Browse Articles
Discover research articles across all indexed journals
An intelligent SCADA-integrated deep learning framework for bird-safe offshore wind farm operation
Abstract The rapid expansion of offshore wind energy has intensified concerns regarding avian collisions with turbine blades, particularly for migratory and high-risk bird species. Conventional mitigation approaches—including radar monitoring, manual intervention, and acoustic deterrents—are often limited by high false alarm rates, delayed response times, and the lack of species-level identification. To address these challenges, this study proposes an intelligent framework that integrates a Supervisory Control and Data Acquisition (SCADA) system with a Deep Convolutional Neural Network (DCNN)-based Bird Detection and Classification (BDC) model. The proposed system performs automated image-based bird classification and translates detection outputs into SCADA-driven turbine control actions through a multi-zone proximity assessment strategy. The model is trained and evaluated on a dataset comprising 525 avian species with over 90,000 images. Comparative analysis against conventional classifiers—including Support Vector Machines (SVM), Random Forest, K-Nearest Neighbor, and VGG16—demonstrates that the proposed BDC model achieves superior performance, with an accuracy of 99.62%, precision of 99.92%, recall of 100%, and an F1-score of 99.93%. In addition to classification performance, the system demonstrates a simulation-based system, achieving inference latency below 30 ms and SCADA response execution within 40 ms. These results highlight the potential of integrating deep learning with operational control systems to enable automated, risk-aware turbine response mechanisms for wildlife protection. It is important to note that the evaluation is conducted under controlled dataset conditions, and the dataset does not fully represent real offshore environments characterized by long-distance detection, motion blur, occlusion, and complex backgrounds. Therefore, the reported performance should be interpreted as an upper-bound estimate, and future validation using real-world offshore data is required to confirm deployment robustness. Overall, the proposed framework provides a simulation-based proof-of-concept approach for bridging AI-based avian monitoring with SCADA-enabled turbine control, contributing toward environmentally sustainable offshore wind farm operation.
Koumine inhibits osteoclastogenesis and prevents ovariectomy-induced bone loss via suppression of the MAPK signaling pathways
Abstract Excessive osteoclast formation drives osteolytic bone diseases such as osteoporosis. Koumine (KM), an alkaloid derived from Gelsemium elegans, exhibits various bioactivities; however, its role in bone homeostasis remains unknown. This study investigated the effects of KM on RANKL-induced osteoclastogenesis in bone marrow-derived macrophages (BMMs). Cell viability, differentiation (TRAcP staining), and function (F-actin ring formation, bone resorption pit assay) were assessed. The underlying mechanisms were explored using Western blot and qPCR. The in vivo efficacy of KM was evaluated in an ovariectomized (OVX) mouse model using micro-CT and histological analyses. KM dose-dependently inhibited osteoclast formation and bone resorption without cytotoxicity. It suppressed RANKL-induced activation of the mitogen-activated protein kinase (MAPK) pathway, downregulating c-Fos, NFATc1, TRAP, and CTSK. In contrast, KM had no significant effect on RANKL-induced NF-κB activation. KM did not impair osteoblast differentiation or mineralization. In vivo, KM treatment prevented OVX-induced bone loss, improved trabecular microarchitecture, and reduced osteoclast numbers. KM suppresses osteoclastogenesis and protects against bone loss, an effect associated with inhibition of MAPK signaling, highlighting its potential as a novel therapeutic for osteolytic diseases.
