonstrate superior performance. In contrast to prior surveys, this work uniquely emphasizes several key differences and contributions. Firstly, it provides a comprehensive and systematic taxonomy of STR methods, explicitly highlighting the trade-offs between detection accuracy, computational efficiency, and geometric adaptability across different paradigms. Secondly, it delves into the nuances of text recognition, illustrating how transformer-based models have revolutionized the field by capturing long-range dependencies and contextual information, thereby addressing challenges in recognizing complex text layouts and multilingual scripts. Furthermore, the survey pioneers the exploration of critical research frontiers, such as multilingual text adaptation, enhancing model robustness against environmental variations (e.g., lighting conditions, occlusions), and devising data-efficient learning strategies to mitigate the dependency on large-scale annotated datasets. By synthesizing insights from technical advancements across 28 benchmark datasets and standardized evaluation protocols, this study offers researchers a holistic perspective on the current state-of-the-art, persistent challenges, and promising avenues for future research, with the ultimate goal of achieving human-level scene text comprehension. Full article

This study investigates the impact of arc-hardening parameters on a groove-shaped S45C steel tube, with a focus on surface hardness and microstructure. According to the findings, when arc quenching occurs, the tube’s surface hardness increases significantly compared to its original hardness. The surface layer hardness can increase to 50.3 HRC, which is 3.4 times greater than the untreated surface. Changing arc quenching parameters such as current intensity, gas flow rate, arc length, scan speed, heating angle, and cooling angle causes a variation in surface hardness due to the balance of heat input and cooling value. Moreover, the microhardness distribution is divided into three zones: the hardened zone (with a high hardness value), the heat-affected zone (HAZ), which has rapidly declining hardness, and the base metal (with a low hardness value). The hardened zone could have a hardness with a load of 0.3 N of 440 HV and a case depth of about 900 μm. The next zone is the HAZ, where the hardness with a load of 0.3 N drops significantly. The hardness in the base metal zone recovers to its original value of 152 HV. Interestingly, the microstructure, under the hardness distribution, illustrates the relationship between the hardness value and its phases. The hardened zone consists of martensite and residual austenite phases, resulting in a high hardness value. The bainite phase constitutes the HAZ, which correlates to the zone of rapid hardness reduction. Finally, the base metal zone has ferrite and pearlite microstructures, indicating the softest zone. The investigation’s findings may increase our understanding of the arc-hardening process and widen its industrial applications. Full article

This study explores how different electrode shapes affect plasma flow in a non-transferred plasma torch. Various cathode geometries—including conical, tapered, flat, and cylindrical—were examined alongside stepped anode designs. A 2D axisymmetric computational model was employed to assess the impact of these shapes on plasma behavior. The results reveal that different cathode designs require varying current levels to maintain a consistent power output. This paper presents the changes in electric conductivity and electric potential for different input currents across the arc formation path (from the cathode tip to the anode beginning) and relating to Ohm’s law. Significant variations in plasma jet velocity and temperature were observed, especially near the cathode tip. The study concludes by evaluating thermal efficiency across geometry configurations. Flat cathodes demonstrated the highest efficiency, while the anode shape had minimal impact. Full article

THz communication stands as a pivotal technology for 6G networks, designed to address the critical challenge of data demands surpassing current microwave and millimeter-wave (mmWave) capabilities. However, realizing Tbps and kilometer-range transmission confronts the “dual attenuation dilemma” comprising severe free-space path loss (FSPL) (>120 dB/km) and atmospheric absorption. This review comprehensively summarizes our group′s advancements in overcoming fundamental challenges of long-distance THz communication. Through systematic photonic–electronic co-optimization, we report key enabling technologies including photonically assisted THz signal generation, polarization-multiplexed multiple-input multiple-output (MIMO) systems with maximal ratio combining (MRC), high-gain antenna–lens configurations, and InP amplifier systems for complex weather resilience. Critical experimental milestones encompass record-breaking 1.0488 Tbps throughput using probabilistically shaped 64QAM (PS-64QAM) in the 330–500 GHz band; 30.2 km D-band transmission (18 Gbps with 543.6 Gbps·km capacity–distance product); a 3 km fog-penetrating link at 312 GHz; and high-sensitivity SIMO-validated 100 Gbps satellite-terrestrial communication beyond 36,000 km. These findings demonstrate THz communication′s viability for 6G networks requiring extreme-capacity backhaul and ultra-long-haul connectivity. Full article

With the increase in the penetration rate of distributed sources and loads, the sensor monitoring data is increasing dramatically. Power grid maintenance services require a rapid response in power quality data analysis. To achieve a rapid response and highly accurate classification of power quality disturbances (PQDs), this paper proposes an efficient classification algorithm for PQDs based on Lissajous trajectory (LT) and a lightweight DenseNet, which utilizes the concept of Lissajous curves to construct an ideal reference signal and combines it with the original PQD signal to synthesize a feature trajectory with a distinctive shape. Meanwhile, to enhance the ability and efficiency of capturing trajectory features, a lightweight L-DenseNet skeleton model is designed, and its feature extraction capability is further improved by integrating an attention mechanism with L-DenseNet. Finally, the LT image is input into the fusion model for training, and PQD classification is achieved using the optimally trained model. The experimental results demonstrate that, compared with current mainstream PQD classification methods, the proposed algorithm not only achieves superior disturbance classification accuracy and noise robustness but also significantly improves response speed in PQD classification tasks through its concise visualization conversion process and lightweight model design. Full article

