Applied Sciences

22 pages, 7712 KB

Open AccessArticle

CT-Net: A Hybrid ConvNeXt–Transformer Approach for ASL Alphabet Classification

by Zhuofan Yang, Houjin Lu and Samaneh Shamshiri

Appl. Sci. 2026, 16(10), 5168; https://doi.org/10.3390/app16105168 (registering DOI) - 21 May 2026

Recognition of the American Sign Language (ASL) alphabet is of utmost importance in bridging the communication gap between the hearing-impaired and the hearing. However, robust classification remains difficult because some hand gestures are morphologically very similar. To address this problem, this study presents [...] Read more.

Recognition of the American Sign Language (ASL) alphabet is of utmost importance in bridging the communication gap between the hearing-impaired and the hearing. However, robust classification remains difficult because some hand gestures are morphologically very similar. To address this problem, this study presents CT-Net, a hybrid deep learning architecture that integrates ConvNeXt-Tiny with a lightweight Transformer encoder. CT-Net combines convolutional feature extraction and self-attention mechanisms, which enable it to capture fine-grained local patterns and long-range spatial dependencies effectively. The proposed model was extensively compared with various architectures including traditional CNNs, Transformer-based models, hybrid machine-learning approaches and recent lightweight hybrid networks. The experimental results show that CT-Net achieved the best overall performance with a peak accuracy of 95.67% on the enhanced ASL dataset. Ablation studies demonstrate the effectiveness of our design choice. CT-Net achieves a strong trade-off between recognition accuracy and computational efficiency with an inference rate of 163.55 Frames Per Second (FPS). These findings highlight the potential of hybrid frameworks as a powerful tool for fine-grained gesture recognition tasks. Full article

► Show Figures

Figure 1

17 pages, 1446 KB

Open AccessArticle

A Point Cloud Registration Method Based on Triangular Mesh Features

by Wenguang Wang, Jinshuo Zhao, Changshun Yuan and Haoran Wang

Appl. Sci. 2026, 16(10), 5167; https://doi.org/10.3390/app16105167 - 21 May 2026

Abstract

To address the issue where conventional initial registration methods combined with the Iterative Closest Point (ICP) algorithm are highly sensitive to initial parameters such as point cloud density and pose—often resulting in degraded accuracy or even misregistration—this paper proposes a Lidar point cloud [...] Read more.

To address the issue where conventional initial registration methods combined with the Iterative Closest Point (ICP) algorithm are highly sensitive to initial parameters such as point cloud density and pose—often resulting in degraded accuracy or even misregistration—this paper proposes a Lidar point cloud registration method integrating triangular mesh features. First, Crust Triangulation is employed to construct a triangular mesh from the input point cloud. Then, the keypoint extraction based on triangular meshes, termed KE-TM, is introduced. Subsequently, a Triangle Feature Histogram (TFH) is constructed as the feature descriptor. Based on this, an initial alignment method grounded in triangular meshes, referred to as TM-IA, is developed to achieve coarse registration of point clouds. Finally, the ICP algorithm is applied to refine the alignment. Comparative experiments conducted on incomplete Bunny point clouds demonstrate that the proposed KE-TM maintains a higher keypoint repeatability under reduced point cloud density. The combined TM-IA and ICP registration method can achieve rotation errors within 1° and translation errors within 1 mm under small initial pose deviations, while also maintaining robust performance under larger initial misalignments. Compared with traditional methods, the proposed method significantly reduces sensitivity to initial parameters and improves the accuracy. This method has certain practical significance for the precise alignment of 3D point clouds. Full article

21 pages, 3529 KB

Open AccessArticle

Multi-Task Learning-Based Speech Emotion Recognition Using Pre-Trained Acoustic Model

by Xiaoyu Wang, Kai Yao and Ying Yi

Appl. Sci. 2026, 16(10), 5166; https://doi.org/10.3390/app16105166 - 21 May 2026

Abstract

Accurate recognition of human emotions is crucial for human–computer interaction, and speech, as an important external manifestation of emotion, has attracted significant attention. Existing speech emotion recognition (SER) methods are predominantly based on single-task learning, which inadequately model speaker variability and other latent [...] Read more.

Accurate recognition of human emotions is crucial for human–computer interaction, and speech, as an important external manifestation of emotion, has attracted significant attention. Existing speech emotion recognition (SER) methods are predominantly based on single-task learning, which inadequately model speaker variability and other latent factors in speech, thereby limiting recognition performance. In this paper, a multi-task learning-based SER method leveraging a pre-trained acoustic model is proposed. Speech emotion recognition is treated as the primary task, while speaker recognition, gender recognition, and automatic speech recognition are introduced as auxiliary tasks. A multi-task learning framework based on hard parameter sharing is constructed to guide the model to learn shared acoustic representations that simultaneously encode emotional category characteristics, speaker identity, and other relevant information. Experiments conducted on the IEMOCAP dataset demonstrate that the proposed model achieves weighted accuracy (WA) and unweighted accuracy (UA) of 83.24% and 83.36%, respectively, under five-fold cross-validation, and 83.86% and 84.23%, respectively, under ten-fold cross-validation. In both settings, the proposed method consistently outperforms the baseline models, confirming its effectiveness in improving speech emotion recognition performance. Full article