Whole genome sequencing-based multi-locus association mapping for kernel iron, zinc and protein content in groundnut
Abstract Malnutrition is a major global challenge, especially in the developing regions, where improving the nutritional content of staple crops is an important step towards alleviating hidden hunger. Groundnut, a nutrient rich legume, contains several essential nutrients, high protein, essential amino-acids and vitamins required for human health. In this study, multi-season phenotyping data for kernel iron (Fe), zinc (Zn) and protein content (PC), along with whole genome re-sequencing (WGRS) data from a mini-core collection, were used to perform genome-wide association study (GWAS) analysis. Phenotypic variability analysis revealed a large variation in Fe (7.6 – 42.8 ppm), Zn (10.9 – 62.4 ppm) and PC (12.7 – 33.6%). GWAS analysis identified a total of 15 marker-trait associations (MTAs) and 28 candidate genes for pooled season data, and 44 MTAs and 62 candidate genes for individual seasons. Key candidate genes like MYB transcription factor ( Arahy.QI0PHV , Arahy.1I6ZSS ), Zn finger MYM type protein , RING finger MYM type protein ( Arahy.7P97F6 , Arahy.9R964H , Arahy.I3B88T ) and NAC domain protein ( Arahy.LV3APC ), were found to be associated with the Fe and Zn homeostasis pathway. In addition, genes related to protein homeostasis, such as Protein kinase family protein ( Arahy.4D7KBI ), and E3 ubiquitin-protein ligase ( Arahy.PE3CF6 ), were identified within significant MTA regions. These findings provide basis for the detection and characterization of potential candidate genes associated with nutritional quality traits. Haplo-pheno analysis revealed that accessions with superior haplotypes for kernel Fe and Zn content were predominantly found in the Spanish Bunch/Valencia Bunch types, whereas the accessions with inferior haplotypes were more common in the Virginia Bunch/Virginia Runner types. Single nucleotide polymorphism (SNP)-based KASP (Kompetitive Allele Specific Polymerase Chain Reaction) markers for 9 MTAs were designed and validated for Fe and Zn. Among these, three markers (snpAH00636 and snpAH00641 for Fe, and snpAH00644 for Zn) showed polymorphism, and could be utilized in genomics-assisted breeding to develop nutrient-rich groundnut varieties.
Lattice-Embedded Single-Atom Sr–O–Ni Channels Enable Photon–Phonon Coupling for Photothermal N <sub>2</sub> O Decomposition
Efficient material selection for training occluded mmWave radar-based gesture recognisers
Abstract Radar-based gesture recognition has emerged as a promising approach for unobtrusive interaction. Unlike camera-based systems, radar sensors can detect gestures through opaque materials, enabling seamless embedding into various everyday objects. However, it remains unclear how to train models efficiently for robust gesture recognition through diverse materials. To investigate this, we collected a dataset of 17,520 gesture recordings performed through 73 everyday materials. By comparing several material-sampling and data-augmentation strategies, we found that a small carefully selected representative subset of training materials was sufficient to match the performance of a classifier trained on the full material dataset. Our results showed that the models trained on 14 quota-sampled materials achieved accuracies of 95.8% and 91.2%, comparable to training on all 73 materials (96.8% and 91.6%) and significantly better than training without material data (66.8% and 65.8%). Among the evaluated sampling approaches, Quota sampling also provided the best overall trade-off between performance and practicality. In contrast, classifiers trained on augmented data performed worse than those trained on actual material-specific data. Taken together, these findings indicate that, for the tested sensor, gesture set, and material collection, carefully selected real-material data offer a practical route to reducing material-specific data collection in radar-based gesture recognition while preserving generalisation. Code, models, and data are available in the public repository, with additional details provided in the supplementary materials: https://gitlab.com/hicuplab/seeing-through .
Detection of recurrence of HPV-driven oropharyngeal cancer by HPV cell-free DNA
Abstract The incidence of oropharyngeal cancer (OPC) driven by human papillomavirus (HPV) is increasing in high-income countries. Recurrent tumors remain a major cause of HPV-OPC mortality, yet reliable biomarkers for post-treatment monitoring are lacking. This study evaluated HPV cell-free DNA (cfDNA) from blood plasma as a monitoring biomarker for the detection of HPV-OPC recurrence. Blood plasma from 59 OPC patients was collected at diagnosis and during follow-up. HPV cfDNA was quantified using a multiplex digital PCR assay targeting HPV16 and seven other high-risk types. Tumor HPV status was confirmed by p16 INK4A immunohistochemistry, HPV DNA PCR, and E6 serology. At diagnosis, HPV cfDNA was detected with 95% sensitivity (36/38; 95% CI 82–99%) and 95% specificity (19/20; 95% CI 75–100%). For 24 HPV-OPC patients, follow-up samples were available (median follow-up 1.5 years), with four patients experiencing recurrence or persistence of HPV-OPC. In two patients, HPV cfDNA detection after initial clearance preceded clinical recurrence by 3 and 8 months, respectively. The third patient experienced recurrence 7 months after collection of the last blood sample, which was HPV cfDNA-negative. The fourth patient had persistent disease without HPV cfDNA detection. HPV cfDNA positivity after therapy had a positive predictive value of 75% for HPV-OPC recurrence within one year on a per-test basis. HPV cfDNA has been shown to have great potential as a minimally invasive monitoring biomarker for HPV-OPC surveillance. Monitoring HPV cfDNA levels may identify patients at risk of recurrence early, thereby potentially facilitating timely intervention. Studies with more participants are needed to establish a solid evidence base for surveillance protocols that incorporate HPV cfDNA.
