Nothing Special   »   [go: up one dir, main page]
Next Issue
Volume 14, February-1
Previous Issue
Volume 14, January-1
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
5.3
2.6
Journals
Electronics
Volume 14
Issue 2
share announcement

Electronics, Volume 14, Issue 2 (January-2 2025) – 188 articles

Cover Story (view full-size image): Balancing currents between paralleled silicon carbide (SiC) MOSFETs requires innovative approaches, as current balancing techniques proposed by literature cannot autonomously minimize the entire current imbalance from the first operating cycles without requiring the extraction of the device parameters. Also, the effectiveness of most methods has been tested under a symmetrical printed circuit board (PCB) layout. This paper proposes a novel and self-sustaining current balancing approach that balances currents between parallel-connected SiC MOSFETs. The proposed solution actively detects and automatically minimizes the entire current imbalance under both symmetrical and asymmetrical layouts. Additionally, it can be realized without measuring the parameters of SiC MOSFETs. View this paper
  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
34 pages, 361 KiB  
Article
Model Checking Using Large Language Models—Evaluation and Future Directions
by Sotiris Batsakis, Ilias Tachmazidis, Matthew Mantle, Nikolaos Papadakis and Grigoris Antoniou
Electronics 2025, 14(2), 401; https://doi.org/10.3390/electronics14020401 - 20 Jan 2025
Viewed by 533
Abstract
Large language models (LLMs) such as ChatGPT have risen in prominence recently, leading to the need to analyze their strengths and limitations for various tasks. The objective of this work was to evaluate the performance of large language models for model checking, which [...] Read more.
Large language models (LLMs) such as ChatGPT have risen in prominence recently, leading to the need to analyze their strengths and limitations for various tasks. The objective of this work was to evaluate the performance of large language models for model checking, which is used extensively in various critical tasks such as software and hardware verification. A set of problems were proposed as a benchmark in this work and three LLMs (GPT-4, Claude, and Gemini) were evaluated with respect to their ability to solve these problems. The evaluation was conducted by comparing the responses of the three LLMs with the gold standard provided by model checking tools. The results illustrate the limitations of LLMs in these tasks, identifying directions for future research. Specifically, the best overall performance (ratio of problems solved correctly) was 60%, indicating a high probability of reasoning errors by the LLMs, especially when dealing with more complex scenarios requiring many reasoning steps, and the LLMs typically performed better when generating scripts for solving the problems rather than solving them directly. Full article
(This article belongs to the Special Issue Advances in Information, Intelligence, Systems and Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>Graph of problem 7 (Source: ProB model checking system examples).</p> Full article ">Figure 2
<p>Solution of problem 8.</p> Full article ">Figure 3
<p>Solution of problem 11.</p> Full article ">
29 pages, 5038 KiB  
Article
An Evolutionary Deep Learning Framework for Accurate Remaining Capacity Prediction in Lithium-Ion Batteries
by Yang Liu, Liangyu Han, Yuzhu Wang, Jinqi Zhu, Bo Zhang and Jia Guo
Electronics 2025, 14(2), 400; https://doi.org/10.3390/electronics14020400 - 20 Jan 2025
Viewed by 420
Abstract
Accurate remaining capacity prediction (RCP) of lithium-ion batteries (LIBs) is crucial for ensuring their safety, reliability, and performance, particularly amidst the growing energy crisis and environmental concerns. However, the complex aging processes of LIBs significantly hinder accurate RCP, as traditional prediction methods struggle [...] Read more.
Accurate remaining capacity prediction (RCP) of lithium-ion batteries (LIBs) is crucial for ensuring their safety, reliability, and performance, particularly amidst the growing energy crisis and environmental concerns. However, the complex aging processes of LIBs significantly hinder accurate RCP, as traditional prediction methods struggle to effectively capture nonlinear degradation patterns and long-term dependencies. To tackle these challenges, we introduce an innovative framework that combines evolutionary learning with deep learning for RCP. This framework integrates Temporal Convolutional Networks (TCNs), Bidirectional Gated Recurrent Units (BiGRUs), and an attention mechanism to extract comprehensive time-series features and improve prediction accuracy. Additionally, we introduce a hybrid optimization algorithm that combines the Sparrow Search Algorithm (SSA) with Bayesian Optimization (BO) to enhance the performance of the model. The experimental results validate the superiority of our framework, demonstrating its capability to achieve significantly improved prediction accuracy compared to existing methods. This study provides researchers in battery management systems, electric vehicles, and renewable energy storage with a reliable tool for optimizing lithium-ion battery performance, enhancing system reliability, and addressing the challenges of the new energy industry. Full article
(This article belongs to the Special Issue Advanced Machine Learning, Pattern Recognition, and Deep Learning Technologies: Methodologies and Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>The framework of the proposed evolutionary deep learning method.</p> Full article ">Figure 2
<p>Sliding window technique for time-series data.</p> Full article ">Figure 3
<p>Architecture of the TCN-BiGRU-Attention model. (<b>a</b>) Represents input sequence data. (<b>b</b>) Extracts temporal features with dilated convolutions, capturing dependencies at different scales. d = 1, d = 2, d = 4, …, indicate the dilation rates, corresponding to progressively larger receptive fields for long-term dependency modeling. The white and grey squares are used to highlight dilated convolutions in the TCN. (<b>c</b>) Processes sequences in both forward and backward directions, extracting contextual dependencies. (<b>d</b>) Highlights critical features from the BiGRU output by assigning attention weights (α<sub>1,1</sub>, α<sub>1,2</sub>, …). (<b>e</b>) Provides the final prediction after passing through the fully connected layer indicated by cross signs.</p> Full article ">Figure 4
<p>Flowchart of the hybrid optimization algorithm.</p> Full article ">Figure 5
<p>Final test results of the twelve experimental groups. (<b>a</b>–<b>d</b>) Results on the NASA dataset. (<b>e</b>–<b>h</b>) Results on the Half-Cell (Si/CNTs) dataset. (<b>i</b>–<b>l</b>) Results on the Half-Cell (Si/Graphene) dataset.</p> Full article ">Figure 5 Cont.
<p>Final test results of the twelve experimental groups. (<b>a</b>–<b>d</b>) Results on the NASA dataset. (<b>e</b>–<b>h</b>) Results on the Half-Cell (Si/CNTs) dataset. (<b>i</b>–<b>l</b>) Results on the Half-Cell (Si/Graphene) dataset.</p> Full article ">Figure 5 Cont.
<p>Final test results of the twelve experimental groups. (<b>a</b>–<b>d</b>) Results on the NASA dataset. (<b>e</b>–<b>h</b>) Results on the Half-Cell (Si/CNTs) dataset. (<b>i</b>–<b>l</b>) Results on the Half-Cell (Si/Graphene) dataset.</p> Full article ">Figure 5 Cont.
<p>Final test results of the twelve experimental groups. (<b>a</b>–<b>d</b>) Results on the NASA dataset. (<b>e</b>–<b>h</b>) Results on the Half-Cell (Si/CNTs) dataset. (<b>i</b>–<b>l</b>) Results on the Half-Cell (Si/Graphene) dataset.</p> Full article ">Figure 6
<p>Radar chart comparing the performance of the three optimization methods. (<b>a</b>) Performance on the NASA dataset. (<b>b</b>) Performance on the Half-Cell (Si/CNTs) dataset. (<b>c</b>) Performance on the Half-Cell (Si/Graphene) dataset.</p> Full article ">Figure 7
<p>Box plots comparing the performance of the proposed model and ablation models. (<b>a</b>) Results on the NASA dataset. (<b>b</b>) Results on the Half-Cell (Si/CNTs) dataset. (<b>c</b>) Results on the Half-Cell (Si/Graphene) dataset.</p> Full article ">Figure 8
<p>Box plots comparing the performance of the BO algorithm and the hybrid algorithm (SSA + BO). (<b>a</b>) Results on the NASA dataset. (<b>b</b>) Results on the Half-Cell (Si/CNTs) dataset. (<b>c</b>) Results on the Half-Cell (Si/Graphene) dataset.</p> Full article ">
19 pages, 5395 KiB  
Article
Optimizing 3D Point Cloud Reconstruction Through Integrating Deep Learning and Clustering Models
by Seyyedbehrad Emadi and Marco Limongiello
Electronics 2025, 14(2), 399; https://doi.org/10.3390/electronics14020399 - 20 Jan 2025
Viewed by 395
Abstract
Noise in 3D photogrammetric point clouds—both close-range and UAV-generated—poses a significant challenge to the accuracy and usability of digital models. This study presents a novel deep learning-based approach to improve the quality of point clouds by addressing this issue. We propose a two-step [...] Read more.
Noise in 3D photogrammetric point clouds—both close-range and UAV-generated—poses a significant challenge to the accuracy and usability of digital models. This study presents a novel deep learning-based approach to improve the quality of point clouds by addressing this issue. We propose a two-step methodology: first, a variational autoencoder reduces features, followed by clustering models to assess and mitigate noise in the point clouds. This study evaluates four clustering methods—k-means, agglomerative clustering, Spectral clustering, and Gaussian mixture model—based on photogrammetric parameters, reprojection error, projection accuracy, angles of intersection, distance, and the number of cameras used in tie point calculations. The approach is validated using point cloud data from the Temple of Neptune in Paestum, Italy. The results show that the proposed method significantly improves 3D reconstruction quality, with k-means outperforming other clustering techniques based on three evaluation metrics. This method offers superior versatility and performance compared to traditional and machine learning techniques, demonstrating its potential to enhance UAV-based surveying and inspection practices. Full article
(This article belongs to the Special Issue Point Cloud Data Processing and Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>Illustration of the reprojection error.</p> Full article ">Figure 2
<p>Illustration of the angle of intersection.</p> Full article ">Figure 3
<p>Conceptual illustration of the proposed methodology, highlighting the main steps involved in optimizing point cloud data using deep learning clustering models.</p> Full article ">Figure 4
<p>Application example: Temple of Neptune in Paestum (Italy).</p> Full article ">Figure 5
<p>A selected section of the Temple of Neptune.</p> Full article ">Figure 6
<p>Visualizations of point cloud data under different single-parameter noise reduction analyses: (<b>a</b>) reprojection errors, (<b>b</b>) average intersection angles, (<b>c</b>) number of images, and (<b>d</b>) projection accuracy.</p> Full article ">Figure 7
<p>Distribution of clusters generated by: (<b>a</b>) GMM clustering algorithms, (<b>b</b>) k-means clustering algorithms, (<b>c</b>) agglomerative clustering algorithms, and (<b>d</b>) Spectral clustering algorithms.</p> Full article ">
16 pages, 12452 KiB  
Article
Scaling Nuclear Magnetic Resonance with Integrated Planar Coil and Transceiver Front-End: Co-Design Considerations
by Natachai Terawatsakul, Alireza Saberkari, Yuttapoom Puttisong and Morgan Madec
Electronics 2025, 14(2), 398; https://doi.org/10.3390/electronics14020398 - 20 Jan 2025
Viewed by 382
Abstract
A comprehensive framework for designing a micro-nuclear magnetic resonance (NMR) front-end is presented. Key radio frequency (RF) engineering principles are established to enable efficient excitation and detection of NMR signals. This foundation aims to guide the optimal design of novel handheld NMR devices [...] Read more.
A comprehensive framework for designing a micro-nuclear magnetic resonance (NMR) front-end is presented. Key radio frequency (RF) engineering principles are established to enable efficient excitation and detection of NMR signals. This foundation aims to guide the optimal design of novel handheld NMR devices operating with magnetic fields (B0) below 0.5 Tesla and RF frequencies under 30 MHz. To address the complexities of signal-to-noise ratio optimization in this regime, a specialized metric called the coil performance factor (CPF) is introduced, emphasizing the role of coil design. Through systematic optimization under realistic constraints, an optimal coil configuration maximizing the CPF is identified. This design, with three turns, a coil width of 0.22 mm, and a coil spacing of 0.15 mm, achieves an optimal balance between magnetic field strength, homogeneity, and noise. This work serves as a valuable resource for engineers developing optimized coil designs and RF solutions for handheld NMR devices, providing clear explanations of essential concepts and a practical design methodology. Full article
(This article belongs to the Special Issue RF/MM-Wave Circuits Design and Applications, 2nd Edition)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Nuclear spin sublevel of hydrogen (<sup>1</sup>H) subjected into the static magnetic field <math display="inline"><semantics> <msub> <mi>B</mi> <mn>0</mn> </msub> </semantics></math>. (<b>b</b>) Classical vectorized picture of net-magnetization along the <math display="inline"><semantics> <msub> <mi>B</mi> <mn>0</mn> </msub> </semantics></math> (z-axis) direction.</p> Full article ">Figure 2
<p>(<b>a</b>) FID detection via NMR coils. (<b>b</b>) FID signal in the xy plane.</p> Full article ">Figure 3
<p>Sinusoidal <math display="inline"><semantics> <msub> <mi>B</mi> <mn>1</mn> </msub> </semantics></math> decomposed into two rotating fields.</p> Full article ">Figure 4
<p>(<b>a</b>) Rotating frame with a static field <math display="inline"><semantics> <msubsup> <mi>B</mi> <mn>1</mn> <mo>+</mo> </msubsup> </semantics></math>; (<b>b</b>) NMR vector model.</p> Full article ">Figure 5
<p>Impact of on- and off-resonance on the flipping angle.</p> Full article ">Figure 6
<p>Pulse bandwidth.</p> Full article ">Figure 7
<p>Frequency planning of (<b>a</b>) single- and (<b>b</b>) dual-synthesizer TRX.</p> Full article ">Figure 8
<p>Handheld NMR RX’s scenario.</p> Full article ">Figure 9
<p>(<b>a</b>) Simulated magnetic field within the sample volume. (<b>b</b>) Temperature distribution on the coil.</p> Full article ">Figure 10
<p>Optimization flowchart.</p> Full article ">Figure 11
<p>Three-turn coil’s (<b>a</b>) <math display="inline"><semantics> <msub> <mi>B</mi> <mn>1</mn> </msub> </semantics></math>, (<b>b</b>) <span class="html-italic">K</span>, (<b>c</b>) <math display="inline"><semantics> <msub> <mi>R</mi> <mi>c</mi> </msub> </semantics></math> as a function of <math display="inline"><semantics> <msub> <mi mathvariant="normal">W</mi> <mi>coil</mi> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi mathvariant="normal">S</mi> <mi>coil</mi> </msub> </semantics></math>.</p> Full article ">Figure 12
<p>Three-turn coil’s CPF as a function of <math display="inline"><semantics> <msub> <mi mathvariant="normal">W</mi> <mi>coil</mi> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi mathvariant="normal">S</mi> <mi>coil</mi> </msub> </semantics></math>.</p> Full article ">Figure 13
<p>Three-turn coil’s optimal <math display="inline"><semantics> <msub> <mi mathvariant="normal">W</mi> <mi>coil</mi> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi mathvariant="normal">S</mi> <mi>coil</mi> </msub> </semantics></math> as a function of RX noise.</p> Full article ">
24 pages, 7528 KiB  
Article
EOS: Edge-Based Operation Skip Scheme for Real-Time Object Detection Using Viola-Jones Classifier
by Cheol-Ho Choi, Joonhwan Han, Hyun Woo Oh, Jeongwoo Cha and Jungho Shin
Electronics 2025, 14(2), 397; https://doi.org/10.3390/electronics14020397 - 20 Jan 2025
Viewed by 381
Abstract
Machine learning-based object detection systems are preferred due to their cost-effectiveness compared to deep learning approaches. Among machine learning methods, the Viola-Jones classifier stands out for its reasonable accuracy and efficient resource utilization. However, as the number of classification iterations increases or the [...] Read more.
Machine learning-based object detection systems are preferred due to their cost-effectiveness compared to deep learning approaches. Among machine learning methods, the Viola-Jones classifier stands out for its reasonable accuracy and efficient resource utilization. However, as the number of classification iterations increases or the resolution of the input image increases, the detection processing speed may decrease. To address the detection speed issue related to input image resolution, an improved edge component calibration method is applied. Additionally, an edge-based operation skip scheme is proposed to overcome the detection processing speed problem caused by the number of classification iterations. Our experiments using the FDDB public dataset show that our method reduces classification iterations by 24.6157% to 84.1288% compared to conventional methods, except for our previous study. Importantly, our method maintains detection accuracy while reducing classification iterations. This result implies that our method can realize almost real-time object detection when implemented on field-programmable gate arrays. Full article
(This article belongs to the Special Issue Artificial Intelligence in Image and Video Processing)
►▼ Show Figures