In this study, a novel shape memory alloy actuated bearing active preload system (SMA-BAPS) was proposed and experimentally demonstrated. SMA actuators placed in a single or antagonistic configuration were employed to drive the screw pair and thus fulfill one-way or bidirectional preload adjustment. Moreover, the self-locking screw pair was used to maintain the bearing preload without external energy input. To determine the parameters of screw pair and SMA actuators, a detailed design process was conducted based on analytical models of the proposed system. Finally, a screw pair with a lead of 3 mm and SMA actuators with a diameter of 0.5 mm and a length of 130 mm were adopted. Prototype tests were conducted to validate and evaluate the performance of the preload adjustment with the SMA-BAPS. The resistive torque and the natural frequency of spindles were recorded to represent the preload level of the bearing. Through the performance tests, the SMA-BAPS induced a maximum 47% variation in the resistive torque and a 20% variation in the spindle’s natural frequency. The response time of the SMA-BAPS was less than 5 s when the heating current of 5 A was applied on the SMA actuator. This design highlighted the compact size, quick response, as well as the bidirectional preload adjustment, representing its potential use in aerospace mechanisms and advanced motors. Full article

Background/Objectives: Neuronal oscillations play a key role in the symptoms of Parkinson’s disease (PD). This study investigates the effects of random synaptic inputs, their correlations, and the interaction with synaptic dynamics and spike timing-dependent plasticity (STDP) on the membrane potential and firing patterns of subthalamic nucleus (STN) neurons, both in healthy and PD-affected states. Methods: We used a modified Hodgkin–Huxley model with a Langevin stochastic framework to study how synaptic conductance, random input fluctuations, and STDP affect STN neuron firing and membrane potential, including sensitivity to refractory period and synaptic depression variability. Results: Our results show that random inputs significantly affect the firing patterns of STN neurons, both in healthy cells and those with PD under DBS treatment. STDP, along with random refractory periods and fluctuating input currents, increases the irregularity of inter-spike intervals (ISIs) in output neuron spike trains. Sensitivity analyses highlight the key role of synaptic depression and refractory period variability in shaping firing patterns. Combining random inputs with STDP boosts the correlation between neuron activities. Furthermore, at fixed input noise levels, the model’s output closely matches experimental firing rate and ISI variability data from PD patients and animals, with statistical tests confirming significant effects of STDP on firing regularity. Conclusions: The findings suggest that the stochastic dynamics of STN neurons, combined with STDP, are crucial for shaping neuronal firing patterns in both healthy and PD-affected states. These insights improve our understanding of how noise and plasticity contribute to neural function and dysfunction, with implications for PD symptom management. Full article

The output energy of Laser Active Optical Systems (LAOSs), in which image brightness is amplified within the laser-active medium, is always higher than the input energy. This contrasts with conventional optical systems (OSs). As a result, a LAOS enables the creation of laser beams with tailored energy distribution across the aperture, making them ideal for material processing applications. This concept was first successfully implemented using metal vapor lasers as the gain medium. In these systems, material processing was achieved by using a laser beam that either carried the required energy profile or the image of the object itself. Later, other laser media were utilized for LAOSs, including barium vapor, strontium vapor, excimer XeCl lasers, and solid-state media. Additionally, during the development of these systems, several modifications were introduced. For example, Space-Time Light Modulators (STLMs) and CCD cameras were incorporated, along with the use of multipass amplifiers, disk-shaped or thin-disk (TD) solid-state laser amplifiers, and other advancements. These techniques have significantly expanded the range of power, energy, pulse durations, and operating wavelengths. Currently, TD laser amplifiers and STLMs based on Digital Light Processor (DLP) technology or Digital Micromirror Devices (DMDs) enhance the potential to develop LAOS devices for Subtractive and Additive Technologies (ST, AT), applicable in both macromachining (cutting, welding, drilling) and micro-nano processing. This review presents comparable characteristics and requirements for these various LAOS applications. Full article

The four-switch buck–boost (FSBB) converter, which possesses both step-up and step-down capabilities, is highly suitable for applications where input and output voltages have overlapping ranges. Correspondingly, the current shaping control (CSC) strategy is investigated for the FSBB converter, which shapes a quadrilateral inductor current waveform featuring the minimum RMS value to improve efficiency and power density. However, the small-signal model for the CSC algorithm has not yet been established, and the traditional and common modeling method requires considering multiple duty cycles and phase shifts of the FSBB converter, whose calculation is complex and inconvenient to use. For the special case of the CSC strategy using cycle-by-cycle current detection, an additional constraint of the averaged volt-second on the inductor can be regarded as zero, making the inductor current no longer a variable of the state-space, which eliminates the pole generated by the inductor and reduces the order of the small-signal model. Thus, this paper greatly simplifies the computation and design of the compensator by using the constraint condition mentioned above. This one-pole first-order model is simplified, maintains enough accuracy in the low-frequency domain, and can be corrected using only a simple PI controller. Finally, a prototype of the 300 W FSBB converter under the digital CSC algorithm was built to validate the precision and dynamic performance of the proposed first-order small-signal model. Full article