► Show Figures

Figure 1

31 pages, 2920 KB

Open AccessArticle

An Efficient Reliability Analysis Method for Steel Structures Based on Support Vector Machines and Hyperparameter Optimization

by Yingshun Fang, Chengshu Yang, Cunpeng Liu and Dalian Bai

Appl. Sci. 2026, 16(10), 5165; https://doi.org/10.3390/app16105165 - 21 May 2026

Abstract

To address the challenge of exorbitant computational costs in the reliability analysis of complex steel structures, which stems from the impact of multiple sources of uncertainty throughout their entire lifecycle, this paper presents a comparative evaluation of the explicit reconstruction of the Limit [...] Read more.

To address the challenge of exorbitant computational costs in the reliability analysis of complex steel structures, which stems from the impact of multiple sources of uncertainty throughout their entire lifecycle, this paper presents a comparative evaluation of the explicit reconstruction of the Limit State Function (LSF) using SVM combined with Hyperparameter Optimization (HPO) for structural reliability analysis under constrained computational budgets. Although traditional Monte Carlo simulation (MCS) exhibits high accuracy, it requires a substantial number of finite element calculations, rendering it difficult to satisfy the efficiency requirements of engineering projects. Conversely, the first-order and second-order reliability methods (FORM/SORM) offer high computational efficiency but rely on explicit limit state functions, posing challenges for their direct application to complex structural systems. Thus, this study initially acquires response samples of the structure under various combinations of random variables through a limited number of finite element analyses (FEA). Subsequently, it employs an SVM to develop a highly accurate equivalent explicit limit state function, which serves as a substitute for the original implicit limit state function. Finally, it integrates Monte Carlo simulation to efficiently evaluate the structure’s failure probability and reliability index. Meanwhile, to tackle the problem of SVM model performance being highly susceptible to hyperparameters, this study presents a comparative analysis of four strategies: Bayesian Optimization (BO), Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Random Search (RS), aiming to identify the optimal parameter combination and improve the model’s generalization capability. Through verification with four progressive examples, including linear, nonlinear, truss, and multistory frame structures, the results demonstrate that the proposed method can accurately characterize the nonlinearity of structural responses. The obtained failure probabilities and reliability indices are in close agreement with those obtained from the direct Monte Carlo simulation (MCS) and existing research. Moreover, while maintaining computational accuracy, the method significantly reduces computational costs, thereby providing an efficient and practical solution for structural reliability analysis in engineering practice. Full article

16 pages, 978 KB

Open AccessArticle

Driving Waveform as a Design Variable for PFAS Plasma Degradation: Electron-Density-Driven Versus Reactive-Species-Driven Pathways

by Yejin Lee, Juncheol Kim, Hwanho Kim, Ki Ho Baek, Juyeon Choi, Yunchan Jang, Kwiyong Kim, Seunghun Lee, Sunghoon Jung, Oi Lun Li, Holak Kim, Joo Young Park and Sarnai Odsuren

Appl. Sci. 2026, 16(10), 5164; https://doi.org/10.3390/app16105164 - 21 May 2026

Abstract

Per- and polyfluoroalkyl substances (PFASs) are persistent micropollutants whose carbon–fluorine bonds resist conventional advanced oxidation. Nonthermal plasmas have emerged as a promising option for PFAS degradation, but the relative contributions of reactive oxygen species (ROS) and electrons are still being investigated. Herein, we [...] Read more.

Per- and polyfluoroalkyl substances (PFASs) are persistent micropollutants whose carbon–fluorine bonds resist conventional advanced oxidation. Nonthermal plasmas have emerged as a promising option for PFAS degradation, but the relative contributions of reactive oxygen species (ROS) and electrons are still being investigated. Herein, we compared sinusoidal alternating-current (AC) and nanosecond-pulsed discharges―in an identical plasma reactor with the same input power (30 W)―through diagnostics including voltage–current characterization, optical emission spectroscopy with vibrational and rotational temperatures and Hα Stark broadening for electron density, and aqueous H₂O₂ quantification. AC discharges produced more aqueous H₂O₂, stronger ·OH emission, and higher vibrational and rotational temperatures, yet showed lower perfluorooctanoic acid (PFOA) removal (85% ± 2%) and lower defluorination (61% ± 1%) than the pulsed discharge (96% ± 2% and 80% ± 2%, respectively). Among the diagnostics examined, electron density tracked the removal trend, being higher under pulsed operation (1.2 × 10¹⁶ vs. 8.3 × 10¹⁵ under AC operation). A pseudo-first-order kinetic model based on electron density qualitatively reproduced the observed PFOA decay rate, suggesting that the waveform may serve as a design variable for tuning electron and ROS-mediated pathways in plasma–water reactors. Full article