Figure 1

Figure 1
<p>Analyzing machine learning and deep learning approaches in artificial intelligence research and development: A comparative perspective (Red box: Detected result).</p> Full article ">Figure 2
<p>Process overview: Integrating the proposed edge-based skip scheme with the Viola-Jones classifier algorithm.</p> Full article ">Figure 3
<p>Reference coordinate selection approach for the proposed edge-based operation skip scheme: (<b>a</b>) window for edge component-calibrated image and (<b>b</b>) window for merged edge component image.</p> Full article ">Figure 4
<p>Experimental result using the proposed and conventional methods for ‘Lena’ test frame: (<b>a</b>) Viola-Jones classifier [<a href="#B11-electronics-14-00397" class="html-bibr">11</a>,<a href="#B12-electronics-14-00397" class="html-bibr">12</a>], (<b>b</b>) Hyun [<a href="#B20-electronics-14-00397" class="html-bibr">20</a>], (<b>c</b>) Choi [<a href="#B14-electronics-14-00397" class="html-bibr">14</a>], (<b>d</b>) our previous work [<a href="#B15-electronics-14-00397" class="html-bibr">15</a>], and (<b>e</b>) the proposed method (Red box: Detected result).</p> Full article ">Figure 5
<p>Experimental result using the proposed and conventional methods for ‘Solvay Conference 1927’ test frame: (<b>a</b>) Viola-Jones classifier [<a href="#B11-electronics-14-00397" class="html-bibr">11</a>,<a href="#B12-electronics-14-00397" class="html-bibr">12</a>], (<b>b</b>) Hyun [<a href="#B20-electronics-14-00397" class="html-bibr">20</a>], (<b>c</b>) Choi [<a href="#B14-electronics-14-00397" class="html-bibr">14</a>], (<b>d</b>) our previous work [<a href="#B15-electronics-14-00397" class="html-bibr">15</a>], and (<b>e</b>) proposed method (Red box: Detected result).</p> Full article ">Figure 6
<p>Experimental result using the proposed and conventional methods for FDDB public dataset with various IoU threshold values: (<b>a</b>) precision, (<b>b</b>) recall, and (<b>c</b>) <math display="inline"><semantics> <msub> <mi>F</mi> <mn>1</mn> </msub> </semantics></math> score.</p> Full article ">Figure 7
<p>Mean performance degradation of proposed and conventional methods using the FDDB public dataset at various IoU threshold values: (<b>a</b>) precision, (<b>b</b>) recall, and (<b>c</b>) <math display="inline"><semantics> <msub> <mi>F</mi> <mn>1</mn> </msub> </semantics></math> score.</p> Full article ">Figure 8
<p>Experimental results using the proposed method with an operating frequency of 30 frames per second: (<b>a</b>) Lena, (<b>b</b>) Solvay conference 1927, and (<b>c</b>) FDDB public dataset.</p> Full article ">Figure 9
<p>Experimental results using the proposed method with an operating frequency of 60 frames per second: (<b>a</b>) Lena, (<b>b</b>) Solvay conference 1927, and (<b>c</b>) FDDB public dataset.</p> Full article ">
16 pages, 1161 KiB  
Article
Multiplex Graph Contrastive Learning with Soft Negatives
by Zhenhao Zhao, Minhong Zhu, Chen Wang, Sijia Wang, Jiqiang Zhang, Li Chen and Weiran Cai
Electronics 2025, 14(2), 396; https://doi.org/10.3390/electronics14020396 - 20 Jan 2025
Viewed by 282
Abstract
Graph Contrastive Learning (GCL) seeks to learn nodal or graph representations that contain maximal consistent information from graph-structured data. While node-level contrasting modes are dominating, some efforts have commenced to explore consistency across different scales. Yet, they tend to lose consistent information and [...] Read more.
Graph Contrastive Learning (GCL) seeks to learn nodal or graph representations that contain maximal consistent information from graph-structured data. While node-level contrasting modes are dominating, some efforts have commenced to explore consistency across different scales. Yet, they tend to lose consistent information and be contaminated by disturbing features. We propose MUX-GCL, a novel cross-scale contrastive learning framework that addresses these key challenges in GCL by leveraging multiplex representations as effective patches to enhance information consistency. Our method introduces a soft-negative contrasting strategy based on positional affinities to reduce false negatives, thereby minimizing information loss during multi-scale contrasts. While this learning mode minimizes contaminating noises, a commensurate contrasting strategy using positional affinities further avoids information loss by correcting false negative pairs across scales. Extensive downstream experiments demonstrate that MUX-GCL yields multiple state-of-the-art results on public datasets. Our theoretical analysis further guarantees the new objective function as a stricter lower bound of mutual information of raw input features and output embeddings, which rationalizes this paradigm. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Similarity distributions of cross-layer embeddings between two augmented views (for GRACE). <span class="html-italic">u</span> and <span class="html-italic">v</span> denote two augmented views derived from the original graph. The vertical axis represents the node similarity calculated using the dot product. The vertical axis represents the probability density function values estimated using a Gaussian kernel. All positive pairs are substantially more similar than negative pairs, labeled as <math display="inline"><semantics> <mrow> <msub> <mi>u</mi> <mi>m</mi> </msub> <msub> <mi>v</mi> <mi>n</mi> </msub> <mspace width="3.33333pt"/> <mi>p</mi> <mi>o</mi> <mi>s</mi> <mo>/</mo> <mi>n</mi> <mi>e</mi> <mi>g</mi> </mrow> </semantics></math> with <span class="html-italic">m</span> and <span class="html-italic">n</span> numbering the layers.</p> Full article ">Figure 2
<p>Overall architecture of MUX-GCL. Contrasts are executed between “effective patches” constructed from all representations of the multiplex encoder as illustrated by the links. The pairwise affinities of topological embedding estimate the likelihood of being false negatives. Augmentations are implemented as in GRACE. Positive and negative pairs are labeled in the figure.</p> Full article ">Figure 3
<p>Hyper-parameter <math display="inline"><semantics> <mi>λ</mi> </semantics></math> analysis (Acc (%) for 3 seeds) on Cora and Photo. The combinations of <math display="inline"><semantics> <mi>λ</mi> </semantics></math> that cannot be taken among them are set to the average.</p> Full article ">Figure A1
<p><math display="inline"><semantics> <msubsup> <mi>T</mi> <mrow> <mi>D</mi> <mo>,</mo> <mi>i</mi> <mi>j</mi> </mrow> <mrow> <mi>L</mi> <mi>k</mi> </mrow> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>T</mi> <mrow> <mi>S</mi> <mo>,</mo> <mi>i</mi> <mi>j</mi> </mrow> <mrow> <mi>L</mi> <mi>k</mi> </mrow> </msubsup> </semantics></math> values distribution during training process of Cora. We take out the embeddings of our encoder trained at epoch 10 and 300, respectively, to calculate <math display="inline"><semantics> <msubsup> <mi>T</mi> <mrow> <mi>D</mi> <mo>,</mo> <mi>i</mi> <mi>j</mi> </mrow> <mrow> <mi>L</mi> <mi>k</mi> </mrow> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>T</mi> <mrow> <mi>S</mi> <mo>,</mo> <mi>i</mi> <mi>j</mi> </mrow> <mrow> <mi>L</mi> <mi>k</mi> </mrow> </msubsup> </semantics></math>, respectively. Gaussian curves are fitted to the values of <math display="inline"><semantics> <msubsup> <mi>T</mi> <mrow> <mi>D</mi> <mo>,</mo> <mi>i</mi> <mi>j</mi> </mrow> <mrow> <mi>L</mi> <mi>k</mi> </mrow> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>T</mi> <mrow> <mi>S</mi> <mo>,</mo> <mi>i</mi> <mi>j</mi> </mrow> <mrow> <mi>L</mi> <mi>k</mi> </mrow> </msubsup> </semantics></math> at epochs 10 and 300, respectively.</p> Full article ">
26 pages, 23657 KiB  
Article
A Digital Twin Approach for Soil Moisture Measurement with Physically Based Rendering Simulations and Machine Learning
by Ismail Parewai and Mario Köppen
Electronics 2025, 14(2), 395; https://doi.org/10.3390/electronics14020395 - 20 Jan 2025
Viewed by 427
Abstract
Soil is one of the most important factors of agricultural productivity, directly influencing crop growth, water management, and overall yield. However, inefficient soil moisture monitoring methods, such as manual observation and gravimetric in rural areas, often lead to overwatering or underwatering, wasting resources [...] Read more.
Soil is one of the most important factors of agricultural productivity, directly influencing crop growth, water management, and overall yield. However, inefficient soil moisture monitoring methods, such as manual observation and gravimetric in rural areas, often lead to overwatering or underwatering, wasting resources and reduced yields, and harming soil health. This study offers a digital twin approach for soil moisture measurement, integrating real-time physical data, virtual simulations, and machine learning to classify soil moisture conditions. The digital twin is proposed as a virtual representation of physical soil designed to replicate real-world behavior. We used a multispectral rotocam, and high-resolution soil images were captured under controlled conditions. Physically based rendering (PBR) materials were created from these data and implemented in a game engine to simulate soil properties accurately. Image processing techniques were applied to extract key features, followed by machine learning algorithms to classify soil moisture levels (wet, normal, dry). Our results demonstrate that the Soil Digital Twin replicates real-world behavior, with the Random Forest model achieving a high classification accuracy of 96.66% compared to actual soil. This data-driven approach conveys the potential of the Soil Digital Twin to enhance precision farming initiatives and water use efficiency for sustainable agriculture. Full article
(This article belongs to the Special Issue Pattern Recognition and Image Processing: Latest Advances and Prospects)
►▼ Show Figures

