Search Results (1,682)

Designers traditionally perform the preliminary sizing of foundations based on engineering judgement, relying on parameters such as superstructure loads and the characteristics of the supporting soil. This process must adhere to strict guidelines for wind turbine foundations to ensure structural stability and compliance with regulatory standards. This study proposes a computational model based on genetic algorithms to optimise the dimensions of wind turbine foundations. The fitness function combines two normalised objectives, namely concrete volume and bending moment, while structural and geotechnical requirements are imposed as constraints. The model was validated using six real-world case studies, achieving consistent reductions in concrete volume compared with the original designs, with an average reduction of 19.5%. Each case was run 10 times to assess the consistency of the solutions obtained. The results demonstrate the effectiveness of the proposed approach in identifying more material-efficient foundation geometries while satisfying the adopted design constraints. It should be emphasised that the reported savings refer specifically to concrete volume reduction and should not be interpreted as total foundation cost savings, since reinforcement design and detailing are outside the scope of the present model. This study is restricted to foundations with circular cross-sections, thereby opening avenues for future research aimed at extending the optimisation framework to alternative geometric configurations and incorporating reinforcement design. Full article

Cross-view geo-localization (CVGL) between unmanned aerial vehicle (UAV) imagery and satellite imagery is a key technique for autonomous UAV navigation in Global Navigation Satellite System (GNSS)-denied environments. However, most existing methods rely on energy-intensive Artificial Neural Networks (ANNs), making them difficult to deploy on resource-constrained edge computing platforms. Spiking Neural Networks (SNNs) provide a promising alternative for energy-efficient inference, but their application to CVGL still faces two challenges that remain insufficiently addressed. First, the isotropic computation used by existing SNN backbones is mismatched with the directional characteristics of spike activations. Spike activations tend to form oriented aggregation patterns along elongated geographic structures, and isotropic computation can therefore dilute directional signals. Second, the limited representational capacity of SNNs further increases the sensitivity during training optimization. However, the standard triplet loss adopts a static weighting strategy and assigns the same weight to all triplets that violate the margin constraint, which is unfavorable for learning from hard negatives. To address these challenges, we propose a framework with two core contributions. At the feature extraction level, the Directional Adaptive Convolution Module (DACM) processes spike feature maps by sequentially performing horizontal strip convolution and vertical strip convolution, thereby capturing a more complete geometric structure of directional spike clusters. At the training supervision level, we propose a Dual-dimensional Progressive Reweighting (DPR) loss, which jointly characterizes sample difficulty from pairwise difficulty and positive-pair quality difficulty. A learnable fusion parameter is used to adaptively balance these two types of difficulty information. Experimental results on the University-1652 and SUES-200 benchmarks show that the proposed framework, when equipped with the same representation learning head as its ANN counterparts, achieves competitive and, in many settings, superior performance. In terms of energy efficiency, its estimated theoretical energy consumption is over $8.8\times$ lower than that of published ANN methods under their original configurations. Under a more rigorous matched ANN control that shares the identical architecture, the estimated energy is reduced from 29.84 mJ to 6.36 mJ, an approximately $4.7\times$ reduction obtained at a cost of only 2.29 percentage points in R@1. Full article

Urban stormwater poses a threat to urban safety and development. The systematic spatial planning of low-impact development (LID) facilities is increasingly recognized as a sustainable approach to enhancing urban flood resilience. Previous studies have largely focused on empirically-based design approaches or have employed uncoupled computational methods, resulting in a lack of accuracy in flood simulation results. In this study, a novel framework was proposed. The Non-dominated Sorting Genetic Algorithm II was employed to perform multi-objective optimization of the spatial layout of LID facilities, and a coupled model of SWMM and TELEMAC was developed to simulate surface flooding based on the optimized schemes. Under three rainfall scenarios, three schemes on the Pareto Front—representing the lowest cost, the optimal compromise and the least overflow—were selected to investigate how scheme parameters and overflow are influenced by rainfall intensity and design preferences. The results indicate that scheme parameters and node overflow show greater variation under the influence of different design preferences than under different rainfall conditions. Taking the schemes selected in this study as examples, under the lowest-cost scheme, LID coverage was less than 10%, resulting in a limited reduction in node overflow; but when ‘minimum overflow’ was set as the design preference, node overflow was virtually eliminated, with peak water levels at flood-prone locations reduced to 0.05 m, 0.08 m and 0.10 m under 50-year, 100-year and 200-year storm conditions, respectively. The framework for urban flooding simulation and flood control scheme design proposed in this study is potentially applicable to comparable settings, subject to similar data availability and physical conditions, serving as a reference for enhancing urban resilience to flooding. Full article