(This article belongs to the Section Applied Physics General)

18 pages, 330 KB

Open AccessReview

Shared Autonomous Vehicles (SAVs): A Multivocal Literature Review

by António Pedro Ribeiro Camacho, António Reis Pereira and Miguel Mira da Silva

Appl. Sci. 2026, 16(10), 5163; https://doi.org/10.3390/app16105163 - 21 May 2026

Abstract

This study presents a multivocal literature review (MLR) on the implementation of Shared Autonomous Vehicles (SAVs), a relatively new concept in urban mobility that merges autonomous driving with shared transportation. The purpose of this review is to analyse the feasibility, challenges and potential [...] Read more.

This study presents a multivocal literature review (MLR) on the implementation of Shared Autonomous Vehicles (SAVs), a relatively new concept in urban mobility that merges autonomous driving with shared transportation. The purpose of this review is to analyse the feasibility, challenges and potential impacts of SAV deployment by aggregating and synthesising insights from the academic literature and grey sources. The review addresses factors influencing deployment, including social acceptance, environmental impact, business models, policy frameworks, needs and barriers, and lessons from existing pilot programmes. The findings reveal that successful SAV implementation depends on combining technology, regulation and infrastructure. Public trust and perception of safety, cost and convenience can also significantly influence the adoption of this technology, as well as potential sustainability benefits (like reduced emissions and fewer private vehicles). Case studies from cities like Phoenix, San Francisco and Singapore show promising results but also context-specific challenges. This study concludes that future research should apply these insights to specific cities, where urban layouts and public transport reliance demand customised approaches to successfully deploy SAVs. Full article

45 pages, 11691 KB

Open AccessArticle

Multi-Objective Robotics Optimization Using Improved MO-BxR Algorithms

by Ravipudi Venkata Rao, Harishankar Morazha Variam and Joao Paulo Davim

Appl. Sci. 2026, 16(10), 5162; https://doi.org/10.3390/app16105162 - 21 May 2026

Abstract

Robotics optimization is essential for improving the performance, efficiency, and reliability of robotic systems, especially when dealing with complex engineering problems involving multiple conflicting objectives. Algorithm-specific parameter-free metaheuristic algorithms have gained attention in such applications because they eliminate the need for problem-specific parameter [...] Read more.

Robotics optimization is essential for improving the performance, efficiency, and reliability of robotic systems, especially when dealing with complex engineering problems involving multiple conflicting objectives. Algorithm-specific parameter-free metaheuristic algorithms have gained attention in such applications because they eliminate the need for problem-specific parameter tuning. However, their performance can be further enhanced by improving convergence and maintaining solution diversity in multi-objective optimization. This paper proposes three multi-objective variants—archive, opposition, and self-adaptive multi-population (SAMP)—for the algorithm-specific parameter-free BxR algorithms such as Best–Mean–Random (BMR), Best–Worst–Random (BWR), and Best–Mean–Worst–Random (BMWR). The proposed variants are evaluated on five robotic optimization problems spanning two to six objectives, including Autonomous Underwater Vehicle shape optimization, power line inspection robot design, inverse kinematics of a 4-DOF manipulator, wall-building robot trajectory planning, and optimization of a reconfigurable parallel cutting and grinding mechanism. Their performance is compared with several established multi-objective algorithms using metrics such as GD, IGD, SPC, and HV, supported by rigorous statistical testing involving Friedman tests, Conover post hoc analysis with Holm correction, and Vargha–Delaney A₁₂ effect sizes over 30 independent runs. The results show that archive variants achieve the best IGD rank in four of the five case studies and the best HV rank in three of them, with the five-objective trajectory planning problem being the sole exception where SAMP and base BxR variants show improved IGD performance. The base BxR algorithms prove to be strong competitors, consistently outperforming established parameter-dependent methods on IGD across all five problems. The opposition variants do not provide consistent improvement; however, they also do not cause catastrophic degradation, suggesting that refined opposition strategies warrant further investigation. The study demonstrates the effectiveness of the proposed algorithms as practical optimization tools for complex robotic optimization problems. Full article

(This article belongs to the Section Mechanical Engineering)

20 pages, 7106 KB

Open AccessArticle

Six-Degrees-of-Freedom Compensation for Microphone Array Installation Uncertainty in Rotating Sound Source Localization

by Cheng Wei Lee and Wei Ma

Appl. Sci. 2026, 16(10), 5161; https://doi.org/10.3390/app16105161 - 21 May 2026

Abstract

Accurate rotating sound source localization requires precise knowledge of the relative pose between the microphone array and the rotating object. In practice, six-degrees-of-freedom (6-DOF) installation uncertainties often arise, degrading beamforming performance. This paper presents a 6-DOF compensation method formulated as an optimization problem. [...] Read more.