Figure 1

Figure 1
<p>The soil digital twin development scheme.</p> Full article ">Figure 2
<p>Experimental setup for real soil: (<b>Left</b>) Indoor setup for soil moisture analysis with LED lighting and real-time data display. (<b>Right</b>) Low-light indoor setup capturing soil properties with LED lighting for digital twin modeling.</p> Full article ">Figure 3
<p>Experimental setup in Unreal Engine for soil texture and lighting analysis: (<b>Left</b>) daylight scene with camera placement and spotlight direction shown by arrows; (<b>Right</b>) simulation of blue LED lighting effect on soil texture.</p> Full article ">Figure 4
<p>Comparison of real soil images (top row) and digital twin images (bottom row) across different color channels: (<b>a</b>,<b>e</b>) blue; (<b>b</b>,<b>f</b>) green; (<b>c</b>,<b>g</b>) red; (<b>d</b>,<b>h</b>) yellow.</p> Full article ">Figure 5
<p>Accuracy comparison across soil types.</p> Full article ">Figure 6
<p>Confusion matrix results for Artificial Neural Networks (ANNs): real soil (top row) vs. digital twin (bottom row).</p> Full article ">Figure 7
<p>Confusion matrix results for Random Forest (RF): real soil (top row) vs. digital twin (bottom row).</p> Full article ">Figure 8
<p>Confusion matrix results for Support Vector Machine (SVM): real soil (top row) vs. digital twin (bottom row).</p> Full article ">
24 pages, 5413 KiB  
Systematic Review
Internet of Things Ontologies for Well-Being, Aging and Health: A Scoping Literature Review
by Hrvoje Belani, Petar Šolić, Eftim Zdravevski and Vladimir Trajkovik
Electronics 2025, 14(2), 394; https://doi.org/10.3390/electronics14020394 - 20 Jan 2025
Viewed by 497
Abstract
Internet of Things aims to simplify and automate complicated tasks by using sensors and other inputs for collecting huge amounts of data, processing them in the cloud and on the edge networks, and allowing decision making toward further interactions via actuators and other [...] Read more.
Internet of Things aims to simplify and automate complicated tasks by using sensors and other inputs for collecting huge amounts of data, processing them in the cloud and on the edge networks, and allowing decision making toward further interactions via actuators and other outputs. As connected IoT devices rank in billions, semantic interoperability remains one of the permanent challenges, where ontologies can provide a great contribution. The main goal of this paper is to analyze the state of research on semantic interoperability in well-being, aging, and health IoT services by using ontologies. This was achieved by analyzing the following research questions: “Which IoT ontologies have been used to implement well-being, aging and health services?” and “What is the dominant approach to achieve semantic interoperability of IoT solutions for well-being, aging and health?’ We conducted a scoping literature review of research papers from 2013 to 2024 by applying the PRISMA-ScR meta-analysis methodology with a custom-built software tool for an exhaustive search through the following digital libraries: IEEE Xplore, PubMed, MDPI, Elsevier ScienceDirect, and Springer Nature Link. By thoroughly analyzing 30 studies from an initial pool of more than 80,000 studies, we conclude that IoT ontologies for well-being, aging, and health services increasingly adopt Semantic Web of Things standards to achieve semantic interoperability by integrating heterogeneous data through unified semantic models. Emerging approaches, like semantic communication, Large Language Models Edge Intelligence, and sustainability-driven IoT analytics, can further enhance service efficiency and promote a holistic “One Well-Being, Aging, and Health” framework. Full article
(This article belongs to the Special Issue Internet of Things for E-health)
►▼ Show Figures

Figure 1

Figure 1
<p>Search strategy used for the selection of particular publishers in the scoping review.</p> Full article ">Figure 2
<p>Heatmap about how many articles are there in the initially included article set containing a pair of properties. The value of each pair is represented using a particular shade of orange—the darker the orange, the higher the value.</p> Full article ">Figure 3
<p>The distribution of 30 resulting articles by the publication year and by the property groups.</p> Full article ">Figure 4
<p>The distribution of 30 resulting articles by the publication year and by the digital library that contained the articles.</p> Full article ">Figure 5
<p>The distribution of 30 resulting articles by the publication year and by the keyword used in the scoping literature review.</p> Full article ">Figure 6
<p>The keyword co-occurrence within the 30 selected articles.</p> Full article ">Figure 7
<p>The graph of keyword co-occurrence within the 30 selected articles.</p> Full article ">Figure 8
<p>(<b>a</b>) The keyword occurrence ratio within the 30 selected articles; (<b>b</b>) the technical challenges ratio within the 30 selected articles.</p> Full article ">
22 pages, 4095 KiB  
Article
A Reliable Routing Algorithm Based on Path Satisfaction in the Energy Internet
by Baoju Liu, Xiangqian Wei, Haifeng Hu, Peng Yu and Lei Shi
Electronics 2025, 14(2), 393; https://doi.org/10.3390/electronics14020393 - 20 Jan 2025
Viewed by 395
Abstract
To meet the ever-increasing strict transmission requirements of services in the Energy Internet (EI), reliable routing algorithms for service are necessary. Most of the existing routing algorithms in the Internet Protocol (IP) layer concentrate on service requirements and network topology features while neglecting [...] Read more.
To meet the ever-increasing strict transmission requirements of services in the Energy Internet (EI), reliable routing algorithms for service are necessary. Most of the existing routing algorithms in the Internet Protocol (IP) layer concentrate on service requirements and network topology features while neglecting spectrum resource utilization in the optical transport layer. The status of spectrum resources in the optical transport layer also affects the availability of the routing path. However, there are few studies that combined service transmission requirements and network structure with spectrum resources of the link. In light of this, it is more practical to design routing algorithms integrated with the IP layer and the optical layer. There are three main innovations as follows: (1) The indicator of path satisfaction is proposed meanwhile the system model and service model are constructed. (2) Searching routing paths for services is abstracted into a constrained optimization problem. The optimal objective is to maximize path satisfaction. At the same time, various service requirements, such as end-to-end latency and bandwidth, should be satisfied. (3) To reduce computational complexity, a heuristic path satisfaction-based service-aware routing algorithm (PSSRA) is designed to resolve it. Extensive experiments are carried out with varied service requests on different network topologies. The final results demonstrate that the proposed algorithm outperforms the existing algorithms regarding the service blocking ratio and service distribution fairness index. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Architecture diagram for the Energy Internet. It describes a cross-over architecture involving an IP layer and a transmissions layer, which covers three layers of the communication network from the prospect of Software Defined Networking.</p> Full article ">Figure 2
<p>Schematic diagram of the SRLG. This network topology has seven nodes, and it is a part of backbone power cable network in the Jiangsu Province in China.</p> Full article ">Figure 3
<p>The schematic diagram of the PSSRA.</p> Full article ">Figure 4
<p>Sequence diagram of the PSSRA.</p> Full article ">Figure 5
<p>14-node NSFnet.</p> Full article ">Figure 6
<p>28-node network.</p> Full article ">Figure 7
<p>The SBR with the PSSRA at different levels in the 7-node network.</p> Full article ">Figure 8
<p>JFI variation under different approaches in the 7-node network. The oblique line denotes the result of our approach, the rhomboid form is the results of SPA, and the green denotes the result of the SCRA.</p> Full article ">Figure 9
<p>The SBR of the PSSRA with different service levels in the 14-node network. It can be found that services with higher levels have lower SBRs, and vice versa.</p> Full article ">Figure 10
<p>JFI variation under the different approaches in the 14-node network.</p> Full article ">Figure 11
<p>Computation time in the 14-node network. To test the algorithm performance, we conducted and compared it with other algorithms in this network topology.</p> Full article ">
21 pages, 1171 KiB  
Article
Statistical Analysis of the Sum of Double Random Variables for Security Applications in RIS-Assisted NOMA Networks with a Direct Link
by Sang-Quang Nguyen, Phuong T. Tran, Bui Vu Minh, Tran Trung Duy, Anh-Tu Le, Lubos Rejfek and Lam-Thanh Tu
Electronics 2025, 14(2), 392; https://doi.org/10.3390/electronics14020392 - 20 Jan 2025
Viewed by 516
Abstract
Next- generation wireless communications are projected to integrate reconfigurable intelligent surfaces (RISs) to perpetrate enhanced spectral and energy efficiencies. To quantify the performance of RIS-aided wireless networks, the statistics of a single random variable plus the sum of double random variables becomes a [...] Read more.
Next- generation wireless communications are projected to integrate reconfigurable intelligent surfaces (RISs) to perpetrate enhanced spectral and energy efficiencies. To quantify the performance of RIS-aided wireless networks, the statistics of a single random variable plus the sum of double random variables becomes a core approach to reflect how communication links from RISs improve wireless-based systems versus direct ones. With this in mind, the work applies the statistics of a single random variable plus the sum of double random variables in the secure performance of RIS-based non-orthogonal multi-access (NOMA) systems with the presence of untrusted users. We propose a new communication strategy by jointly considering NOMA encoding and RIS’s phase shift design to enhance the communication of legitimate nodes while degrading the channel capacity of untrusted elements but with sufficient power resources for signal recovery. Following that, we analyze and derive the closed-form expressions of the secrecy effective capacity (SEC) and secrecy outage probability (SOP). All analyses are supported by extensive Monte Carlo simulation outcomes, which facilitate an understanding of system communication behavior, such as the transmit signal-to-noise ratio, the number of RIS elements, the power allocation coefficients, the target data rate of the communication channels, and secure data rate. Finally, the results demonstrate that our proposed communication can be improved significantly with an increase in the number of RIS elements, irrespective of the presence of untrusted proximate or distant users. Full article
(This article belongs to the Special Issue Wireless Sensor Network: Latest Advances and Prospects)
►▼ Show Figures

Figure 1

Figure 1
<p>PDF and CDF validation under different settings of RIS elements. (<b>a</b>,<b>c</b>,<b>e</b>) illustrate the CDF of <math display="inline"><semantics> <mrow> <mrow> <mo stretchy="false">|</mo> </mrow> <msub> <mi>τ</mi> <mi mathvariant="sans-serif">B</mi> </msub> <msup> <mrow> <mo stretchy="false">|</mo> </mrow> <mn>2</mn> </msup> </mrow> </semantics></math> and (<b>b</b>,<b>d</b>,<b>f</b>) show the PDF of <math display="inline"><semantics> <mrow> <mrow> <mo stretchy="false">|</mo> </mrow> <msub> <mi>τ</mi> <mi mathvariant="sans-serif">W</mi> </msub> <msup> <mrow> <mo stretchy="false">|</mo> </mrow> <mn>2</mn> </msup> </mrow> </semantics></math> under different settings of RIS elements and channel distributions.</p> Full article ">Figure 2
<p>SOP performance against transmit SNR and power allocation. (<b>a</b>) shows the SOP with respect to (w.r.t.) the SNR <math display="inline"><semantics> <mover> <mi>γ</mi> <mo>¯</mo> </mover> </semantics></math> and (<b>b</b>) illustrates the SOP w.r.t. power allocation coefficient <math display="inline"><semantics> <msub> <mi>ρ</mi> <mi mathvariant="sans-serif">W</mi> </msub> </semantics></math>.</p> Full article ">Figure 3
<p>SOP performance against target rate transmission. (<b>a</b>) shows the SOP w.r.t. the Willie’s target data rates and (<b>b</b>) illustrates the SOP regarding the Bob’s secure rate <math display="inline"><semantics> <msub> <mi>R</mi> <mi mathvariant="sans-serif">B</mi> </msub> </semantics></math>.</p> Full article ">Figure 4
<p>SEC performance against transmit SNR and power allocation. (<b>a</b>) shows the SEC with respect to (w.r.t.) the SNR <math display="inline"><semantics> <mover> <mi>γ</mi> <mo>¯</mo> </mover> </semantics></math> and (<b>b</b>) illustrates the SEC w.r.t. power allocation coefficient <math display="inline"><semantics> <msub> <mi>ρ</mi> <mi mathvariant="sans-serif">W</mi> </msub> </semantics></math>.</p> Full article ">
15 pages, 4180 KiB  
Article
Evaluation Method and Modeling Analysis of the Common Mode Noise Suppression Capability of Full-Bridge Transformers
by Yipeng Kong and Wei Chen
Electronics 2025, 14(2), 391; https://doi.org/10.3390/electronics14020391 - 20 Jan 2025
Viewed by 337
Abstract
The effective capacitance of the common mode port serves as a critical metric for assessing the common mode noise suppression capability of transformers in power converters. Conventionally, the evaluation of transformers in single-ended topologies, such as flyback converters, using a network analyzer necessitates [...] Read more.
The effective capacitance of the common mode port serves as a critical metric for assessing the common mode noise suppression capability of transformers in power converters. Conventionally, the evaluation of transformers in single-ended topologies, such as flyback converters, using a network analyzer necessitates a reference static point and a dynamic point at the transformer port. However, a full-bridge transformer without a center tap lacks a reference static point in both the primary and secondary stages. Consequently, this paper proposes an innovative measurement technique to evaluate the common mode noise suppression capability of full-bridge transformers. This method accounts for the intrinsic parameters of the transformer and refines the high-frequency equivalent circuit model for accurate measurement. Ultimately, the validity of the proposed model is confirmed through experiments conducted on a CLLC converter prototype, offering the industry a straightforward and efficient approach to assessing and testing the common mode noise suppression performance of transformers without a center tap. Full article
(This article belongs to the Section Power Electronics)
►▼ Show Figures

