Search Results (403)

Construction material descriptions in engineering projects are often heterogeneous due to inconsistent naming conventions, supplier-specific terminology, and incomplete specification information, making it difficult to automatically map non-standard descriptions to standardized material dictionaries. Existing rule-based approaches rely heavily on manually constructed rules and predefined dictionaries, limiting their adaptability, whereas direct large language model (LLM)-based methods often lack domain-specific constraints, resulting in less reliable material classification and attribute decomposition. To address these challenges, this paper proposes LLM-MSMA, an LLM-based multi-agent framework for construction material classification and attribute decomposition. The framework adopts a multi-scale task decomposition strategy that integrates hybrid retrieval, logic validation, and template-guided attribute decomposition. Hybrid retrieval generates candidate material categories, which are subsequently refined through LLM-based semantic reasoning. A Logic Validation Agent further verifies classification results using engineering rules, while an Attribute Decomposition Agent extracts structured material attributes according to classification-specific templates. The proposed framework is evaluated using a real-world bill-of-materials dataset containing 2912 material records collected from a large-scale engineering project as a case study. Experimental results show that, compared with the rule-based baseline, LLM-MSMA improves material classification accuracy by 43.39 percentage points and increases the attribute decomposition F1-score by 68.89 percentage points. Compared with the direct LLM baseline, the proposed framework further improves material classification accuracy by 8.24 percentage points and the attribute decomposition F1-score by 7.31 percentage points, while achieving an Exact Match Accuracy of 82.83%. These results indicate that combining hybrid retrieval with domain-specific structured knowledge and multi-agent LLM reasoning improves the reliability of construction material classification and attribute decomposition for heterogeneous material descriptions within the evaluated engineering context. Full article

Fine-scale mapping of forest types is a prerequisite for accurately assessing forest biomass, biodiversity, ecosystem service values, and carbon budgets. Shihe County, located within an ecotone transitioning from subtropical to temperate monsoon climates in China, was selected as the case study area. By integrating Sentinel-1 SAR and Sentinel-2 multispectral imagery, along with derived vegetation indices, texture features, and backscattering coefficients, as well as statistical features extracted from the 2022 NDVI time-series, we employed a hierarchical classification framework and a random forest algorithm to generate the forest type map. The results indicated the following: (1) With a single-date multi-feature dataset, late winter (3 March) was determined to be the optimal period for forest type classification in Shihe County. (2) Shortwave infrared bands, red-edge bands, the modified vegetation index, the normalized difference red-edge index, mean texture features, and VH-polarization data were the most influential variables. (3) Incorporating yearly NDVI time-series statistical features into the single-date winter subset significantly improved classification performance, yielding an overall accuracy of 86.39% and a Kappa of 0.781, which represents a 10.83% improvement over the baseline. Persistent misclassification between bamboo forests and tea plantations remained a primary constraint on further accuracy enhancement, while the classification accuracy for evergreen broadleaf forest and deciduous coniferous forest exhibited considerable uncertainty, likely attributable to limited reference sample sizes. (4) Deciduous broadleaf forests constituted the dominant land cover type in Shihe County, whereas evergreen coniferous forests, bamboo, and tea plantations were also widely distributed, collectively reflecting the region’s transitional ecological characteristics. Full article

We present ESNformer, a hybrid architecture that couples an Echo State Network (ESN) reservoir with a Transformer encoder for classification of structured, multi-indicator assessment data: a fixed-order vector of complementary indicators per assessment instance rather than a repeated-measures time series. The reservoir acts as a fixed nonlinear feature map over the indicator vector, while self-attention, made position-aware over the fixed column order, learns how each indicator’s evidence contributes to the final decision, so the two components, together, capture local, indicator-level detail and global, cross-indicator interactions within a single, end-to-end trainable model. Interpretability is treated as a first-class design requirement rather than an afterthought: the architecture is paired with an explainability layer combining SHAP feature attribution (reported both globally and per class), the model’s own attention weights, a deletion/insertion faithfulness test that quantitatively verifies which inputs the model actually relies on, and counterfactual maps that translate a prediction into an actionable, inspectable recommendation. We evaluate the architecture on a concrete case study, classifying Braille-literacy instructional recommendations from 15 pedagogical indicators grouped into three categories (Mangold’s, ABKL, and Progresar), using a benchmark of 900 real assessment instances (630 used, together with a class-conditional augmentation procedure, to build a 2100-instance training set) with validation and test partitions (135 instances each) kept exclusively real. On this benchmark, the tuned model reached 85.33% accuracy, 85.90% macro-precision, 85.33% macro-recall, an F1 score of 85.25%, and an AUC of 0.95 on the real test set. SHAP attribution, attention weights, and the faithfulness test converge on the same two dominant indicators (response time and error count): removing them alone collapses accuracy to chance, while retaining only them recovers most of the model’s accuracy. We report this transparently alongside a comparison against ESN-only, Transformer-only, and tabular baselines (logistic regression, decision tree, random forest, XGBoost, and an MLP) on the same data and discuss what the hybrid architecture and its explainability pipeline add beyond what the two dominant indicators already explain and how the approach generalizes to other tabular and mixed-granularity assessment settings that require both predictive accuracy and a verifiable account of what drove each decision. Full article