Accurate rotating sound source localization requires precise knowledge of the relative pose between the microphone array and the rotating object. In practice, six-degrees-of-freedom (6-DOF) installation uncertainties often arise, degrading beamforming performance. This paper presents a 6-DOF compensation method formulated as an optimization problem. The method treats the array as a rigid body and estimates its translational and rotational offsets by minimizing the sum of Euclidean distances between beamforming peaks and a small set of stationary single-frequency reference sources. A novel cascaded local-Bayesian optimization (CLBO) algorithm is proposed, initialized via coordinate descent local search followed by Bayesian optimization refinement. Our simulations show that CLBO yields the lowest residual error and requires the fewest evaluations, outperforming Bayesian optimization, simulated annealing, and local search. Mode composition beamforming (MCB) maps confirm that 6-DOF compensation restores spatial resolution and dynamic range to near-ideal levels, even under perturbed reference-source positions. Our experimental validation on a UAV rotor confirms practical feasibility, restoring focal peaks at blade tips. The proposed approach requires no external metrology, it is robust in practice, and it offers an efficient solution to array installation uncertainty. Full article

(This article belongs to the Special Issue Sound and Vibration: Measurement, Perception, and Control)

► Show Figures

Figure 1

57 pages, 1310 KB

Open AccessReview

On the Challenges and Opportunities of Fuzzing via Large Language Models: A Review

by Yiqing Sun, Villads H. Andersen, Gaurav Choudhary and Nicola Dragoni

Appl. Sci. 2026, 16(10), 5160; https://doi.org/10.3390/app16105160 - 21 May 2026

Abstract

Fuzzing is an important automated testing technique for discovering vulnerabilities and abnormal software behavior, but conventional and pre-LLM learning-based approaches often struggle with strict validity constraints, limited semantic understanding, and poor adaptability across targets. Recent advances in large language models (LLMs) create new [...] Read more.

Fuzzing is an important automated testing technique for discovering vulnerabilities and abnormal software behavior, but conventional and pre-LLM learning-based approaches often struggle with strict validity constraints, limited semantic understanding, and poor adaptability across targets. Recent advances in large language models (LLMs) create new opportunities for fuzzing by supplying semantic guidance from source code, documentation, traces, and execution feedback. As the literature on LLM-based fuzzing has grown rapidly, a systematic synthesis is needed. This paper presents a literature review on the integration of LLMs into fuzzing workflows. We review how LLMs are applied across the fuzzing workflow, compare these approaches with earlier deep learning-based fuzzing, and summarize the main patterns in task coverage, application strategies, targets, testing modes, and model choices. The survey shows that current work is concentrated in test case generation, while defect detection and post-processing remain less represented, and broader workflow integration is uneven. Based on these findings, we identify four main directions for future research as follows: broader coverage of the fuzzing workflow, stronger context construction and workflow engineering, richer reasoning for fuzzing control, and shared evaluation standards and benchmarks. Full article

(This article belongs to the Section Electrical, Electronics and Communications Engineering)

40 pages, 1920 KB

Open AccessArticle

A Generative AI-Driven Predictive Analytics Framework for Modelling Creativity and Performance in Engineering Design Systems

by Kavita Behara and Puramanathan Naidoo

Appl. Sci. 2026, 16(10), 5159; https://doi.org/10.3390/app16105159 - 21 May 2026

Abstract

Engineering education is increasingly shifting toward data-driven and creativity-centred pedagogies that foster innovation, communication, ethical awareness, and teamwork. However, traditional Problem-Based Learning and Design Thinking approaches rely heavily on subjective evaluation and lack scalable mechanisms for monitoring learning progression and creativity development. These [...] Read more.

Engineering education is increasingly shifting toward data-driven and creativity-centred pedagogies that foster innovation, communication, ethical awareness, and teamwork. However, traditional Problem-Based Learning and Design Thinking approaches rely heavily on subjective evaluation and lack scalable mechanisms for monitoring learning progression and creativity development. These pedagogical limitations highlight the need for data-driven approaches that can support iterative learning processes, continuous feedback, and objective evaluation of creativity and performance. This study proposes a Generative Artificial Intelligence (GenAI)-driven predictive analytics framework for modelling student performance and creativity in engineering design systems. The framework integrates deep learning architectures, including Long Short-Term Memory (LSTM) networks and Transformer-based multimodal fusion, to analyze temporal and heterogeneous learning data. The novel Creativity Index (CI) is introduced to quantify design innovation by combining novelty and feasibility metrics derived from AI-assisted interactions and project milestones. The model was evaluated on a longitudinal dataset comprising 450 students across 10 semesters (~5400 time-series observations). Experimental results demonstrate strong predictive performance, achieving 89% classification accuracy and RMSE of 3.8. Comparative analysis shows significant improvements in engineering design (+15%), communication (+16%), ethical awareness (+17%), and teamwork (+16%) (p < 0.01). The proposed framework enables real-time feedback, early risk detection, and adaptive learning optimization. These findings highlight the potential of integrating generative AI and predictive analytics to develop scalable, data-driven intelligent learning systems. Full article