Figure 1

Figure 1
<p>Mechanism of common mode noise transmission in transformers.</p> Full article ">Figure 2
<p>Schematic diagram of impedance analyzer measurement.</p> Full article ">Figure 3
<p>Equivalent circuit diagram of common mode port effective capacitance: (<b>a</b>) potential distribution diagram of the transformer and (<b>b</b>) effective capacitance equivalent diagram.</p> Full article ">Figure 4
<p>Transformer Insertion Loss Measurement Diagram and Its Equivalent Circuit.</p> Full article ">Figure 5
<p>Measurement method based on the RF transformer.</p> Full article ">Figure 6
<p>Analysis of coupling capability in single-sided winding of the RF transformer: (<b>a</b>) single-side winding insertion loss test and (<b>b</b>) test schematics.</p> Full article ">Figure 7
<p>Comparison of S21 curves with different inductance measurements.</p> Full article ">Figure 8
<p>Measurement of common-mode effective capacitance in full-bridge transformer ports (<b>a</b>) with a center tap and (<b>b</b>) without a center tap.</p> Full article ">Figure 9
<p>Simplified equivalent circuit model.</p> Full article ">Figure 10
<p>Measured test object.</p> Full article ">Figure 11
<p>Insertion loss calculation and measurement for different schemes (<b>a</b>) comparison of test results of Scheme A and (<b>b</b>) comparison of test results of Scheme B.</p> Full article ">Figure 12
<p>Comparison of actual measurements for two solutions.</p> Full article ">Figure 13
<p>Principle diagram of the conducted electromagnetic interference test.</p> Full article ">Figure 14
<p>Noise testing platform.</p> Full article ">Figure 15
<p>Comparison of noise results from two methods.</p> Full article ">
11 pages, 4700 KiB  
Article
Optimized Design and Experimental Study of an Axis-Encircling Beam with Gently Varying Cusp Magnetic Field
by Chaojun Lei, E’Feng Wang, Qixiang Zhao, Shaoliang Shi, Dongshuo Gao, Shufeng Li, Yichi Zhang and Jinjun Feng
Electronics 2025, 14(2), 390; https://doi.org/10.3390/electronics14020390 - 20 Jan 2025
Viewed by 362
Abstract
Mode competition is a significant barrier to advancing gyrotrons towards high frequency, high power, and high efficiency. An axis-encircling beam enables gyrotrons to achieve high interaction efficiency while maintaining stable operation at higher-order harmonics. However, generating a high-quality axis-encircling beam requires an ideal [...] Read more.
Mode competition is a significant barrier to advancing gyrotrons towards high frequency, high power, and high efficiency. An axis-encircling beam enables gyrotrons to achieve high interaction efficiency while maintaining stable operation at higher-order harmonics. However, generating a high-quality axis-encircling beam requires an ideal cusp magnetic field, which is challenging to achieve experimentally. This paper discusses the optimization design of an axis-encircling beam with a gently varying cusp magnetic field. A non-ideal cusp magnetic field is designed using the existing magnetic field and power supply in the laboratory. Under this magnetic field, a large-orbit electronic optical system with 20 kV, 0.5 A, an axis-encircling radius of 3.3 mm at a guiding magnetic field of 0.122 T, and a velocity spread (both transverse and longitudinal) of less than 1.2% was obtained and tested. Full article
(This article belongs to the Section Power Electronics)
►▼ Show Figures

Figure 1

Figure 1
<p>The major component of a gyro-TWT (1—cusp gun; 2—reverse coil; 3—cavity coil; 4—collector coil; 5—input window; 6—high-frequency structure; 7—mode converter; 8—collector; 9—gently varying magnetic field; 10—output window).</p> Full article ">Figure 2
<p>The distribution of an axial magnetic field.</p> Full article ">Figure 3
<p>The distribution of the electric potential.</p> Full article ">Figure 4
<p>The structure of electrodes and the trajectory of the axis-encircling beam.</p> Full article ">Figure 5
<p>The distribution of the longitudinal momentums of the beam along the z-axis.</p> Full article ">Figure 6
<p>The distribution of the transverse momentums of the beam along the z-axis.</p> Full article ">Figure 7
<p>The distribution of the pitch factor of the beam along the z-axis.</p> Full article ">Figure 8
<p>The velocity spread of the beam vs. the cathode magnetic field.</p> Full article ">Figure 9
<p>The central spread of the beam changes with the distribution of magnetic field near the cathode.</p> Full article ">Figure 10
<p>The central spread of the beam changes with the magnetic field near the cathode.</p> Full article ">Figure 11
<p>The structure and the trajectory of the small orbit beam.</p> Full article ">Figure 12
<p>The photo of the MIG. (1—output window; 2—concentric mold of magnetic field and electron beam; 3—insulating ceramic; 4—titanium getter pump).</p> Full article ">Figure 13
<p>(<b>a</b>) The orbit of the axis-encircling beam at the exit of the MIG in the experiment; and (<b>b</b>) the small orbit beam at the exit of the MIG in the experiment.</p> Full article ">Figure 14
<p>Voltage and current waveform in the test. (1—waveform of the beam voltage; 2—waveform of the beam current).</p> Full article ">
39 pages, 18582 KiB  
Review
Recent Developments in Bidirectional DC-DC Converter Topologies, Control Strategies, and Applications in Photovoltaic Power Generation Systems: A Comparative Review and Analysis
by Ayiguzhali Tuluhong, Zhisen Xu, Qingpu Chang and Tengfei Song
Electronics 2025, 14(2), 389; https://doi.org/10.3390/electronics14020389 - 20 Jan 2025
Viewed by 453
Abstract
As an important piece of equipment in photovoltaic power generation systems, the bidirectional DC-DC converter plays a vital role in improving the conversion efficiency of photovoltaic power generation system. The energy transfer in PV systems heavily relies on efficient bidirectional DC-DC converters. To [...] Read more.
As an important piece of equipment in photovoltaic power generation systems, the bidirectional DC-DC converter plays a vital role in improving the conversion efficiency of photovoltaic power generation system. The energy transfer in PV systems heavily relies on efficient bidirectional DC-DC converters. To ensure stable operation, converters with high reliability and power density are required. This paper introduces the basic principles and topologies of bidirectional DC-DC converters and provides a comparative analysis. And it examines the characteristics of the converters’ control schemes and switching strategies, summarizes the existing research findings and current issues. Finally, it looks ahead to the future development trends of bidirectional DC-DC converters in PV systems. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>New renewable energy electricity capacity by technology from 2016 to 2028.</p> Full article ">Figure 2
<p>Renewable energy share by technology from 2000 to 2028.</p> Full article ">Figure 3
<p>Overall structure diagram of bidirectional DC-DC converter.</p> Full article ">Figure 4
<p>Classification of DC-DC converters.</p> Full article ">Figure 5
<p>Topology of bidirectional buck/boost converter.</p> Full article ">Figure 6
<p>Topology of bidirectional buck–boost converter.</p> Full article ">Figure 7
<p>Topology of bidirectional Cuk converter.</p> Full article ">Figure 8
<p>Topology of bidirectional Zeta/Sepic converter.</p> Full article ">Figure 9
<p>Topology of basic high-increment DC-DC converter with switched capacitor unit.</p> Full article ">Figure 10
<p>Topology of basic high-gain DC converter with switched inductance unit.</p> Full article ">Figure 11
<p>Topology of the basic high-gain DC converter with switched capacitor and inductance units.</p> Full article ">Figure 12
<p>Basic topology of high-gain DC-DC converter with coupled inductance.</p> Full article ">Figure 13
<p>Topology of cascade converter.</p> Full article ">Figure 14
<p>Topology of interleaved converter.</p> Full article ">Figure 15
<p>Voltage gain and duty cycle curves.</p> Full article ">Figure 16
<p>Basic topology of multilevel converter.</p> Full article ">Figure 17
<p>Topology of bidirectional flyback converter.</p> Full article ">Figure 18
<p>Topology of bidirectional forward converter.</p> Full article ">Figure 19
<p>RCD forward converter structure.</p> Full article ">Figure 20
<p>Structure of active clamp forward converter.</p> Full article ">Figure 21
<p>Resonant magnetic reset forward converter topology.</p> Full article ">Figure 22
<p>Push–pull converter topology.</p> Full article ">Figure 23
<p>Topology of half-bridge converter.</p> Full article ">Figure 24
<p>Topology of full-bridge converter.</p> Full article ">Figure 25
<p>Dual active bridge converter topology.</p> Full article ">Figure 26
<p>Bidirectional LLC resonant converter.</p> Full article ">Figure 27
<p>Bidirectional CLLC resonant converter.</p> Full article ">Figure 28
<p>Bidirectional CLLLC resonant converter.</p> Full article ">Figure 29
<p>Three-active full-bridge DC converter topology.</p> Full article ">Figure 30
<p>Single-phase shift waveform diagram.</p> Full article ">Figure 31
<p>Extended-phase shift waveform diagram.</p> Full article ">Figure 32
<p>Double-phase shift waveform diagram.</p> Full article ">Figure 33
<p>Triple-phase shift waveform diagram.</p> Full article ">Figure 34
<p>Relationship diagram of four modulation methods.</p> Full article ">Figure 35
<p>PID control block diagram.</p> Full article ">Figure 36
<p>Overall structure of sliding mode controller.</p> Full article ">Figure 37
<p>Overall structure of a fuzzy controller.</p> Full article ">Figure 38
<p>Overall structure of model predictive controller.</p> Full article ">
19 pages, 2645 KiB  
Article
Power Grid Faults Diagnosis Based on Improved Synchrosqueezing Wavelet Transform and ConvNeXt-v2 Network
by Zhizhong Liu, Zhuo Zhao, Guangyu Huang, Fei Wang, Peng Wang and Jiayue Liang
Electronics 2025, 14(2), 388; https://doi.org/10.3390/electronics14020388 - 20 Jan 2025
Viewed by 345
Abstract
The increasing demand on electrical power consumption all over the world makes the need for stable and reliable electrical power grids is indispensable. Meanwhile, power grid fault diagnosis based on fault recording data is an important technology to ensure the normal operation of [...] Read more.
The increasing demand on electrical power consumption all over the world makes the need for stable and reliable electrical power grids is indispensable. Meanwhile, power grid fault diagnosis based on fault recording data is an important technology to ensure the normal operation of the power grid. Despite the fact that dozens of studies have been put forward to detect electrical faults, these studies still suffer from several downsides, such as fuzzy characteristics of complex fault samples with small inter-class differences and large intra-class differences in different topology structures of distribution networks. To tackle the above issues, this work proposes a power grid fault diagnosis method based on an improved Synchrosqueezing Wavelet Transform (SWT) and ConvNeXt-v2 network (named PGFDSC). Firstly, PGFDSC extracts fault features from the fault recording data with an improved SWT method, and outputs the vector signal to enhance the instantaneous frequency. Then, PGFDSC inputs the extracted feature vectors into the improved ConvNeXt-v2 network for power grid faults recognition. The improved ConvNeXt-v2 network is a self-supervised learning model with the advantages of fast speed and high accuracy, which can effectively solve the problem of inaccurate judgment caused by the high dimensionality of data samples. Finally, extensive experiments were conducted and the experimental results show that PGFDSC improves the accuracy of fault diagnosis by two percentage points compared to other baseline models. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Schematic diagram of electrical quantity data transmission in the SWT—ConvNeXt model.</p> Full article ">Figure 2
<p>SWT signal decomposition process.</p> Full article ">Figure 3
<p>Schematic diagram of the ConvNeXt-V2 structure.</p> Full article ">Figure 4
<p>Schematic diagram of feature fusion under SimAM.</p> Full article ">Figure 5
<p>Schematic diagram of fault recording diagram.</p> Full article ">Figure 6
<p>Some fault characteristics are shown, where (<b>a</b>) represents the voltage signal of the three-phase short circuit fault recording, (<b>b</b>) represents the current signal of the three-phase short circuit fault recording, (<b>c</b>) represents the voltage signal of the two-phase short circuit fault recording, and (<b>d</b>) represents the current signal of the two-phase short circuit fault recording.</p> Full article ">Figure 7
<p>Model accuracy comparison with varying compressibility coefficients.</p> Full article ">
13 pages, 10090 KiB  
Article
Dual-Band Dual-Circularly Polarized Shared-Aperture Phased Array for S-/C-Band Satellite Communications
by Yuanming Xiao, Lianxing He and Xiaoli Wei
Electronics 2025, 14(2), 387; https://doi.org/10.3390/electronics14020387 - 20 Jan 2025
Viewed by 444
Abstract
In this article, a novel method of achieving a single-layer, dual-band, dual-circularly polarized (CP) shared-aperture phased array antenna with wide beam scanning coverage is presented. The space antenna was designed to provide direct-to-cellular communications services at S-/C-bands with a frequency ratio of 1:1.8. [...] Read more.
In this article, a novel method of achieving a single-layer, dual-band, dual-circularly polarized (CP) shared-aperture phased array antenna with wide beam scanning coverage is presented. The space antenna was designed to provide direct-to-cellular communications services at S-/C-bands with a frequency ratio of 1:1.8. Using novel ceramic substrates with high dielectric constants for antenna miniaturization, the optimum interelement spacing can be ensured in one single layer to meet the large-angle scanning demand. The CP characteristic of the phased array is improved by the sequential rotation technique. A prototype of phased array, which is composed of an 8 × 8 S-band Rx array and a 16 × 16 C-band Tx array, is fabricated to verify this design. The measured results show that the shared-aperture phased array can provide ±50° beam scanning coverage at both the S- and C-bands simultaneously to meet the direct-to-cellular communication demand in low earth orbit (LEO) satellites. Full article
(This article belongs to the Special Issue Antenna Designs for 5G/IoT and Space Applications, 2nd Edition)
►▼ Show Figures