With the widespread adoption of cloud computing technology, modern cloud platforms have become increasingly complex and dynamic, posing significant challenges for efficient resource management. Accurate forecasting of cloud resource loads has therefore become essential for improving service quality, optimizing resource utilization, and reducing operational costs. To address the intrinsic characteristics of cloud load time series, including nonlinear fluctuations, multi-scale temporal dependencies, and redundant high-dimensional features, this paper proposes the MST-iTransformer model, which integrates multi-scale temporal encoding, sparse attention, and adaptive feature selection mechanisms. Specifically, a multi-scale temporal encoding module is developed to capture and fuse temporal dependencies across multiple periodic scales. Furthermore, an adaptive feature selection module is introduced to dynamically assign importance weights to resource features, enhancing informative variables while suppressing redundant ones. Meanwhile, a sparse attention mechanism is incorporated to reduce computational overhead while maintaining forecasting accuracy. The proposed model is evaluated on the Alibaba Cluster Trace dataset. Experimental results demonstrate that MST-iTransformer achieves MSE, RMSE, and MAE values of 0.5559, 0.7456, and 0.4968, respectively. Compared with the original iTransformer, the proposed model achieves simultaneous reductions in prediction errors and inference latency, validating the effectiveness of the multi-scale temporal encoding, sparse attention mechanism, and adaptive feature selection modules in improving forecasting accuracy and computational efficiency. These improvements provide reliable prediction support for resource scheduling and elastic scaling in cloud data centers. Full article

Engine noise is a predominant source of noise in the cabin of internal combustion engine vehicles and new energy hybrid vehicles. The conventional multi-channel active noise control (ANC) system, based on the adaptive notch filtered-X least mean square algorithm, is commonly employed to mitigate such multi-tonal noise. However, the computational efficiency and convergence performance of this system may be significantly hindered by the large estimated secondary path length and the frequency-dependent convergence behavior. To overcome these limitations, this paper proposes a computationally efficient and fast-converging multi-channel ANC system by incorporating a local secondary path (LSP) equalization method. The proposed method enhances the convergence speed by equalizing the magnitude responses of estimated secondary paths and reduces the computational complexity through an improved LSP modeling approach. Accordingly, a set of low-order equalized LSP models with normalized amplitude-frequency responses is generated and employed for reference filtering. A computational complexity analysis comparing the conventional system, a recent cost-effective system, and the proposed system is presented. Numerical simulations are conducted to evaluate the convergence speed and noise attenuation performance of these three systems. Additionally, real vehicle experiments are performed using a digital signal processing controller. The results demonstrate that the proposed multi-channel ANC system achieves a superior noise reduction effect. Under accelerated conditions, the average attenuation of the second-order noise component at the four error microphones is measured at 4.4 dB(A), 6.2 dB(A), 13.4 dB(A), and 10.0 dB(A). These findings confirm the practical effectiveness of the proposed multi-channel ANC system. Full article

In recent years, interest in artificial intelligence has grown significantly, particularly in the development of advanced computational models for supply chain decision-making. Inventory management is one of the areas in which intelligent algorithms can support demand forecasting, replenishment, stock control, and operational optimization under uncertainty. This study presents an updated systematic literature review of intelligent algorithms applied to inventory management. The review followed PRISMA 2020 guidelines and combined database searches in Scopus, ScienceDirect, Web of Science, IEEE Xplore, SpringerLink, Taylor & Francis, and complementary manual searching. The original search covering January 2020 to December 2024 was updated in July 2026 to include studies published or available online up to June 2026. After applying strict eligibility criteria, 37 primary studies with quantitative evidence were included. The updated corpus confirms the predominance of deep learning, reinforcement learning, and hybrid intelligent models, while also showing the recent emergence of Transformer-based, graph neural network, multi-agent reinforcement learning, and prescriptive analytics approaches. The most frequent application areas were inventory control, inventory optimization, replenishment decision-making, and demand forecasting. Reported improvements were mainly associated with cost efficiency, service level, stockout reduction, and system performance; however, the magnitude of improvement varied across algorithms, data sources, sectors, and simulation or real-world settings. Overall, intelligent algorithms represent a relevant tool for improving inventory management, but their adoption requires careful validation, transparent reporting, and alignment with the operational context. Full article