Air-operated double-diaphragm (AODD) pumps in industrial sludge transfer suffer from gradual performance degradation due to rheological variations and component wear, yet conventional monitoring relies on scarce labeled fault data. This paper presents an unsupervised multi-sensor framework that requires no fault labels, integrating physics-informed dual health indices, HI-P (sludge load) and HI-V (mechanical stress), with a Gaussian Mixture Model anomaly detector and a physics residual attribution module. Governing equations motivate the use of these indices from five sensors: inlet and outlet flow meters (100 Hz), an air pressure transducer (100 Hz), and inlet and outlet accelerometers (1652 Hz). Trained on one healthy baseline day (86,218 one-second windows), the Gaussian Mixture Model achieves 100% day-level classification performance on the evaluated dataset (F1 = 1.00) across 455,201 test windows from nine operating days, with window-level receiver operating characteristic area under the curve (ROC-AUC) = 0.8580 and precision–recall AUC (PR-AUC) = 0.9082. Residual attribution analytically confirms that pressure residuals drive Episode 1 (HI-P peak 3.63 times baseline, Cohen’s d = 1.70) and vibration residuals drive Episode 2 (HI-V peak 5.44 times the baseline, d = 4.10), providing empirical support for the proposed physics-informed formulation without requiring fault labels. Comparisons with four unsupervised benchmarks confirm that this is the only approach that simultaneously enables label-free operation, physics-driven features, exact attribution, real-world deployment, and perfect day-level F1. Full article

Carbonate formations host abundant hydrocarbon resources and constitute a key target for current and future petroleum explorations. A massive Paleocene-Eocene carbonate platform system overlies Late Cretaceous Deccan volcanic rocks within the Indus Fan Offshore Basin. Low geophysical exploration maturity and insufficient targeted research on carbonate reef identification and platform evolution hinder hydrocarbon discoveries in this region. Using 2D seismic data, this study employs multiple seismic attribute methods to characterize carbonate reef-related anomalies. An innovative reef identification method based on a weighted superposition of spectrally decomposed amplitude spectra is proposed to delineate reef-prone sedimentary facies and characterize the multi-stage evolutionary patterns of the carbonate platforms. Based on quantitative constraints from seismic reflection and amplitude differences, platform evolution is classified into four clear phases: initial development, platform expansion, gradual decline and platform drowning. The initial development phase shows prominent high-amplitude and low-frequency seismic responses, and intervals with such high-amplitude responses may indicate favorable reservoir potential. Platform margin zones show the most prominent amplitude anomalies and are considered priority exploration targets, suggesting excellent hydrocarbon potential within strata formed during the initial platform stage. This study provides quantitative geophysical support for the stage classification of Paleocene–Eocene carbonate platforms in the Indus Fan Offshore Basin, and offers practical guidance for carbonate reservoir exploration and the prospect evaluation of analogous basins worldwide. Full article

Marine Spatial Planning (MSP) is a key instrument for sustainable ocean governance in Europe, yet the spatial distribution of socioeconomic benefits and trade-offs remains poorly documented. This study applies and extends the Spatial Economic Benefit Analysis (SEBA) framework to offshore wind energy and capture fisheries in France, with a focus on the North Atlantic–Western Channel (NAMO) façade. Three extensions are introduced relative to earlier single-sector, single-scale SEBA applications: simultaneous dual-sector comparison, nesting of project-level firm mapping within regional analysis, and delivery through an open, updatable visualization platform that allows planners to re-run the analysis as registries are updated. Twelve indicator families were selected against four explicit criteria (relevance to a SEBA step, spatial attributability, traceability to a planning instrument, and public availability) and cross-validated against independent registries. Offshore wind employment is more unevenly distributed across regions than fishery revenue (Gini 0.59 versus 0.42), and firm presence is a weak proxy for employment: Brittany hosts 26.6% of mapped offshore wind firms but only 6.5% of sectoral full-time equivalents (2.1 FTE per firm), whereas Île-de-France hosts 4.8% of firms and 19.6% of employment (34.6 FTE per firm). Fishery revenue and offshore wind firm counts are positively rank-correlated across coastal regions (Spearman ρ = 0.90, n = 5, p = 0.037), indicating co-location rather than causation: offshore wind constitutes an additional spatial pressure on a sector already contracting (−9.5% employment 2012–2022; −29% vessels 2000–2023). Uncertainty was addressed through a three-tier data confidence classification and sensitivity testing; concentration rankings were robust to extreme reassignments of multi-role firms (ΔGini ≤ 0.06). The analysis is deliberately restricted to the distribution of economic benefits and does not assess ecological impacts, which must be integrated before SEBA outputs inform zoning. Outputs are directly interpretable for façade plan revision, tender local content scoring, and compensation design, and the workflow is transferable to any jurisdiction with public corporate registries and landings statistics. Full article