Figure 1

Figure 1
<p>Shared-aperture phased array lattice (<b>a</b>) array lattice; (<b>b</b>) array lattice with traditional PCB substrate. (The yellow ones are C band elements and the red ones are S band elements).</p> Full article ">Figure 2
<p>Geometry of the C-band antenna. Dimensions are R<sub>s</sub> = 9.5, R<sub>p</sub> = 8.2, h = 6, a = 11.2, b = 7.5, u = 1.3. (unit: millimeter).</p> Full article ">Figure 3
<p>Geometry of the S-band antenna. Dimensions are Ws = 32, Wp = 28, h = 6, S1 = 1.7, S2 = 13.5, L1 = 3.8, L2 = 3.7. (unit: millimeter).</p> Full article ">Figure 4
<p>Principles of sequential rotation technique.</p> Full article ">Figure 5
<p>The final layout of the S-/C-band CP antenna subarray.</p> Full article ">Figure 6
<p>Simulated current distributions of (<b>a</b>) the S-band element at 1.975 GHz; (<b>b</b>) the C-band element at 3.55 GHz.</p> Full article ">Figure 7
<p>Simulated electric field distributions of (<b>a</b>) the S-band element at 1.975 GHz; and (<b>b</b>) the C-band element at 3.55 GHz.</p> Full article ">Figure 8
<p>Simulated S<sub>11</sub> and gains (<b>a</b>) the S-band antenna element; and (<b>b</b>) the C-band antenna element.</p> Full article ">Figure 9
<p>The simulated ARs of the S-/C-band CP antenna array.</p> Full article ">Figure 10
<p>Prototype of the fabricated as-proposed antenna in a near-field anechoic chamber.</p> Full article ">Figure 11
<p>Simulated and measured S-parameters of the antenna (<b>a</b>) at S-band; and (<b>b</b>) at C-band.</p> Full article ">Figure 12
<p>Simulated and measured isolations between the S-/C-band.</p> Full article ">Figure 13
<p>Simulated and measured radiation pattern of the S-band array at 1.97 GHz in φ = 0° plane when scanning to different angles.</p> Full article ">Figure 14
<p>Array radiation patterns of the C-band array at 3.55 GHz in φ = 0° plane when scanning to different angles. (<b>a</b>) 64−element array; (<b>b</b>) 128−element array.</p> Full article ">Figure 15
<p>Simulated and measured AR at the different scanning directions (<b>a</b>) at 1.97 GHz; (<b>b</b>) 64-element array at 3.55 GHz.</p> Full article ">
19 pages, 2157 KiB  
Article
Using the Retrieval-Augmented Generation to Improve the Question-Answering System in Human Health Risk Assessment: The Development and Application
by Wenjun Meng, Yuzhe Li, Lili Chen and Zhaomin Dong
Electronics 2025, 14(2), 386; https://doi.org/10.3390/electronics14020386 - 20 Jan 2025
Viewed by 902
Abstract
While large language models (LLMs) are vital for retrieving relevant information from extensive knowledge bases, they always face challenges, including high costs and issues of credibility. Here, we developed a question answering system focused on human health risk using Retrieval-Augmented Generation (RAG). We [...] Read more.
While large language models (LLMs) are vital for retrieving relevant information from extensive knowledge bases, they always face challenges, including high costs and issues of credibility. Here, we developed a question answering system focused on human health risk using Retrieval-Augmented Generation (RAG). We first proposed a framework to generate question–answer pairs, resulting in 300 high-quality pairs across six subfields. Subsequently, we created both a Naive RAG and an Advanced RAG-based Question-Answering (Q&A) system. Performance evaluation of the 300 question–answer pairs in individual research subfields demonstrated that the Advanced RAG outperformed traditional LLMs (including ChatGPT and ChatGLM) and Naive RAG. Finally, we integrated the developed module for a single subfield to launch a multi-knowledge base question answering system. Our study represents a novel application of RAG technology and LLMs to optimize knowledge retrieval methods in human health risk assessment. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>The study framework on establishing a retrieval-augmented generation-based question-answering (Q&amp;A) system.</p> Full article ">Figure 2
<p>The process on the generation of question–answer pairs.</p> Full article ">Figure 3
<p>The flow of the Naive Retrieval-Augmented Generation-based question-answering (Q&amp;A) system.</p> Full article ">Figure 4
<p>The design of the Advanced RAG question-answering system.</p> Full article ">Figure 5
<p>The flowchart of a multi-knowledge base integrated question-answering system. Abbreviations: IR, intent recognition; TD, task distribution.</p> Full article ">
14 pages, 1305 KiB  
Systematic Review
Role of Virtual Reality in Improving Home Cancer Care: A Systematic Literature Review
by Safa Elkefi and Avishek Choudhury
Electronics 2025, 14(2), 385; https://doi.org/10.3390/electronics14020385 - 20 Jan 2025
Viewed by 553
Abstract
Virtual reality (VR) can play an important role in supporting remote care for cancer patients. The purpose of this study is to explore the possible applications of VR in-home cancer care to support patients and healthcare practitioners. We conducted a systematic literature search [...] Read more.
Virtual reality (VR) can play an important role in supporting remote care for cancer patients. The purpose of this study is to explore the possible applications of VR in-home cancer care to support patients and healthcare practitioners. We conducted a systematic literature search following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We searched for peer-reviewed publications in PubMed, Web of Science, and IEEE Xplore for research articles published in the last two decades. After the final list of relevant articles was identified, we adopted an inductive approach to categorize and report our findings. We identified 15 relevant research articles and categorized them into three themes: medical treatment, emotional support, and education and training. Six articles leveraged VR to support medical treatment such as outpatient physical therapy and rehabilitation, and pain management. Five used VR to provide emotional support to patients by uplifting their feelings or providing psychological guidance. Lastly, four leveraged VR for education and training purposes. Overall, all studies reported positive outcomes of VR for home cancer care. Our review advocates for VR integration in in-home cancer care. The findings of this review acknowledge the potential of VR in augmenting medical treatment, providing emotional support to patients, and facilitating the education and training of patients and clinicians. VR in-home cancer care can provide patients with a better quality of care. However, just like any other technological integration, VR can also introduce different challenges, including usability, affordability, and acceptance. Therefore, more studies evaluating VR’s usability and acceptance should be conducted. Additionally, stakeholders such as regulatory bodies, patients, and payors should be involved in ensuring VR’s affordability and developing protocols for its safe use. Full article
(This article belongs to the Special Issue Deep Reinforcement Learning and Its Latest Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>Search terms used in the review.</p> Full article ">Figure 2
<p>PRISMA diagram for the search and selection process.</p> Full article ">Figure 3
<p>Conceptual framework followed in the classification of the selected studies.</p> Full article ">
19 pages, 2156 KiB  
Article
Low-Delay AES Key Expansion Units Based on DDBT Structure
by Xinxing Zheng, Han Yan, Zhiwei Peng and Xiaoqiang Zhang
Electronics 2025, 14(2), 384; https://doi.org/10.3390/electronics14020384 - 19 Jan 2025
Viewed by 431
Abstract
Advanced Encryption Standard (AES) key expansion unit is usually implemented by chain structure with a long critical path length. That makes key expansion unit become the bottleneck of high-speed AES implementations. In this paper, a design method of low-delay AES key expansion unit [...] Read more.
Advanced Encryption Standard (AES) key expansion unit is usually implemented by chain structure with a long critical path length. That makes key expansion unit become the bottleneck of high-speed AES implementations. In this paper, a design method of low-delay AES key expansion unit is proposed. The proposed design method is based on a delay-drive binary tree (DDBT) structure, which has been proven that it has the shortest critical path length. Based on the proposed design method, a low-delay AES encryption key expansion unit and a low-delay AES encryption/decryption unified key expansion unit are designed in this paper. Both hardware complexity analysis and integrated circuit synthesis indicate that our DDBT-structure-based designs can reduce the delay greatly compared to traditional chain structures. Furthermore, compared to previous works, our designs can achieve the largest throughput. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>AES encryption key expansion unit based on the chain structure.</p> Full article ">Figure 2
<p>AES decryption key expansion unit.</p> Full article ">Figure 3
<p>AES encryption/decryption unified key expansion unit based on the chain structure.</p> Full article ">Figure 4
<p>The DDBT structure of input signal <math display="inline"><semantics> <mrow> <msubsup> <mi>K</mi> <mn>3</mn> <mi>i</mi> </msubsup> </mrow> </semantics></math>.</p> Full article ">Figure 5
<p>The AES encryption key expansion unit based on the DDBT structure.</p> Full article ">Figure 6
<p>The AES encryption key expansion unit after extraction of CS.</p> Full article ">Figure 7
<p>AES encryption/decryption unified key expansion unit based on the DDBT structure.</p> Full article ">Figure 8
<p>An example of AES encryption/decryption circuit.</p> Full article ">
32 pages, 8060 KiB  
Article
Study on Robust Path-Tracking Control for an Unmanned Articulated Road Roller Under Low-Adhesion Conditions
by Wei Qiang, Wei Yu, Quanzhi Xu and Hui Xie
Electronics 2025, 14(2), 383; https://doi.org/10.3390/electronics14020383 - 19 Jan 2025
Viewed by 422
Abstract
To enhance the path-tracking accuracy of unmanned articulated road roller (UARR) operating on low-adhesion, slippery surfaces, this paper proposes a hierarchical cascaded control (HCC) architecture integrated with real-time ground adhesion coefficient estimation. Addressing the complex nonlinear dynamics between the two rigid bodies of [...] Read more.
To enhance the path-tracking accuracy of unmanned articulated road roller (UARR) operating on low-adhesion, slippery surfaces, this paper proposes a hierarchical cascaded control (HCC) architecture integrated with real-time ground adhesion coefficient estimation. Addressing the complex nonlinear dynamics between the two rigid bodies of the vehicle and its interaction with the ground, an upper-layer nonlinear model predictive controller (NMPC) is designed. This layer, based on a 4-degree-of-freedom (4-DOF) dynamic model, calculates the required steering torque using position and heading errors. The lower layer employs a second-order sliding mode controller (SOSMC) to precisely track the steering torque and output the corresponding steering wheel angle. To accommodate the anisotropic and time-varying nature of slippery surfaces, a strong-tracking unscented Kalman filter (ST-UKF) observer is introduced for ground adhesion coefficient estimation. By dynamically adjusting the covariance matrix, the observer reduces reliance on historical data while increasing the weight of new data, significantly improving real-time estimation accuracy. The estimated adhesion coefficient is fed back to the upper-layer NMPC, enhancing the control system’s adaptability and robustness under slippery conditions. The HCC is validated through simulation and real-vehicle experiments and compared with LQR and PID controllers. The results demonstrate that HCC achieves the fastest response time and smallest steady-state error on both dry and slippery gravel soil surfaces. Under slippery conditions, while control performance decreases compared to dry surfaces, incorporating ground adhesion coefficient observation reduces steady-state error by 20.62%. Full article
(This article belongs to the Section Electrical and Autonomous Vehicles)
►▼ Show Figures

Figure 1

Figure 1
<p>UARR hardware layout.</p> Full article ">Figure 2
<p>Causality-based modeling simulation platform for road roller.</p> Full article ">Figure 3
<p>Force analysis of UARR dual bodies: (<b>a</b>) 3D force analysis; (<b>b</b>) planar force analysis and structural parameters.</p> Full article ">Figure 4
<p>Fitting the Dugoff model to shearing stress–shearing displacement data [<a href="#B39-electronics-14-00383" class="html-bibr">39</a>].</p> Full article ">Figure 5
<p>Feasibility verification of the Dugoff model for the drum: (<b>a</b>) circular test; (<b>b</b>) X coordinate.</p> Full article ">Figure 6
<p>Equivalent schematic of the UARR hydraulic steering system.</p> Full article ">Figure 7
<p>Relative displacement between the valve spool and valve sleeve.</p> Full article ">Figure 8
<p>Principle diagram of piston rod movement.</p> Full article ">Figure 9
<p>Validation of the dynamics model on wet dirt road: (<b>a</b>) yaw angle; (<b>b</b>) yaw rate.</p> Full article ">Figure 10
<p>Validation of the dynamics model on wet gravel road: (<b>a</b>) yaw angle; (<b>b</b>) yaw rate.</p> Full article ">Figure 11
<p>Validation of the dynamics model on wet dirt road: (<b>a</b>) drum centroid latitude; (<b>b</b>) drum centroid longitude.</p> Full article ">Figure 12
<p>Validation of the dynamics model on wet gravel road: (<b>a</b>) drum centroid latitude; (<b>b</b>) drum centroid longitude.</p> Full article ">Figure 13
<p>Hierarchical cascaded framework integrating NMPC and SOSMC with adhesion coefficient estimation.</p> Full article ">Figure 14
<p>SOSMC tracking performance verification: (<b>a</b>) tracking target steering torque <span class="html-italic">M<sub>j</sub></span>; (<b>b</b>) corresponding steering wheel angle output.</p> Full article ">Figure 15
<p>(<b>a</b>) Experimental scenario; (<b>b</b>) dry surface; (<b>c</b>) slippery surface.</p> Full article ">Figure 16
<p>Step-tracking experiment under dry conditions: (<b>a</b>) lateral error; (<b>b</b>) steady-state error distribution.</p> Full article ">Figure 17
<p>Straight line experiment under dry conditions: (<b>a</b>) lateral error; (<b>b</b>) lateral error distribution.</p> Full article ">Figure 18
<p>Step-tracking experiment under dry wet and slippery conditions: (<b>a</b>) lateral error; (<b>b</b>) steady-state error distribution.</p> Full article ">Figure 19
<p>Straight line experiment under wet and slippery conditions: (<b>a</b>) lateral error; (<b>b</b>) lateral error distribution.</p> Full article ">Figure 20
<p>Ground surface adhesion coefficient estimation based on ST-UKF.</p> Full article ">Figure 21
<p>Comparison of lateral errors across controllers on roads with varying adhesion coefficients.</p> Full article ">Figure 22
<p>Lateral error distribution of different controllers under varying adhesion coefficients.</p> Full article ">
23 pages, 1147 KiB  
Article
Mutation-Based Approach to Supporting Human–Machine Pair Inspection
by Yujun Dai, Shaoying Liu and Haiyi Liu
Electronics 2025, 14(2), 382; https://doi.org/10.3390/electronics14020382 - 19 Jan 2025
Viewed by 262
Abstract
Human–machine pair inspection refers to a technique that supports programmers and machines working together as a “pair” in source code inspection tasks. The machine provides guidance, while the programmer performs the inspection based on this guidance. Although programmers are often best suited to [...] Read more.
Human–machine pair inspection refers to a technique that supports programmers and machines working together as a “pair” in source code inspection tasks. The machine provides guidance, while the programmer performs the inspection based on this guidance. Although programmers are often best suited to inspect their own code due to familiarity, overconfidence may lead them to overlook important details. This study introduces a novel mutation-based human–machine pair inspection method, which is designed to direct the programmer’s attention to specific code components by applying targeted mutations. We assess the effectiveness of code inspections by analyzing the programmer’s corrections of these mutations. Our approach involves defining mutation operators for each keyword in the program based on historical defects, developing mutation rules based on program keywords and a strategy for automatically generating mutants, and designing a code comparison strategy to quantitatively evaluate code inspection quality. Through a controlled experiment, we demonstrate the effectiveness of mutation-based human–machine pair inspection in aiding programmers during the inspection process. Full article
(This article belongs to the Special Issue Intelligent Approaches for Solving Software Problems with AI Techniques)
►▼ Show Figures