Background/Objectives: Free-text veterinary pathology diagnoses contain essential information for cancer registration but are difficult to convert into standardized ontology-based codes because of linguistic variability, contextual modifiers, and large ontology search spaces. This study evaluated a hybrid lexical–semantic architecture for the automated assignment of Vet-ICD-O-Canine-1 morphology codes. Methods: A retrospective single-registry benchmark included 211 diagnoses from the São Paulo Animal Cancer Registry. Of these, 190 contained sufficient morphological information for expert-reviewed reference coding, whereas 21 generic or insufficiently specified descriptions were retained as an exploratory challenge subset. Fuzzy lexical matching retrieved Top-10, Top-20, or Top-30 candidates from the complete 971-entry morphology ontology, followed by semantic selection using Claude Haiku 4.5 and structured JSON output. Performance and computational efficiency were compared to direct full-ontology inference. Results: Among the evaluated fuzzy metrics, token_set_ratio achieved the highest Top-30 reference-code retrieval rate of 89.5%. End-to-end exact-match agreement increased from 73.7% with Top-10 to 79.5% with Top-20 and 85.8% with Top-30 (95% CI, 80.1–90.0%). Top-30 generated non-null codes for 93.2% of the 190 evaluable diagnoses and achieved a conditional exact-match agreement of 92.1%. By contrast, the direct full-ontology baseline achieved 71.2% conditional exact-match agreement (42/59) among non-null predictions and 22.1% end-to-end exact-match agreement (42/190) when incorrect predictions, null outputs, and technical failures were considered non-concordant outcomes. Compared to direct full-ontology inference, Top-30 reduced input-token consumption by 92.4%, total token consumption by 92.2%, and inference cost by 91.3%, while avoiding the 118 API rate-limit failures observed with the direct baseline. Among the 21 insufficiently specified diagnoses, Top-30 returned null codes in 38.1% and non-null codes in 61.9%. Conclusions: Ontology-guided candidate reduction improved coding agreement, computational efficiency, and operational robustness within this retrospective single-registry benchmark. However, the reported performance estimates require confirmation in larger independent datasets, and an upstream data-sufficiency or abstention mechanism is needed before prospective operational deployment. Full article

Computational fluid dynamics methods range from Reynolds-Averaged Navier–Stokes (RANS) simulations to Direct Numerical Simulations (DNSs). RANS offers low computational cost at the expense of accuracy, while DNS provides high accuracy but at a prohibitive cost. The aim of this study is to evaluate whether multi-fidelity neural networks can learn a corrective mapping from RANS to DNS for a small canonical dataset and to determine how training-set composition and model architecture influence generalisation across geometries. In this study, multi-fidelity neural networks for data fusion between low-fidelity RANS and high-fidelity DNS data were applied to turbulent flow (Re = 5600) over parameterised periodic hills, defined by a geometry parameter characterising the steepness ratio. The inputs to the models were the coordinates and the corresponding RANS velocity components, and the outputs were the DNS velocity components, with data from both fidelities mapped onto the same mesh. Both a single-branch and a two-branch architecture were considered. Generalisability was assessed within a small dataset of five periodic hills defined by different values of the geometry parameter $\alpha$ (0.5, 0.8, 1.0, 1.2, 1.5). Both model architectures were trained on data from different combinations of the geometry parameter to evaluate interpolation and extrapolation capabilities. Both networks successfully corrected RANS flow fields for unseen geometries in interpolation regimes. When interpolating, the single-branch architecture achieved more than a 69% reduction in error, while the two-branch architecture achieved more than a 60% reduction, with both improving key flow features such as recirculation zones and jet structures. A key finding is that the single-branch architecture consistently outperformed the two-branch formulation, particularly in low-data regimes. The results show that multi-fidelity neural networks can improve RANS predictions using small datasets and simple inputs, provided that the training set spans the relevant geometric space. As the model does not require the geometry parameter as an explicit input, it is applicable to geometries lacking straightforward parameterisation. The demonstrated advantage of the single-branch architecture highlights the importance of architectural simplicity when training data is limited. Full article