Enhancing operational decision-making in hydrocarbon exploration drilling using machine learning for gas data interpretation
Abstract Artificial intelligence is increasingly used to support decision-making during hydrocarbon exploration drilling, but mud-gas interpretation remains challenging because gas signatures are influenced by mud properties, drilling parameters, degassing efficiency, and Drill Bit Metamorphism (DBM). Here, we present a machine-learning-assisted workflow for assessing how well reservoir-fluid signals are represented in Advanced Gas (AG) measurements acquired while drilling. A multi-domain dataset from 104 Brazilian exploration wells was quality controlled, harmonized, and integrated with laboratory pressure-volume-temperature (PVT) fluid compositions and expert geological interpretation. Two predictive products were developed: Reservoir Affinity Curves, which estimate the similarity between AG signatures and reference PVT fluids using C2- and C2C-based targets, and a DBM Severity Curve, which quantifies drilling-induced thermal alteration using ethylene-related behavior and operational variables. Kernel Ridge Regression was selected as the primary deployment model because it produced stable, smooth, and interpretable depth-dependent predictions, whereas XGBoost and LightGBM achieved the highest numerical accuracy as benchmark models. The workflow distinguished intervals dominated by representative formation-fluid signatures from zones affected by DBM or other operational artifacts. This approach supports earlier fluid characterization, improves fluid-sampling decisions, and reduces interpretation uncertainty before laboratory results become available.
A graph-integrated reinforcement learning framework with graph neural networks for tactical decision modeling in professional football
Staged screw removal in external plate fixation for distal tibial fractures: a finite element study on adaptive biomechanics
SAKCL: a deep neural network test data selection method based on self-attention and K-means clustering
Abstract DNNs, similar to traditional software systems, may exhibit defects that can lead to serious consequences, especially in safety-critical scenarios. As a result, the ability to detect such defects reliably has become increasingly important, where the quality of the test dataset plays a central role. In this work, we introduce a test data selection method which combines a self-attention mechanism with K-means clustering (i.e., SAKCL) in a coordinated manner. The self-attention component assigns different levels of importance to feature dimensions, helping highlight informative patterns within samples while reducing the effect of redundant information. Based on these refined features, K-means clustering is applied to organize the data and capture structural relationships across samples. Through this process, the selected test cases achieve a better balance between fault detection capability and diversity. Experiments conducted on four benchmark datasets (MNIST, CIFAR-10, Fashion-MNIST, and SVHN) and multiple DNN architectures (LeNet-1/5, ResNet-20, and VGG-16) show that SAKCL consistently performs better than existing methods. On average, it increases the proportion of error-revealing test cases ( FDR ) by more than 12%, improves diversity-related metrics ( KMNC ) by over 6.5%, and leads to a retraining accuracy improvement ( ΔAcc ) of + 3.319% compared with baseline approaches. Statistical analysis further supports the reliability of these improvements ( p < 0.01, Cliff’s δ > 0.8). Overall, the proposed method provides a practical and scalable solution for selecting effective test data in DNN quality assurance.
Comparing trapezoidal and circular configurations to evaluate the impact of shell geometry on thermal energy storage performance
CASCADENCE: a layered cascade defense mechanism for federated learning
Abstract This paper aims to enhance the security and robustness of Federated Learning (FL) systems through a multi-layered defense. We address the critical challenge of protecting distributed learning environments from adversarial attacks while maintaining high model performance during both the training and operational phases. The proposed framework is based on integrated approaches that utilize a Gaussian filter with Discrete Fourier Transform (DFT), adversarial training with differential privacy, JPEG compression, randomized smoothing, and adversarial logit pairing. It integrates multiple defense mechanisms based on system requirements, focusing on preserving model performance while ensuring robust protection during both training and testing phases. Our approach extends beyond existing solutions by introducing various staged defense implementations and analyzing their synergistic effects. Experimental results demonstrate that the proposed ensemble defense mechanism achieves the highest performance, maintaining 98.21% accuracy and an F1 score of 0.98 under attack conditions, compared to a baseline accuracy of 90.87%.