Figure 1

Figure 1
<p>Overview of the human–machine pair inspection process.</p> Full article ">Figure 2
<p>Detailed steps of the human–machine pair inspection process.</p> Full article ">Figure 3
<p>Mutation testing process.</p> Full article ">Figure 4
<p>Overview of the mutation-based human–machine pair inspection method.</p> Full article ">Figure 5
<p>AST for Python code.</p> Full article ">Figure 6
<p>AST for Python code with mutation.</p> Full article ">
19 pages, 1319 KiB  
Article
Towards Failure-Aware Inference in Harsh Operating Conditions: Robust Mobile Offloading of Pre-Trained Neural Networks
by Wenjing Liu, Zhongmin Chen and Yunzhan Gong
Electronics 2025, 14(2), 381; https://doi.org/10.3390/electronics14020381 - 19 Jan 2025
Viewed by 314
Abstract
Pre-trained neural networks like GPT-4 and Llama2 have revolutionized intelligent information processing, but their deployment in industrial applications faces challenges, particularly in harsh environments. To address these related issues, model offloading, which involves distributing the computational load of pre-trained models across edge devices, [...] Read more.
Pre-trained neural networks like GPT-4 and Llama2 have revolutionized intelligent information processing, but their deployment in industrial applications faces challenges, particularly in harsh environments. To address these related issues, model offloading, which involves distributing the computational load of pre-trained models across edge devices, has emerged as a promising solution. While this approach enables the utilization of more powerful models, it faces significant challenges in harsh environments, where reliability, connectivity, and resilience are critical. This paper introduces failure-resilient inference in mobile networks (FRIM), a framework that ensures robust offloading and inference without the need for model retraining or reconstruction. FRIM leverages graph theory to optimize partition redundancy and incorporates an adaptive failure detection mechanism for mobile inference with efficient fault tolerance. Experimental results on DNN models (AlexNet, ResNet, VGG-16) show that FRIM improves inference performance and resilience, enabling more reliable mobile applications in harsh operating environments. Full article
(This article belongs to the Special Issue New Advances in Distributed Computing and Its Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>A pre-trained model-based mobile inference with redundancy.</p> Full article ">Figure 2
<p>An execution graph of neural network models with redundant offloading.</p> Full article ">Figure 3
<p>An example for partition offloading with redundancy.</p> Full article ">Figure 4
<p>An example for adapting redundancy.</p> Full article ">Figure 5
<p>Execution graph with redundant model partitions.</p> Full article ">Figure 6
<p>Average accuracy of FRIM and other strategies under different numbers of failures.</p> Full article ">Figure 7
<p>Latency for ten inference queries on FRIM and other strategies.</p> Full article ">
22 pages, 1818 KiB  
Article
Cooperative Service Caching and Task Offloading in Mobile Edge Computing: A Novel Hierarchical Reinforcement Learning Approach
by Tan Chen, Jiahao Ai, Xin Xiong and Guangwu Hu
Electronics 2025, 14(2), 380; https://doi.org/10.3390/electronics14020380 - 19 Jan 2025
Viewed by 456
Abstract
In the current mobile edge computing (MEC) system, the user dynamics, diversity of applications, and heterogeneity of services have made cooperative service caching and task offloading decision increasingly important. Service caching and task offloading have a naturally hierarchical structure, and thus, hierarchical reinforcement [...] Read more.
In the current mobile edge computing (MEC) system, the user dynamics, diversity of applications, and heterogeneity of services have made cooperative service caching and task offloading decision increasingly important. Service caching and task offloading have a naturally hierarchical structure, and thus, hierarchical reinforcement learning (HRL) can be used to effectively alleviate the dimensionality curse in it. However, traditional HRL algorithms are designed for short-term missions with sparse rewards, while existing HRL algorithms proposed for MEC lack delicate a coupling structure and perform poorly. This article introduces a novel HRL-based algorithm, named hierarchical service caching and task offloading (HSCTO), to solve the problem of the cooperative optimization of service caching and task offloading in MEC. The upper layer of HSCTO makes decisions on service caching while the lower layer is in charge of task offloading strategies. The upper-layer module learns policies by directly utilizing the rewards of the lower-layer agent, and the tightly coupled design guarantees algorithm performance. Furthermore, we adopt a fixed multiple time step method in the upper layer, which eliminates the dependence on the semi-Markov decision processes (SMDPs) theory and reduces the cost of frequent service replacement. We conducted numerical evaluations and the experimental results show that HSCTO improves the overall performance by 20%, and reduces the average energy consumption by 13% compared with competitive baselines. Full article
(This article belongs to the Special Issue Advanced Technologies in Edge Computing and Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>System model.</p> Full article ">Figure 2
<p>Architecture of hierarchical reinforcement learning.</p> Full article ">Figure 3
<p>Reward of service caching agent during training process.</p> Full article ">Figure 4
<p>Rewards of task offloading agents during training process. (<b>a</b>–<b>c</b>) show the reward curves of the 3 agents respectively.</p> Full article ">Figure 5
<p>Comparison of rewards with different <math display="inline"><semantics> <mi>η</mi> </semantics></math>.</p> Full article ">Figure 6
<p>Comparison of average utility with a different <math display="inline"><semantics> <mi>η</mi> </semantics></math>.</p> Full article ">Figure 7
<p>Comparison of the reward for different algorithms.</p> Full article ">Figure 8
<p>Comparison of average utility for different algorithms.</p> Full article ">
25 pages, 5264 KiB  
Article
Intelligent Gas Risk Assessment and Report Generation for Coal Mines: An Innovative Framework Based on GLM Fine-Tuning
by Yi Sun, Ying Han and Xinke Liu
Electronics 2025, 14(2), 379; https://doi.org/10.3390/electronics14020379 - 19 Jan 2025
Viewed by 380
Abstract
Traditional coal mine gas risk assessment relies on manual operations, leading to inefficiencies, incomplete information integration, and insufficient evaluation accuracy, ultimately affecting safety oversight. This paper proposes an intelligent gas risk assessment and report generation framework (IGRARG) based on fine-tuning a Generative Language [...] Read more.
Traditional coal mine gas risk assessment relies on manual operations, leading to inefficiencies, incomplete information integration, and insufficient evaluation accuracy, ultimately affecting safety oversight. This paper proposes an intelligent gas risk assessment and report generation framework (IGRARG) based on fine-tuning a Generative Language Model (GLM) to address these challenges. The framework integrates multi-source sensor data with the reasoning capabilities of large language models (LLMs). It constructs a gas risk dataset for coal mine safety scenarios, fine-tuned with GLM. Incorporating industry regulations and a domain-specific knowledge base enhanced with a Retrieval-Augmented Generation (RAG) mechanism, the framework automates alarm judgment, suggestion generation, and report creation via a hierarchical graph structure. Real-time human feedback further refines decision making. Experimental results show an evaluation accuracy of 85–93%, with over 300 field tests achieving a 94.46% alarm judgment accuracy and reducing weekly report generation from 90 min to 2–3 min. This framework significantly enhances the intelligence and efficiency of gas risk assessment, providing robust decision support for coal mine safety management. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Overall framework of IGRARG.</p> Full article ">Figure 2
<p>Data collection and transmission process flowchart.</p> Full article ">Figure 3
<p>Time series diagram of sensor monitoring data. (<b>a</b>) Time series data collected by the carbon monoxide sensor over a period of time; (<b>b</b>) Time series data collected by the laser methane sensor over a period of time.</p> Full article ">Figure 4
<p>Flowchart of knowledge retrieval process.</p> Full article ">Figure 5
<p>Main roadmap of the intelligent assessment report generation module.</p> Full article ">Figure 6
<p>Risk assessment distribution of different gases under the IGRARG framework.</p> Full article ">Figure 7
<p>Comparison of evaluation reports for different sections of the coal mine (taking methane as an example).</p> Full article ">Figure 8
<p>Comparison of evaluation reports for different sections of the coal mine (taking carbon monoxide as an example).</p> Full article ">Figure 9
<p>Distribution of expert evaluation results. (<b>a</b>) Multidimensional evaluation distribution of the alarm assessment results by experts; (<b>b</b>) Multidimensional evaluation distribution of the assessment report by experts.</p> Full article ">
51 pages, 10695 KiB  
Article
AI on Wheels: Bibliometric Approach to Mapping of Research on Machine Learning and Deep Learning in Electric Vehicles
by Adrian Domenteanu, Liviu-Adrian Cotfas, Paul Diaconu, George-Aurelian Tudor and Camelia Delcea
Electronics 2025, 14(2), 378; https://doi.org/10.3390/electronics14020378 - 19 Jan 2025
Viewed by 1074
Abstract
The global transition to sustainable energy systems has placed the use of electric vehicles (EVs) among the areas that might contribute to reducing carbon emissions and optimizing energy usage. This paper presents a bibliometric analysis of the interconnected domains of EVs, artificial intelligence [...] Read more.
The global transition to sustainable energy systems has placed the use of electric vehicles (EVs) among the areas that might contribute to reducing carbon emissions and optimizing energy usage. This paper presents a bibliometric analysis of the interconnected domains of EVs, artificial intelligence (AI), machine learning (ML), and deep learning (DL), revealing a significant annual growth rate of 56.4% in research activity. Key findings include the identification of influential journals, authors, countries, and collaborative networks that have driven advancements in this domain. This study highlights emerging trends, such as the integration of renewable energy sources, vehicle-to-grid (V2G) schemes, and the application of AI in EV battery optimization, charging infrastructure, and energy consumption prediction. The analysis also uncovers challenges in addressing information security concerns. By reviewing the top-cited papers, this research underlines the transformative potential of AI-driven solutions in enhancing EV performance and scalability. The results of this study can be useful for practitioners, academics, and policymakers. Full article
(This article belongs to the Special Issue Secure Integration of Artificial Intelligence (AI) and Autonomous Vehicular Networks)
►▼ Show Figures