(This article belongs to the Special Issue Applications of Artificial Intelligence Technologies for Education)

16 pages, 542 KB

Open AccessArticle

Regenerative Endodontic Treatment with Calcium Hydroxide-Assisted Blood Clot Stabilisation in Traumatised Immature Teeth: A Retrospective Study of Clinical and Radiographic Outcomes

by Petra Bučević Sojčić, Domagoj Glavina, Jakov Stojanović, Jelena Bagarić, Dubravka Turjanski, Tomislav Škrinjarić, Kristina Goršeta and Hrvoje Jurić

Appl. Sci. 2026, 16(10), 5158; https://doi.org/10.3390/app16105158 - 21 May 2026

Abstract

Traumatised immature permanent teeth with pulp necrosis pose a clinical challenge due to incomplete root development and limited treatment options. This retrospective observational study evaluated the clinical and radiographic outcomes of regenerative endodontic treatment (RET) incorporating an additional calcium hydroxide-based step for blood [...] Read more.

Traumatised immature permanent teeth with pulp necrosis pose a clinical challenge due to incomplete root development and limited treatment options. This retrospective observational study evaluated the clinical and radiographic outcomes of regenerative endodontic treatment (RET) incorporating an additional calcium hydroxide-based step for blood clot stabilisation compared with apexification in traumatised immature teeth. A total of 49 teeth treated between 2015 and 2025 were analysed. Clinical and radiographic parameters were recorded at baseline and reassessed at 6- and 12-month follow-up, including root length, dentinal wall thickness, apical diameter, periapical status, and apical closure. RET was associated with significantly greater increases in dentinal wall thickness (p = 0.002), root length (p = 0.010), and reductions in apical diameter (p < 0.001) over time compared with apexification. Apical closure was observed significantly more frequently in the RET group (p < 0.001). Periapical healing improved in both groups, while overall clinical outcomes appeared broadly similar. Within the limitations of this retrospective observational study, RET incorporating an additional calcium hydroxide-based step for blood clot stabilisation appears to be a clinically applicable approach for promoting root development in traumatised immature teeth. Further prospective and controlled studies are required to confirm these findings and to better define the role of this adjunctive step. Full article

(This article belongs to the Special Issue Recent Advancements in Novel Dental Materials)

29 pages, 1788 KB

Open AccessArticle

Pomegranate Peel Powder as a Functional Ingredient in Yogurt: Effects on Bioactive Profile, Physicochemical Properties, Microstructure, Texture, and Sensory Quality

by Andra Dorina Șuler, Roxana Nicoleta Rațu, Florina Stoica, Petru Marian Cârlescu, Andreea Bianca Balint, Ioana Cristina Crivei, Ionuț Dumitru Velescu, Iuliana Motrescu, Florin Daniel Lipsa and Gabriela Râpeanu

Appl. Sci. 2026, 16(10), 5157; https://doi.org/10.3390/app16105157 - 21 May 2026

Abstract

Pomegranate peel, an agro-industrial by-product, is a promising source of functional compounds. This study evaluated pomegranate peel powder (PP) as a multifunctional yogurt ingredient and assessed its effects on the phytochemical profile, antioxidant activity, physicochemical properties, color, texture, microstructure, mineral composition, storage stability, [...] Read more.

Pomegranate peel, an agro-industrial by-product, is a promising source of functional compounds. This study evaluated pomegranate peel powder (PP) as a multifunctional yogurt ingredient and assessed its effects on the phytochemical profile, antioxidant activity, physicochemical properties, color, texture, microstructure, mineral composition, storage stability, and sensory acceptability. Yogurts supplemented with 3% and 6% PP were compared with a control. PP contained 12.49 mg GAE/g dw total polyphenols, 9.16 mg CE/g dw flavonoids, 63.66 mg C3G/100 g dw anthocyanins, 17.48% dietary fiber, 341.88 mg/100 g calcium, and 140.99 mg/100 g magnesium. PP addition improved yogurt functionality in a concentration-dependent manner. The 6% formulation showed the highest total polyphenol content (9.71 mg GAE/g dw), antioxidant activity (63.67 µmol TE/g dw), dry matter (19.20 g/100 g), and dietary fiber (1.19 g/100 g). Syneresis decreased from 18.22% in the control to 12.17% and 9.22% in the 3% and 6% PP yogurts, respectively, while firmness increased from 3.85 N to 4.80 N. After 21 days of refrigerated storage, fortified yogurts retained high phytochemical and antioxidant levels. Although the 6% formulation provided greater enrichment, the 3% yogurt offered the best balance between functionality, technological performance, and sensory quality, supporting PP valorization in cleaner-label dairy products. Full article

(This article belongs to the Special Issue Functional Foods and Active Natural Products)

26 pages, 3759 KB

Open AccessArticle

Prediction-Regularized Spatio-Temporal Transformer Framework for Offline Multi-Intersection Traffic Signal Control

by Yueting Deng, Huale Li, Tong Xia, Zhaobin Wang and Ruoming Lei

Appl. Sci. 2026, 16(10), 5156; https://doi.org/10.3390/app16105156 - 21 May 2026

Abstract

Multi-intersection traffic signal control must jointly address local coordination and delayed traffic propagation under strongly time-varying conditions. Existing offline sequence-imitation methods mainly recover actions from historical trajectories and make limited use of short-term future traffic evolution in shared-representation learning. To address this issue, [...] Read more.