Hyperspectral image classification relies heavily on the effective modeling of spatial–spectral representations. Recent deep learning architectures, including Transformers and state space models (SSMs), have shown promise for HSI classification. However, under few-shot scenarios, they may suffer from optimization instability in early-stage feature reduction, weakened local spatial–spectral correlations after direct sequence flattening, and attenuation of center-pixel spectral information caused by deep spatial aggregation. To mitigate these issues, we propose orthogonal projection spatial–spectral Mamba (O-Mamba), a lightweight architecture for few-shot HSI classification. First, we introduce a task-driven orthogonal projection module (TOPM) for learnable end-to-end spectral dimensionality reduction. In this module, orthogonal parameterization, supervised initialization, and an auxiliary loss jointly improve the stability of the projection process, reducing feature redundancy and mitigating representation collapse. Second, we design a 3D spatial–spectral Mamba encoder that employs 3D Convolutional Neural Networks (CNN) as local tokenizers to preserve local spatial–spectral structures and then uses Mamba to capture long-range sequence dependencies with linear complexity with respect to sequence length. Finally, to alleviate over-smoothing in the target-pixel representation, we propose a decoupled target–context fusion strategy. This mechanism separately preserves the original spectral signature of the center pixel and fuses it with high-level contextual features, which may improve the separability of spectrally similar classes. Extensive experiments on four benchmark datasets show that O-Mamba achieves competitive classification performance under the evaluated few-shot settings, while maintaining a relatively small model size and low computational cost compared with representative CNN, Transformer, and Mamba-based methods. Full article

Skin cancer represents a growing global health burden. Current diagnostic pathways combine clinical examination and dermoscopy with histopathological confirmation; however, overlap between benign and malignant lesions can create diagnostic uncertainty and lead to potentially avoidable biopsies. Electrical impedance spectroscopy (EIS) has emerged as a non-invasive technique with potential for portable and cost-efficient implementation that quantifies the dielectric contrast between malignant and healthy tissue, providing objective information that may support clinical decision-making. This review synthesizes the field across four levels. First, we describe the biophysical origins of the impedance contrast in skin cancer, spanning the cellular, tissue architecture, and molecular scales, together with the equivalent circuit and Cole–Cole frameworks used to interpret it. Second, we examine hardware advances, including electrode–skin interface strategies, flexible and wearable architectures, computational electrode design, and the translation from laboratory prototypes to commercial systems such as Nevisense. Third, we critically appraise clinical evidence from large multicenter trials, focusing on the sensitivity–specificity trade-off and the demonstrated reduction in the number needed to excise. Finally, we discuss emerging frontiers, including artificial intelligence-driven analysis and multimodal fusion with dermoscopy, reflectance confocal microscopy, optical coherence tomography, and near-infrared spectroscopy. We conclude that EIS is most valuable as a complementary component within an integrated, AI-supported multimodal diagnostic framework. Full article

This paper presents a comparative study of two approaches for solving second-order ordinary differential equations arising from the membrane theory of shells of revolution. The considered equations originate from the rotational symmetry of shell structures, which enables the reduction of the governing partial differential equations to a sequence of ordinary differential equations corresponding to individual circumferential harmonics. The study compares the classical Differential Transform Method (DTM) with Physics-Informed Neural Networks (PINNs). Both initial value and boundary value problems are investigated, including benchmark examples with known analytical solutions and a systematic analysis of the influence of PINN architecture on the solution accuracy. For the PINN approach, the effects of the number of collocation points, hidden layers, and neurons per layer on the approximation error and training time are examined. The results demonstrate that DTM provides an efficient framework for constructing analytical solutions of initial value problems with minimal computational cost. However, its application to boundary value problems requires the introduction of additional auxiliary parameters and the solution of supplementary nonlinear equations, considerably increasing the analytical complexity of the procedure. In contrast, PINNs achieve high accuracy for both initial and boundary value problems while naturally incorporating boundary conditions through the loss function. The presented results demonstrate how the exploitation of geometric symmetry, combined with modern scientific machine learning techniques, provides an effective computational framework for solving differential equations arising in shell mechanics. Full article

The large-scale integration of photovoltaic generation into distribution grids has introduced significant operational challenges, including voltage excursions, reverse power flows, and increased variability. Battery energy storage systems (BESSs) offer a versatile solution by providing coordinated active- and reactive-power support. However, their scheduling in active distribution networks is challenging because of the non-convex alternating-current (AC) power-flow equations, the nondifferentiability of battery-degradation modeling, and uncertainty in renewable generation and demand. This paper proposes a two-stage methodology for the day-ahead operation of BESSs in ADNs. In the first stage, parallel particle swarm optimization (PPSO) determines the hourly active- and reactive-power schedules of the BESS units. In the second stage, a matrix-based multi-period AC power flow based on successive approximations evaluates the schedules and verifies voltage, thermal, converter-capability, and state-of-charge (SoC) constraints. A rainflow-counting degradation model is incorporated into the objective function to account for cycling and calendar aging costs. The methodology is assessed through ablation analyses comparing active-power-only and coordinated P–Q dispatches, degradation-unaware and degradation-aware scheduling, and serial and parallel PSO implementations. It is validated on modified 33-, 69-, and 136-node systems under deterministic and uncertainty-based operating conditions, including 100 demand and PV-generation scenarios. PPSO is compared with parallel versions of the adaptive Jaya algorithm (AJAYA), genetic algorithm (GA), multi-verse optimizer (MVO), salp swarm algorithm (SSA), grey wolf optimizer (GWO), and vortex search algorithm (VSA), using operating-cost reduction, computational time, solution variability, feasibility indicators, BESS lifetime, and weekly cost analysis. Additionally, exact one-sided Wilcoxon signed-rank tests with Holm adjustment are used to assess the statistical significance of the economic differences between PPSO and the benchmark methods. Results show that PPSO provides the lowest or most competitive operating costs and the shortest computational time in the evaluated cases, while all network and storage constraints remain satisfied. Full article