Figure 1

Figure 1
<p>Steps in analysis.</p> Full article ">Figure 2
<p>Top 10 most relevant sources.</p> Full article ">Figure 3
<p>Core sources by Bradford’s law.</p> Full article ">Figure 4
<p>Top 10 sources of local impact by H-index.</p> Full article ">Figure 5
<p>Top 10 sources of production over time.</p> Full article ">Figure 6
<p>Sources quartiles.</p> Full article ">Figure 7
<p>Top 10 most relevant authors.</p> Full article ">Figure 8
<p>Top 10 most cited local authors.</p> Full article ">Figure 9
<p>Top 10 authors’ production over time.</p> Full article ">Figure 10
<p>Author productivity based on Lotka’s law.</p> Full article ">Figure 11
<p>Top 10 authors’ local impact by H-index.</p> Full article ">Figure 12
<p>Top 10 most important corresponding author’s country (SCP: Single-country publication; MCP: Multiple-country publication).</p> Full article ">Figure 13
<p>Top 10 most cited countries.</p> Full article ">Figure 14
<p>Top 10 countries’ production over time.</p> Full article ">Figure 15
<p>Top 10 most relevant affiliations.</p> Full article ">Figure 16
<p>Top 10 affiliations production over time.</p> Full article ">Figure 17
<p>Collaboration network.</p> Full article ">Figure 18
<p>Countries collaboration world map.</p> Full article ">Figure 19
<p>Factorial analysis for authors’ keywords.</p> Full article ">Figure 20
<p>Factorial analysis for Keywords Plus.</p> Full article ">Figure 21
<p>Thematic map Keywords Plus.</p> Full article ">Figure 22
<p>Thematic map Authors’ Keywords.</p> Full article ">Figure 23
<p>Top 50 words based on authors’ keywords (<b>A</b>), and Keywords Plus (<b>B</b>).</p> Full article ">Figure 24
<p>Three-field plot: countries (<b>left</b>), authors (<b>middle</b>), and keywords (<b>right</b>).</p> Full article ">Figure 25
<p>Three-field plot: affiliations (<b>left</b>), authors (<b>middle</b>), and Keywords Plus (<b>right</b>).</p> Full article ">
17 pages, 16060 KiB  
Article
Channel-Wise Attention-Enhanced Feature Mutual Reconstruction for Few-Shot Fine-Grained Image Classification
by Qianying Ou and Jinmiao Zou
Electronics 2025, 14(2), 377; https://doi.org/10.3390/electronics14020377 - 19 Jan 2025
Viewed by 308
Abstract
Fine-grained image classification is faced with the challenge of significant intra-class differences and subtle similarities between classes, with a limited number of labelled data. Previous few-shot learning approaches, however, often fail to recognize these discriminative details, such as a bird’s eyes and beak. [...] Read more.
Fine-grained image classification is faced with the challenge of significant intra-class differences and subtle similarities between classes, with a limited number of labelled data. Previous few-shot learning approaches, however, often fail to recognize these discriminative details, such as a bird’s eyes and beak. In this paper, we proposed a channel-wise attention-enhanced feature mutual reconstruction mechanism that helps to alleviate these problems for fine-grained image classification. This mechanism first employed a channel-wise attention module (CAM) to learn the channel weights for both the support and query features. We utilized channel-wise self-attention to assign greater importance to object-relevant channels. This helps the model to focus on subtle yet discriminative details, which is essential to the classification process. Then, we introduce a feature mutual reconstruction module (FMRM) to reconstruct features. The support features are reconstructed by a support-weight-enhanced feature map to reduce the intra-class variations, and query features are reconstructed by a query-weight-enhanced feature map to increase inter-class variations. The results of classification depend on the similarity between reconstructed features and enhanced features. We evaluated the performance based on four fine-grained image datasets when Conv-4 and Resnet-12 were used. The experimental results showed that our method outperforms previous few-shot fine-grained classification methods. This proves that our method can improve fine-grained image classification performance and simultaneously balance both the inter-class and intra-class variations. Full article
(This article belongs to the Special Issue AI/Machine Learning in Computer Vision/Image Processing and Natural Language Processing)
►▼ Show Figures

Figure 1

Figure 1
<p>The difference between fine-grained images and general images. Rows represent different species and columns represent different backgrounds. (<b>a</b>) Fine-grained image examples; (<b>b</b>) general image examples.</p> Full article ">Figure 2
<p>Overview of our approach. The channel-wise feature mutual reconstruction contains four sub-modules. <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>c</mi> <mi>S</mi> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>i</mi> <mi>Q</mi> </msubsup> </semantics></math> are extracted features. <math display="inline"><semantics> <msubsup> <mi>w</mi> <mi>c</mi> <mi>S</mi> </msubsup> </semantics></math> represents the attention weight of the <span class="html-italic">c</span>th class support images and <math display="inline"><semantics> <msubsup> <mi>w</mi> <mi>i</mi> <mi>Q</mi> </msubsup> </semantics></math> represents the attention weight of the <span class="html-italic">i</span>th query instance. <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>c</mi> <mrow> <mi>S</mi> <mo>→</mo> <mi>Q</mi> </mrow> </msubsup> </semantics></math> represents the query feature reconstructed by support feature <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>c</mi> <mi>S</mi> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>i</mi> <mrow> <mi>Q</mi> <mo>→</mo> <mi>S</mi> </mrow> </msubsup> </semantics></math> represents the support feature reconstructed by query feature <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>i</mi> <mi>Q</mi> </msubsup> </semantics></math>. <math display="inline"><semantics> <msub> <mi>Q</mi> <mi>i</mi> </msub> </semantics></math> represents the query features enhanced by <math display="inline"><semantics> <msubsup> <mi>w</mi> <mi>c</mi> <mi>S</mi> </msubsup> </semantics></math>, while <math display="inline"><semantics> <msubsup> <mi>S</mi> <mi>c</mi> <mo>′</mo> </msubsup> </semantics></math> represents support features enhanced by <math display="inline"><semantics> <msubsup> <mi>w</mi> <mi>i</mi> <mi>Q</mi> </msubsup> </semantics></math>. After that, we calculate the similarity between <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>c</mi> <mi>S</mi> </msubsup> </semantics></math> and <math display="inline"><semantics> <msub> <mi>Q</mi> <mi>i</mi> </msub> </semantics></math>, as well as <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>i</mi> <mi>Q</mi> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>S</mi> <mi>c</mi> <mo>′</mo> </msubsup> </semantics></math>, to obtain the results.</p> Full article ">Figure 3
<p>Channel-wise attention module. <span class="html-italic">Q</span>, <span class="html-italic">K</span>, and <span class="html-italic">V</span> are obtained through a <math display="inline"><semantics> <mrow> <mn>1</mn> <mo>×</mo> <mn>1</mn> </mrow> </semantics></math> convolution kernel, representing the query, key, and value of the images in the channel dimension, respectively.</p> Full article ">Figure 4
<p>Overview of our feature mutual reconstruction module. <math display="inline"><semantics> <msub> <mi>S</mi> <mi>c</mi> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi>Q</mi> <mi>i</mi> </msub> </semantics></math> are the features reassigned by the support attention weight <math display="inline"><semantics> <msubsup> <mi>w</mi> <mi>c</mi> <mi>s</mi> </msubsup> </semantics></math>. <math display="inline"><semantics> <msubsup> <mi>S</mi> <mi>c</mi> <mo>′</mo> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>Q</mi> <mi>i</mi> <mo>′</mo> </msubsup> </semantics></math> are the features reassigned by the support attention weight <math display="inline"><semantics> <msubsup> <mi>w</mi> <mi>i</mi> <mi>s</mi> </msubsup> </semantics></math>. <math display="inline"><semantics> <msubsup> <mi>d</mi> <mrow> <mi>i</mi> <mo>.</mo> <mi>c</mi> </mrow> <mrow> <mi>S</mi> <mo>→</mo> <mi>Q</mi> </mrow> </msubsup> </semantics></math> calculates the similarity between <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>c</mi> <mrow> <mi>S</mi> <mo>→</mo> <mi>Q</mi> </mrow> </msubsup> </semantics></math> and <math display="inline"><semantics> <msub> <mi>Q</mi> <mi>i</mi> </msub> </semantics></math>, while <math display="inline"><semantics> <msubsup> <mi>d</mi> <mrow> <mi>i</mi> <mo>.</mo> <mi>c</mi> </mrow> <mrow> <mi>Q</mi> <mo>→</mo> <mi>S</mi> </mrow> </msubsup> </semantics></math> calculates the similarity between <math display="inline"><semantics> <msubsup> <mi>S</mi> <mi>c</mi> <mo>′</mo> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>F</mi> <mi>i</mi> <mrow> <mi>Q</mi> <mo>→</mo> <mi>S</mi> </mrow> </msubsup> </semantics></math>.</p> Full article ">Figure 5
<p>Five-way K-shot classification performance. We obtained the data by testing the same model on a 5-way K-shot classification task. This chosen model was trained on 5-way 5-shot classification with a Resnet-12 backbone on CUB.</p> Full article ">Figure 6
<p>C-way 5-shot classification performance. We obtained the data by testing the same model on a C-way 5-shot classification task. This chosen model was trained on 5-way 5-shot classification with a Resnet-12 backbone on CUB.</p> Full article ">Figure 7
<p>Visualization of image features recovered by inverse Resnet-12 in the CUB datasets. From left to right, the images represent the original support features, enhanced support features, reconstructed support features, reconstructed query features, enhanced query features, and the original query features.</p> Full article ">Figure 8
<p>Visualization of the discriminative regions by GradCAM in the CUB datasets. From top to bottom, the images represent the original CUB image, ProtoNet’s visualization, and our visualization.</p> Full article ">
18 pages, 8288 KiB  
Article
Improved Sage–Husa Unscented Kalman Filter for Harmonic State Estimation in Distribution Grid
by Peixuan Yu and Jianjun Sun
Electronics 2025, 14(2), 376; https://doi.org/10.3390/electronics14020376 - 19 Jan 2025
Viewed by 393
Abstract
In distribution grids with the large-scale integration of renewable energy sources and energy storage systems, power signals are often contaminated with time-varying noise and frequency deviation caused by low-frequency inertia. To achieve an accurate dynamic harmonic state estimate (HSE), a novel method based [...] Read more.
In distribution grids with the large-scale integration of renewable energy sources and energy storage systems, power signals are often contaminated with time-varying noise and frequency deviation caused by low-frequency inertia. To achieve an accurate dynamic harmonic state estimate (HSE), a novel method based on an improved Sage–Husa unscented Kalman filter (ISHUKF) is proposed. Considering the frequency deviation, a nonlinear filter model for power signal is proposed, and a UKF is used to address the nonlinear estimation. A Sage–Husa noise estimator is incorporated to enhance the robustness of the UKF-based HSE against the time-varying noise. Additionally, the noise covariance of the Sage–Husa algorithm is modified to ensure the rapid convergence of the estimation. Then, the performance of the proposed method is validated using an IEEE 14-node system. Finally, the method is applied to evaluate the harmonic states of grid-connected inverter faults in real-world scenarios. The simulation and experiment results demonstrate that the proposed method provides an accurate dynamic HSE even in the presence of time-varying noise and frequency deviation. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>The flowchart of the ISHUKF for HSE.</p> Full article ">Figure 2
<p>IEEE 14 node diagram including harmonic loads at busbars 10 and 13.</p> Full article ">Figure 3
<p>The harmonic current and the increasing noise.</p> Full article ">Figure 4
<p>The performance comparison of methods for the 5-th harmonic estimation. The blue curve is the estimation and the dashed black one is the true value.</p> Full article ">Figure 5
<p>The performance comparison of methods for the 7-th harmonic estimation. The blue curve is the estimation and the dashed black one is the true value.</p> Full article ">Figure 6
<p>The performance comparison of methods for the 11-th harmonic estimation. The blue curve is the estimation and the dashed black one is the true value.</p> Full article ">Figure 7
<p>The performance comparison of methods for the 13-th harmonic estimation. The blue curve is the estimation and the dashed black one is the true value.</p> Full article ">Figure 8
<p>Experiment platform for harmonic state estimation test.</p> Full article ">Figure 9
<p>The field faulty currents at 49.5 Hz.</p> Full article ">Figure 10
<p>The estimation comparison of the second harmonic and the third harmonic of Phase A (faulty phase) current with single switch open-circuited.</p> Full article ">Figure 11
<p>The estimation comparison of the second harmonic and the third harmonic of phase B (unfaulty phase) current.</p> Full article ">Figure 12
<p>The estimation comparison of the second harmonic and the third harmonic of phase C (unfaulty phase) current.</p> Full article ">Figure 13
<p>The currents of the inverter when the upper arms of Phase A and Phase B are open-circuited.</p> Full article ">Figure 14
<p>The estimation comparison of the second harmonic and the third harmonic of Phase A (faulty phase) current with multiple switch open-circuited.</p> Full article ">
6 pages, 138 KiB  
Editorial
Control and Applications of Intelligent Unmanned Aerial Vehicles
by Yunda Yan, Dewei Yi, Hao Lu and Lan Gao
Electronics 2025, 14(2), 375; https://doi.org/10.3390/electronics14020375 - 19 Jan 2025
Viewed by 496
Abstract
Embedding intelligence into the control design of unmanned aerial vehicles (UAVs) has become increasingly critical as these systems are deployed in complex and uncertain environments, often requiring them to adapt dynamically to unpredictable events [...] Full article
(This article belongs to the Special Issue Control and Applications of Intelligent Unmanned Aerial Vehicle)
24 pages, 5134 KiB  
Article
A Novel Data Sanitization Method Based on Dynamic Dataset Partition and Inspection Against Data Poisoning Attacks
by Jaehyun Lee, Youngho Cho, Ryungeon Lee, Simon Yuk, Jaepil Youn, Hansol Park and Dongkyoo Shin
Electronics 2025, 14(2), 374; https://doi.org/10.3390/electronics14020374 - 18 Jan 2025
Viewed by 566
Abstract
Deep learning (DL) technology has shown outstanding performance in various fields such as object recognition and classification, speech recognition, and natural language processing. However, it is well known that DL models are vulnerable to data poisoning attacks, where adversaries modify or inject data [...] Read more.
Deep learning (DL) technology has shown outstanding performance in various fields such as object recognition and classification, speech recognition, and natural language processing. However, it is well known that DL models are vulnerable to data poisoning attacks, where adversaries modify or inject data samples maliciously during the training phase, leading to degraded classification accuracy or misclassification. Since data poisoning attacks keep evolving to avoid existing defense methods, security researchers thoroughly examine data poisoning attack models and devise more reliable and effective detection methods accordingly. In particular, data poisoning attacks can be realistic in an adversarial situation where we retrain a DL model with a new dataset obtained from an external source during transfer learning. By this motivation, we propose a novel defense method that partitions and inspects the new dataset and then removes malicious sub-datasets. Specifically, our proposed method first divides a new dataset into n sub-datasets either evenly or randomly, inspects them by using the clean DL model as a poisoned dataset detector, and finally removes malicious sub-datasets classified by the detector. For partition and inspection, we design two dynamic defensive algorithms: the Sequential Partitioning and Inspection Algorithm (SPIA) and the Randomized Partitioning and Inspection Algorithm (RPIA). With this approach, a resulting cleaned dataset can be used reliably for retraining a DL model. In addition, we conducted two experiments in the Python and DL environment to show that our proposed methods effectively defend against two data poisoning attack models (concentrated poisoning attacks and random poisoning attacks) in terms of various evaluation metrics such as removed poison rate (RPR), attack success rate (ASR), and classification accuracy (ACC). Specifically, the SPIA completely removed all poisoned data under concentrated poisoning attacks in both Python and DL environments. In addition, the RPIA removed up to 91.1% and 99.1% of poisoned data under random poisoning attacks in Python and DL environments, respectively. Full article
(This article belongs to the Special Issue Big Data Analytics and Information Technology for Smart Cities)
►▼ Show Figures