Multi-intersection traffic signal control must jointly address local coordination and delayed traffic propagation under strongly time-varying conditions. Existing offline sequence-imitation methods mainly recover actions from historical trajectories and make limited use of short-term future traffic evolution in shared-representation learning. To address this issue, we propose PR-STLight, a prediction-regularized spatio-temporal extension of TransformerLight for offline multi-intersection traffic signal control. PR-STLight introduces short-term future inbound-queue evolution as structural supervision for shared representation learning. The model combines neighborhood-constrained spatial self-attention, causal temporal self-attention, and a Topology-Recurrent Queue Predictor (TRQP) to capture topology-aware spatio-temporal dependencies and near-future congestion dynamics. Training adopts a two-stage strategy, namely queue-prediction pretraining followed by joint control-prediction optimization, to improve optimization stability on a fixed offline replay buffer. In experiments on the adopted CityFlow benchmarks, PR-STLight obtains average travel times of 274.39 s on Jinan

3 \times 4

and 288.09 s on Hangzhou

4 \times 4

, corresponding to 1.14% and 2.82% lower travel times than the strongest non-PR baseline, and 21.27% and 22.54% lower travel time than the TransformerLight backbone, respectively. It also achieves the lowest average inbound queue on Hangzhou and remains competitive on Jinan. These results show that PR-STLight provides an effective offline spatio-temporal sequence framework for coordinated multi-intersection signal control. Full article

(This article belongs to the Special Issue Advances in Intelligent Decision-Making Systems)

► Show Figures

Figure 1

17 pages, 22288 KB

Open AccessArticle

Surface Defect Detection of Copper Tube Based on YOLOX with Convolutional Block Attention and Adaptive Spatial Feature Fusion

by Jianjun He and Ji Wang

Appl. Sci. 2026, 16(10), 5155; https://doi.org/10.3390/app16105155 - 21 May 2026

Abstract

Surface defect detection technology is important to improve product quality and save production costs. In order to realize automatic detection of copper tube surface defects, an improved YOLOX algorithm is proposed based on convolutional block attention (CBA) and adaptive spatial feature fusion (ASFF), [...] Read more.

Surface defect detection technology is important to improve product quality and save production costs. In order to realize automatic detection of copper tube surface defects, an improved YOLOX algorithm is proposed based on convolutional block attention (CBA) and adaptive spatial feature fusion (ASFF), named as CBA-ASFF-YOLOX. The improved YOLOX backbone feature extraction network is replaced with CSPDarknet-53. Then, we construct a convolutional block attention module and an adaptively spatial feature fusion module in the feature fusion part to enhance the spatial position correlation between features by learning the connections between different feature maps. To solve the unbalanced differences in real label samples, we utilize Focal Loss function to replace the cross-entropy loss function. In addition, the bounding box regression loss in this study is based on the SIoU formulation, and a simplified variant is adopted to improve regression quality and facilitate more stable convergence. Finally, the algorithm is applied to copper tube surface defect detection task. Experimental results show that the accuracy of CBA-ASFF-YOLOX algorithm is higher than those of other series of YOLO algorithm. Full article

► Show Figures

Figure 1

32 pages, 6072 KB

Open AccessArticle

Assessing Urban Vulnerability Through a Multi-Hazard Framework with Independent Events Modelling

by Glenda Mascheri, Nicola Chieffo, Cláudia Pinto and Paulo B. Lourenço

Appl. Sci. 2026, 16(10), 5154; https://doi.org/10.3390/app16105154 - 21 May 2026

Abstract

Natural hazards and their negative impacts on assets are increasing because of a variety of causes, including climate change, population expansion, and urbanization. Moreover, several areas are susceptible to multiple hazards that interact spatially and/or temporally, necessitating a multi-hazard assessment to adequately mitigate [...] Read more.