Performance evaluation of grid-forming battery energy storage systems for stability enhancement in solar PV plants
Benchmarking hybrid CNN and transformer backbones with graph convolution networks (GCN) for flower growth-stage classification
Investigating bioactive potential of the Serratia sp. CS01 and its pigment for anti-inflammatory, antioxidant and antibacterial action
Ischemic injury and liver graft congestion increase post-transplant interleukin-6 and tumor necrosis factor-alpha in hepatocellular carcinoma
Hydrodynamic evaluation of a spider-inspired underwater robot using distributed flapping fin propulsion
Abstract This paper presents the design and hydrodynamic evaluation of a spider-inspired underwater hexapod robot employing distributed flapping-fin propulsion. Unlike conventional bio-inspired underwater robots with centralized fin actuation, the proposed system integrates flexible bilateral side fins between leg triads, enabling decentralized lift generation and inherent body stabilization. A three-tier validation framework is adopted, comprising (i) quasi-steady analytical estimation of lift, drag, Reynolds number, and lift-to-power scaling, (ii) transient Computational Fluid Dynamics (CFD) simulations in ANSYS Fluent resolving pressure distribution, vortex shedding, and wake evolution, and (iii) preliminary experimental validation using a laboratory-scale prototype in controlled water-tank conditions. Theoretical, numerical, and experimental results show close agreement, with lift predictions within ± 10–15% across operating regimes. CFD results indicate stable hydrodynamic performance over fin-tip velocities U = 0.6–0.8 m/s, yielding a nearly constant lift-to-drag ratio of ≈ 1.3. Non-dimensional analysis confirms operation within an efficient Strouhal number range. The study establishes the hydrodynamic feasibility and energetic advantages of distributed flapping-fin propulsion integrated with a multi-legged body architecture. The results provide validated design insights for the development of manoeuvrable and energy-efficient underwater robotic platforms for inspection, monitoring, and exploration tasks.
An intelligent ethereum blockchain technology for pest detection and smart irrigation in IoT using hybrid deep learning model
Abstract This research discusses the incorporation of IoT with blockchain technique to enhance the efficiency of smart farming systems, particularly focusing on plant disease classification, pest detection, and smart irrigation. The study aims to develop a secure and effective IoT-based smart farming framework using the Ethereum blockchain to store and transmit data, and a Hybrid Convolution Adaptive Recurrent MobileNet (HC-ARMNet) model for predictive analytics, optimized by the Improved Secretary Bird Optimization (ISBO) algorithm. The research employs IoT sensors to acquire real-time data, which is then stored in the Ethereum blockchain to ensure security. The HC-ARMNet model, combining 1D/2D convolutions with recurrent connections, processes this data for pest detection and irrigation management. The ISBO algorithm is leveraged to fine-tune the technique’s parameters. Datasets used: The proposed system utilizes three standard datasets for evaluation. The PlantifyDr Dataset is used for classifying plant disease, and the Pest Detection Dataset is used for recognizing pests. Also, for the smart irrigation process, the significant field images are collected manually. The accuracy, precision, and FNR rates of the ISBO-HC-ARMNet-aided plant disease classification are 94.16%, 94.2% and 5.87%. At the same time, the ISBO-HC-ARMNet-based pest detection process’s accuracy, sensitivity, and specificity are 93.78%, 93.79% and 93.76%, respectively. In addition, the ISBO-HC-ARMNet-based smart irrigation task’s MSE is 3.21, SMAPE is 0.03, and MASE is 30.23. Thus, the designed system showcases promising performance over classical approaches in terms of accuracy and error rates for plant disease classification, pest detection, and smart irrigation. The research concludes that the IoT-aided smart farming framework with blockchain and the HC-ARMNet model provides a robust solution for secure and efficient agricultural management. The system’s predictive capabilities provide accurate and timely data analysis, facilitating to the improvement of precision agriculture. Future work will focus on improving the system with advanced feature extraction strategies to reduce processing time.