Figure 1

Figure 1
<p>Classification of adversarial attacks based on attack timing.</p> Full article ">Figure 2
<p>Proposed defense method based on dataset partition and inspection.</p> Full article ">Figure 3
<p>No partition vs. 2-way partition.</p> Full article ">Figure 4
<p>Comparison of 2-way partition and 4-way partition.</p> Full article ">Figure 5
<p>An example of the sequential partition and inspection algorithm (SPIA).</p> Full article ">Figure 6
<p>An example of random partition and inspection algorithm (RPIA).</p> Full article ">Figure 7
<p>Graphical representation of evaluation results for concentrated poisoning attacks (Python simulation).</p> Full article ">Figure 8
<p>Graphical representation of evaluation results for random poisoning attacks (Python simulation).</p> Full article ">Figure 9
<p>Example images from each class in the CIFAR-10 dataset.</p> Full article ">Figure 10
<p>Graphical representation of evaluation results for concentrated poisoning attacks (DL training).</p> Full article ">Figure 11
<p>Graphical representation of evaluation results for random poisoning attacks (DL training).</p> Full article ">Figure 12
<p>This figure illustrates the performance of a transfer-learned model using various defense methods under concentrated poisoning attacks. The accuracy of the pre-trained model (Mt1) is 81.34%. When 20% poisoned data are added to Dt, the following transfer results are observed: (<b>a</b>) <span class="html-italic">No sanitization</span>: Without any defense method, achieves an accuracy of 61.7% under a 0% RPR. (<b>b</b>) <span class="html-italic">SPIA defense</span>: Applying the SPIA defense method with n = 100 reduces the impact of poisoning and improves the accuracy to 81.2%. (<b>c</b>) <span class="html-italic">RPIA defense</span>: Using the RPIA defense method with n = 8000 achieves an accuracy of 80.7%, effectively mitigating the poisoning attack with an RPR of 98.7%. The regions highlighted in red show the attack causes a significant number of misclassifications, particularly affecting specific classes (Class 0 and Class 1).</p> Full article ">Figure 13
<p>This figure illustrates the performance of a transfer-learned model using various defense methods under random poisoning attacks, while maintaining the experimental conditions consistent with those described in <a href="#electronics-14-00374-f007" class="html-fig">Figure 7</a>: (<b>a</b>) <span class="html-italic">No sanitization</span>: Without any defense method, an accuracy of 61.7% with a 0% RPR. (<b>b</b>) <span class="html-italic">SPIA defense</span>: Applying the SPIA defense method with n = 8000 brings the model’s accuracy to 81.2% and an RPR of 97.6%. (<b>c</b>) <span class="html-italic">RPIA defense</span>: Using the RPIA defense method with n = 4000 achieves an accuracy of 80.7% and an RPR of 98.8%.</p> Full article ">
16 pages, 3826 KiB  
Article
Research on Transformer Temperature Early Warning Method Based on Adaptive Sliding Window and Stacking
by Pan Zhang, Qian Zhang, Huan Hu, Huazhi Hu, Runze Peng and Jiaqi Liu
Electronics 2025, 14(2), 373; https://doi.org/10.3390/electronics14020373 - 18 Jan 2025
Viewed by 418
Abstract
This paper proposes a transformer temperature early warning method based on an adaptive sliding window and stacking ensemble learning algorithm, aiming to improve the accuracy and robustness of temperature prediction. The transformer temperature early warning system is crucial for ensuring the safe operation [...] Read more.
This paper proposes a transformer temperature early warning method based on an adaptive sliding window and stacking ensemble learning algorithm, aiming to improve the accuracy and robustness of temperature prediction. The transformer temperature early warning system is crucial for ensuring the safe operation of the power system, and temperature prediction, as the foundation of early warning, directly affects the early warning effectiveness. This paper analyzes the characteristics of transformer temperature using support vector regression, random forest, and gradient boosting regression as base learners and ridge regression as the meta-learner to construct a stacking model. At the same time, Bayesian optimization is used to automatically adjust the sliding window size, achieving adaptive sliding window processing. The experimental results indicate that the temperature prediction method based on adaptive sliding window and stacking significantly reduces prediction errors, enhances the model’s adaptability and generalization ability, and provides more reliable technical support for transformer fault warning. Full article
(This article belongs to the Special Issue Power Electronics in Hybrid AC/DC Grids and Microgrids)
►▼ Show Figures

Figure 1

Figure 1
<p>Stacking model structure diagram.</p> Full article ">Figure 2
<p>Flowchart of transformer temperature warning method based on adaptive sliding window and stacking.</p> Full article ">Figure 3
<p>Transformer temperature prediction results of fixed sliding window and stacking.</p> Full article ">Figure 4
<p>Transformer temperature prediction results of adaptive sliding window and stacking.</p> Full article ">Figure 5
<p>Transformer temperature prediction results of empirical mode decomposition bidirectional long short-term memory.</p> Full article ">Figure 6
<p>Fixed sliding window stacking model transformer temperature prediction error curve graph.</p> Full article ">Figure 7
<p>Adaptive sliding window stacking model transformer temperature prediction error curve graph.</p> Full article ">Figure 8
<p>Line graph of total temperature prediction error of phases A, B, and C as a function of sliding window size.</p> Full article ">
21 pages, 651 KiB  
Article
A Comparative Study of Incremental ΔΣ Analog-to-Digital Converter Architectures with Extended Order and Resolution
by Monica Aziz, Paul Kaesser, Sameh Ibrahim and Maurits Ortmanns
Electronics 2025, 14(2), 372; https://doi.org/10.3390/electronics14020372 - 18 Jan 2025
Viewed by 539
Abstract
Incremental Delta-Sigma (I-DS) analog-to-digital converters (ADCs) are one of the best candidates for integrated sensor interface systems when it comes to high resolution and power efficiency. Advanced architectures such as Multistage noise shaping (MASH) or extended counting (EC) I-DS ADCs can be used [...] Read more.
Incremental Delta-Sigma (I-DS) analog-to-digital converters (ADCs) are one of the best candidates for integrated sensor interface systems when it comes to high resolution and power efficiency. Advanced architectures such as Multistage noise shaping (MASH) or extended counting (EC) I-DS ADCs can be used to achieve a high resolution and fast conversion times and avoid stability issues. Different architectures have been proposed in the state of the art (SoA), but there exists no extensive quantitative or qualitative comparison between them. This manuscript fills this gap by providing a detailed system-level comparison between MASH, EC, and other architectural options in I-DS ADCs, where different performances between these architectures are realized depending on the employed oversampling ratio (OSR) and the chosen number of quantizer bits. Also, for specific MASH designs, the appropriate choice of the digital filter improves the SQNR. The advantages, disadvantages, and limitations of the different architectures are presented including non-idealities such as coefficient mismatch showing that 2-1 MASH-LI is less sensitive to mismatch and provides a high maximum stable amplitude (MSA) relative to the simulated architectures. Furthermore, the 2-1 EC achieves good results and comes with the advantage of a lower noise penalty factor compared to the MASH architectures. This work is intended to assist designers in selecting the most appropriate enhanced I-DS MASH architecture for their specific requirements and applications. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Multi-stage/-step I-DS ADC with three variants of requantization and (<b>b</b>) its timing diagram including the reset.</p> Full article ">Figure 2
<p>(<b>a</b>) Block diagram of an exemplary incremental 2-1 MASH modulator with digital cancellation logic and reconstruction filter. The red dashed line indicates QE extraction from the first-stage quantizer (MASH-QE). The blue dashed line indicates the extraction of the QE from the last integrator output (MASH-LI) and (<b>b</b>) its timing diagram, including the reset.</p> Full article ">Figure 3
<p>(<b>a</b>) Block diagram of an incremental 2-1 EC (Extended Counting) modulator with reconstruction filters and (<b>b</b>) its timing diagram, including the reset.</p> Full article ">Figure 4
<p>Simulated SQNR of 2-0 and 2-1 MASH-QE, LI and EC at an OSR of 60. First stage: 3-bit quantizer. Second stage: 8-bit quantizer (2-0) or 4-bit quantizer (2-1).</p> Full article ">Figure 5
<p>Simulated SQNR of 2-0 and 2-1 MASH-QE, LI and EC at an OSR of 60 including limiter blocks after each integrator and after the adder. First stage: 3-bit quantizer. Second stage: 8-bit quantizer (2-0) or 4-bit quantizer (2-1).</p> Full article ">Figure 6
<p>Histogram of the swing at the output of the adder (before the limiter) at an input of −6 dBFS and at an input of −3 dBFS, respectively.</p> Full article ">Figure 7
<p>Histogram of the swing at the last integrator output and the last sample of the last integrator output at an input amplitude of −3 dBFS, respectively.</p> Full article ">Figure 8
<p>Simulated SQNR of 2-0 and 2-1 MASH-QE, LI and EC at an OSR of 60. An inter-stage gain of 6 is used for all the architectures. First stage: 3-bit quantizer. Second stage: 8-bit quantizer (2-0) or 4-bit quantizer (2-1).</p> Full article ">Figure 9
<p>Simulated SQNR of 2-0 and 2-1 MASH-QE, LI, and EC at an input amplitude of −6 dBFS sinusoidal signal. An inter-stage gain of 6 is used for all the architectures. First stage: 3-bit quantizer. Second stage: 8-bit quantizer (2-0) or 4-bit quantizer (2-1).</p> Full article ">Figure 10
<p>Simulated SQNR of 2-0 and 2-1 MASH-QE, LI and EC at an OSR of 100 and with limiter blocks included. First stage: 1-bit quantizer. Second stage: 8-bit quantizer (2-0) or 4-bit quantizer (2-1). The inter-stage gain is 1 for all architectures.</p> Full article ">Figure 11
<p>NTFs of the exemplary 2-0 MASH-LI for the two different versions of the error cancellation logic <span class="html-italic">ECL</span><sub>LI,1</sub> and <span class="html-italic">ECL</span><sub>LI,2</sub>. Solid: single-bit quantizer in first stage and OSR = 100. Dashed: 3-bit quantizer in first stage and OSR = 60. The second stage is an 8-bit quantizer.</p> Full article ">Figure 12
<p>Digital filter weights of the output of the second stage (<math display="inline"><semantics> <mrow> <msub> <mi>H</mi> <mn>2</mn> </msub> <mrow> <mo>(</mo> <mi>z</mi> <mo>)</mo> </mrow> <mo>·</mo> <mrow> <msub> <mi>H</mi> <mrow> <mi>rec</mi> </mrow> </msub> <mo stretchy="false">(</mo> <mi>z</mi> <mo stretchy="false">)</mo> </mrow> </mrow> </semantics></math>) in the case of the 2-0 MASH-LI at an OSR of 100. The first stage is scaled with the single-bit coefficients of <a href="#electronics-14-00372-t001" class="html-table">Table 1</a>. (<b>a</b>) shows the weights when <math display="inline"><semantics> <mrow> <msub> <mi>H</mi> <mn>2</mn> </msub> <mrow> <mo>(</mo> <mi>z</mi> <mo>)</mo> </mrow> </mrow> </semantics></math> is designed after <span class="html-italic">ECL</span><sub>LI,1</sub> and (<b>b</b>) for <span class="html-italic">ECL</span><sub>LI,2</sub>.</p> Full article ">Figure 13
<p>Simulated mean and distribution of SQNR of the exemplary 2-0 MASH-QE, 2-0 MASH-LI and 2-0 EC architectures over the variation of the analog coefficients <math display="inline"><semantics> <msub> <mi>σ</mi> <mi>c</mi> </msub> </semantics></math> in percent at an input amplitude of −6 dBFS sinusoidal signal. OSR = 60, 3-bit first-stage and 8-bit second-stage quantizers, g = 6.</p> Full article ">Figure 14
<p>Simulated mean and distribution of SQNR of the exemplary 2-1 MASH-QE, 2-1 MASH-LI, and 2-1 EC architectures at −6 dBFS sinusoidal input signal over the variation of analog coefficients with standard deviation <math display="inline"><semantics> <msub> <mi>σ</mi> <mi>c</mi> </msub> </semantics></math>. OSR = 60, 3-bit first stage and 4-bit second stage quantizers, g = 6.</p> Full article ">Figure 15
<p>Simulated SQNR of 2-0 and 2-1 MASH-QE, LI and EC at an OSR of 60. First stage: 3-bit quantizer. Second stage: 8-bit quantizer (2-0) or 4-bit quantizer (2-1). The input is −6 dBFS sinusoidal signal and g = 6.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-188
Previous Issue
Next Issue
Back to TopTop