Topology optimization has been widely applied to dielectric graded insulation design in gas-insulated switchgear (GIS); however, the selection of optimization parameters remains challenging due to strong coupling among design variables and the high computational cost of conventional parametric scanning methods. To address this issue, a hybrid optimization framework integrating Latin hypercube sampling (LHS) and response surface methodology (RSM) is proposed for efficient parameter selection in dielectric gradient topology design. The proposed framework combines global parameter space exploration, parameter space reduction, and multi-stage response surface optimization to construct surrogate models for efficient parameter optimization. The results show that the maximum electric field of the optimized insulator is reduced from 3.336 kV/mm to 1.400 kV/mm, demonstrating the effectiveness of the proposed method in improving electric field uniformity and optimization efficiency. Full article

Land-based hyperspectral imaging provides high spatial and spectral resolution for detecting camouflaged targets, but practical deployment remains limited by strong target background spectral similarity, scarce annotated hyperspectral samples, and the computational cost of full-band processing. To address these issues, this paper proposes HCTDNet (Hyperspectral Camouflaged Target Detection Network), a land-based hyperspectral image analysis framework. The method first employs band extraction for data dimensionality reduction, compressing multi-channel hyperspectral images into 3-channel virtual RGB representations, which reduces spectral redundancy while preliminarily enhancing camouflaged target saliency. A pre-trained RGB camouflaged target detector is then adopted as the backbone model, with its parameters frozen to maintain stability, while trainable modality-specific prompts are learned to improve training efficiency. Finally, model fine-tuning is performed using a self-constructed camouflaged target dataset to enhance robustness in detecting camouflaged targets within virtual RGB images. During inference, preprocessed hyperspectral images are fed into the model to generate detection results for camouflaged target regions. The experiments performed on our self-collected land-based hyperspectral dataset with camouflaged targets reveal that HCTDNet achieves superior detection performance compared with seven classical hyperspectral target detection methods while maintaining an average inference speed of approximately 16 FPS. The proposed framework provides an efficient and near-real-time applicable solution for land-based hyperspectral camouflaged target detection, showing significant practical potential. Full article

Graph-based nonnegative matrix factorization (GNMF) has been widely used for dimensionality reduction and data clustering because it can preserve the intrinsic geometric structure of data. However, many existing GNMF-based methods still rely on full sample similarity graphs, resulting in high computational costs; moreover, their sparsity constraints usually treat all features uniformly, making it difficult to distinguish structurally important features from redundant or noisy ones. To address these issues, this paper proposes a feature-graph-guided adaptive Log-L_2,1 sparse NMF with anchor dual graphs under a logarithmic framework. Specifically, anchor-based representations are simultaneously constructed in the sample and feature spaces to approximate the corresponding full-scale graphs. The sample anchor graph preserves the local manifold structure among samples, whereas the feature anchor graph plays a dual role: it preserves structural relationships among features and provides degree information for generating the adaptive weights g_i of the row-wise Log-L_2,1 penalty imposed on the basis matrix U. Consequently, structurally well-connected features receive weaker sparsity penalties, while weakly connected and potentially redundant features are more strongly suppressed. In addition, a logarithmic reconstruction framework is introduced to reduce the influence of large residuals caused by noise and outliers. These mechanisms jointly integrate sample structure preservation, feature structure preservation, and feature-aware sparse learning within a unified graph-NMF model. To optimize the model, multiplicative update rules are derived, while the nonnegativity of the factor matrices is inherently preserved throughout the iterations. Extensive evaluations on several benchmark datasets demonstrate the effectiveness and robustness of the proposed method. Full article