Natural hazards and their negative impacts on assets are increasing because of a variety of causes, including climate change, population expansion, and urbanization. Moreover, several areas are susceptible to multiple hazards that interact spatially and/or temporally, necessitating a multi-hazard assessment to adequately mitigate their effects. The goal of this study is to investigate the direct monetary losses produced by the simultaneous interaction of two independent hazards in Lisbon’s city centre, i.e., earthquake and pluvial flood. Seismic hazard has been assessed in terms of macro-seismic intensity, while flood scenario allows for the prediction of water depth for different return periods through a hydrologic-hydraulic model in HEC-RAS software. The seismic and flood vulnerability of the urban investigated compound was evaluated through MCDM methodology—specifically, AHP and TOPSIS methods. A framework for multi-hazard analysis was subsequently developed, explicitly accounting for the interaction between the two hazards and their joint occurrence probabilities based on historical data from the case study area. The results demonstrate that multi-hazard losses are 108 M€ for a 2-year return period and 232 M€ for a 475/500-year scenario, emphasizing that floods contribute more across all return periods in the research area; however, for longer return periods, the earthquake contribution increases significantly. Full article

(This article belongs to the Special Issue Digital Multi-Hazard Risk Modelling and Life-Cycle Assessment for Next-Generation Resilient and Sustainable Built Environments)

► Show Figures

Figure 1

31 pages, 3022 KB

Open AccessArticle

The Assessment and Enhancement of the Efficiency and Dynamic Responses of LSTM-Guided Yaw-Controlled NREL 5 MW Wind Turbines Subjected to Irregular Waves vs. İzmir–Samos Tsunami Waveform

by Barış Namlı, Cihan Bayındır, Azmi Ali Altintas and Fatih Ozaydin

Appl. Sci. 2026, 16(10), 5153; https://doi.org/10.3390/app16105153 - 21 May 2026

Abstract

Wind energy is one of the clean and sustainable energy sources that can be used to meet global energy demand. However, although wind turbines are used to harness energy in coastal and offshore areas, the extreme environmental conditions make it difficult to utilize [...] Read more.

Wind energy is one of the clean and sustainable energy sources that can be used to meet global energy demand. However, although wind turbines are used to harness energy in coastal and offshore areas, the extreme environmental conditions make it difficult to utilize these structures optimally and also negatively impact their safety. Therefore, in this study, the yaw angles of the National Renewable Energy Laboratory (NREL) 5 MW wind turbine mounted on a monopile platform were controlled using a Long Short-Term Memory (LSTM) artificial intelligence (AI) architecture to optimize power output and ensure structural stability against dynamic responses under irregular wave and Izmir–Samos tsunami conditions. First, the İzmir–Samos tsunami, the NREL 5 MW wind turbine mounted on a monopile platform, the analytical methods employed, the LSTM architecture, and the parameters used in the study were described. The results of the wind direction time series prediction and the aerodynamic, hydrodynamic, and structural responses of the LSTM-based yaw angle control strategy on the wind turbines were investigated and discussed. According to the results, the increase in aerodynamic power achieved using the LSTM-based strategy was approximately

5.18 %

under a scenario with constant wind speed and variable wind direction in two different sea conditions and

5.29 %

under conditions of irregular waves with fully variable wind speed and direction. In addition, more stable responses were observed for most of the parameters examined. The primary goal of this study is to serve as a reference for researchers working on AI-optimized power output and response analysis resulting from varying the yaw angles of wind turbines. Full article

(This article belongs to the Special Issue Advanced Wind Turbine Control and Optimization)

22 pages, 1053 KB

Open AccessArticle

GRAG4PM: Graph Retrieval Augmented Generation Framework Adapted for Process Mining

by Xiaohan Su, Bin Liang, Zhidong Li, Yifei Dong, Justin Wang and Fang Chen

Appl. Sci. 2026, 16(10), 5152; https://doi.org/10.3390/app16105152 - 21 May 2026

Abstract

Recent advancements in generative AI have improved process mining, making workflow analysis more accessible and scalable. However, large language models lack structured reasoning and fail to capture sequential dependencies in workflows, limiting their effectiveness. While retrieval-augmented generation (RAG) improves contextual knowledge integration, it [...] Read more.

Recent advancements in generative AI have improved process mining, making workflow analysis more accessible and scalable. However, large language models lack structured reasoning and fail to capture sequential dependencies in workflows, limiting their effectiveness. While retrieval-augmented generation (RAG) improves contextual knowledge integration, it does not enforce process execution constraints, leading to inconsistencies in workflow modeling. To address this, we propose GRAG4PM, a graph retrieval-augmented generation framework designed for process mining. GRAG4PM introduces three key innovations: a hierarchical graph architecture that captures multi-level process semantics, an adaptive pruning mechanism that refines workflow representations while preserving critical information, and a process-specific dual-indexing scheme that adapts existing hybrid retrieval to the constraints of workflow graphs. Experimental results demonstrate that GRAG4PM significantly improves workflow validation, anomaly detection, and predictive analysis, particularly in handling incomplete information. Across the three evaluated process-aware tasks, workflow-state validation, violation checking, and next-activity prediction, GRAG4PM achieves an average Macro-F1 score of 0.7259, outperforming Graph RAG (0.4394) and standard RAG (0.4036). Under missing-attribute settings, GRAG4PM also maintains the highest average Macro-F1 across masking conditions, confirming its robustness in incomplete data scenarios. By balancing structured workflow constraints with flexible AI-driven retrieval, GRAG4PM enables dynamic, interpretable, and adaptive process mining solutions and explores promising directions for future research on improving workflow solutions for AI. Full article

(This article belongs to the Section Computing and Artificial Intelligence)

► Show Figures

Figure 1

19 pages, 7468 KB

Open AccessArticle

Mechanical Failure of a Bottom Hole Assembly During Composite Plug Milling Operations: A Field Case Study

by Przemysław Toczek, Rafał Wiśniowski, Albert Złotkowski, Krzysztof Pańcikiewicz, Filip Matachowski and Jacek Adamiak

Appl. Sci. 2026, 16(10), 5151; https://doi.org/10.3390/app16105151 - 21 May 2026

Abstract

This paper presents a field case study of a mechanical failure that occurred in the bottom-hole assembly (BHA) during composite plug milling after hydraulic fracturing operations. The failure sequence was reconstructed using field hook load and torque records, operational documentation, and inspection of [...] Read more.

This paper presents a field case study of a mechanical failure that occurred in the bottom-hole assembly (BHA) during composite plug milling after hydraulic fracturing operations. The failure sequence was reconstructed using field hook load and torque records, operational documentation, and inspection of the damaged components recovered from the borehole. The results indicate that the critical condition developed progressively and was associated with increasing resistance to drill string movement, insufficient hole cleaning, and repeated attempts to continue milling and release the partially immobilized assembly. The observed damage pattern, together with the presence of residual cuttings and metallic debris in the borehole, supports the conclusion that the loss of the BHA section at the hydraulic safety sub resulted from the interaction of several adverse operational factors acting simultaneously, particularly the combined action of pull-up force and rotation under deteriorating borehole conditions. A supporting strength assessment of the hydraulic safety sub was used to relate characteristic operating points to the admissible working range of the connector. The study shows that hook load and torque data provide the greatest practical value when interpreted jointly and in their operational context rather than as isolated peak values. The findings support safer planning and execution of plug-milling and stuck-pipe remediation operations in highly deviated wells. Full article

(This article belongs to the Special Issue Drilling Engineering Technologies: Fluid Systems, Automation, and Process Optimization)

► Show Figures

Figure 1

31 pages, 9062 KB

Open AccessArticle

Periodicity of FEM Discrete Models and Its Influence on Solutions to the 1-D Wave Equation

by Wiktor Waszkowiak, Łukasz Doliński, Paweł Kowalski and Arkadiusz Żak

Appl. Sci. 2026, 16(10), 5150; https://doi.org/10.3390/app16105150 - 21 May 2026

Abstract

This paper discusses the influence of the periodicity of finite-element (FE) discrete models and its influence on solutions to the one-dimensional (1-D) wave equation. Numerical solutions to wave-propagation problems obtained via the displacement-based formulations of the finite-element method (FEM) often exhibit high-frequency behavior, [...] Read more.

This paper discusses the influence of the periodicity of finite-element (FE) discrete models and its influence on solutions to the one-dimensional (1-D) wave equation. Numerical solutions to wave-propagation problems obtained via the displacement-based formulations of the finite-element method (FEM) often exhibit high-frequency behavior, which is frequently dismissed in the literature as undesired, spurious, and/or having no physical meaning. In this paper, we verify this notion by demonstrating that this behavior is not merely a computational anomaly but is due to the inherent periodic properties of discrete numerical models. Using Bloch’s theorem, we reveal and demonstrate how, at high frequencies, the discrete nature of FEM numerical models leads to the prevailing behavior governed by the periodic nature of the computational models. In order to illustrate this phenomenon, we investigate 1-D wave propagation in rods, leveraging the non-dispersive nature of the governing equation as a benchmark. In addition to the classical and specialized FEM, we analyze two alternative formulations: the time-domain spectral finite-element method (TD-SFEM) and a novel spline-based finite-element method (spFEM) proposed by the authors. The results obtained and presented explain qualitatively the origins of these numerical anomalies and suggest strategies to mitigate their effects, effectively shifting the periodicity-induced behavior beyond the range of physically relevant frequencies by appropriate selection of approximation polynomials. The authors demonstrate that this can be fully achieved only in the case of spFEM, for which the usable percentage of the available spectra of eigenfrequencies reaches 67%, while in the case of other FEM approaches discussed is significantly smaller as determined by numerical dispersion and the presence of frequency band gaps. Full article

(This article belongs to the Section Mechanical Engineering)

► Show Figures

Figure 1

Journal Description

Applied Sciences

Latest Articles

Journal Menu

Journal Browser

Highly Accessed Articles

Latest Books

E-Mail Alert

News

Topics

Conferences

Special Issues

Topical Collections

Further Information

Guidelines

MDPI Initiatives

Follow MDPI