Large Language Models (LLMs) show considerable potential for materials-science question answering. However, LLM responses may still be affected by unsupported parametric associations, while dense Retrieval-Augmented Generation (RAG) can fragment relational evidence across text chunks. Moreover, general graph-based retrieval does not necessarily preserve the hierarchical relations and factual attributes required to resolve implicit material constraints. To address these limitations, we propose MCTD-KG, a multi-source heterogeneous knowledge graph integrated with a Knowledge-Enhanced RAG framework for complex material question answering. MCTD-KG adopts a Classification–Term–Data ontology to connect disciplinary taxonomies, domain concepts, semantic relations, and empirical records from toolbooks and the scientific literature. Through LLM-assisted knowledge extraction, entity normalization, and multi-source integration, the resulting graph contains more than 530,000 entities across three layers, including 61,768 text-extracted Term-layer entities. During inference, Dual-Channel Retrieval jointly retrieves query-relevant relational paths and associated material attributes, while an explicit semantic filtering stage screens candidate evidence against the query constraints. Evaluation on an expert-validated benchmark of 1577 questions shows that the proposed framework achieves an overall accuracy of 68.48%, compared with 17.40% for the zero-shot Pure LLM, 24.79% for the best Vanilla RAG setting, and 44.96% for GraphRAG. It also achieves 45.22% accuracy on four-hop questions, compared with 39.49% for GraphRAG. These results indicate that integrating multi-source domain knowledge with relation-preserved retrieval and attribute-supported filtering provides more focused and inspectable evidence, thereby supporting more accurate complex material question answering. Full article

This study investigates the damage evolution mechanism affecting the fracture-bearing performance of fly ash cement paste under low-temperature curing conditions. Pre-cut cement paste specimens with fly ash contents of 0%, 15%, and 25% were prepared and subjected to both standard curing and low-temperature curing at 5 °C for 28 days. Three-point bending tests combined with acoustic emission (AE) monitoring were conducted to analyze peak flexural load, AE ring count, cumulative energy, RA-AF crack classification, and b-value evolution. Additionally, scanning electron microscopy (SEM) and thermogravimetric analysis (TGA) were employed to characterize micromorphology and relative changes in hydration product content. The results indicate that both fly ash incorporation and low-temperature curing significantly reduce the flexural bearing capacity of pre-notched specimens. Under low-temperature curing, the peak loads of LF15 and LF25 decrease by 34.83% and 47.19%, respectively, compared to LF0. At the same fly ash replacement level, all low-temperature-cured specimens exhibited lower peak loads than those cured under standard conditions. Overall AE activity was reduced in low-temperature-cured specimens, with crack propagation instability occurring at lower load levels. The addition of fly ash shifted the fracture mode toward a tensile-dominated type, whereas low-temperature curing increased the proportion of shear-type AE events. Fly ash incorporation increased the relative content of calcium silicate hydrate (C-S-H) gel and decreased that of calcium hydroxide (CH); however, this did not result in improved peak flexural load. This outcome is attributed to the insufficient reactivity of fly ash at low temperatures, leading to residual unreacted spherical particles, dilution of clinker, and inadequate interfacial bonding, which collectively weaken the continuous load-bearing skeleton of the matrix. This paper establishes a multi-scale interpretation of the damage mechanisms affecting the low-temperature fracture-bearing performance of fly ash cement paste by correlating macroscopic bearing response, AE damage evolution, crack types, and hydration product composition. Full article

Fine-grained classification of urban green spaces (UGSs) is important for urban ecological assessment and management but remains challenging because of spectral similarity among vegetation types and inaccurate object delineation in complex urban environments. This study proposes a pixel-to-object framework that combines semantic segmentation-guided object construction with multi-periodic phenological modeling. A semantic green-space mask derived from 0.27 m very-high-resolution imagery constrains superpixel segmentation to generate spatially coherent, boundary-aware green space object-level patches (GSOPs). Pixel-level temporal representations are then derived from Sentinel-2 normalized difference vegetation index (NDVI) time series using TimesNet, aggregated into GSOP-level phenological features, and combined with spatial attributes to classify urban trees, grasslands, and farmlands. Applied to the built-up area of Chengdu, China, the framework achieved an overall accuracy of 91.6%, with F1-scores of 92.5%, 91.9%, and 87.6% for urban trees, grasslands, and farmlands, respectively. Ablation experiments showed that removing phenological features reduced overall accuracy by 13.1 percentage points and decreased the F1-scores of grasslands and farmlands by 16.0 and 23.0 percentage points, respectively. These results demonstrate that semantically constrained object delineation and phenological information jointly reduce boundary fragmentation and improve the discrimination of spectrally similar urban vegetation types. Full article

Urban vegetation provides essential ecosystem services, including temperature regulation, air purification, noise reduction, and carbon storage. However, urban densification and climate-induced stresses increasingly threaten these ecosystems. Accurate classification of urban vegetation is critical for sustainable urban planning, yet remains challenging due to the structural complexity and high data density of urban LiDAR (Light Detection and Ranging) point clouds. To address current research gaps, including insufficient model interpretability, high computational demands, and limited generalization capabilities, we introduce transmissivity, a novel feature that combines echo-based LiDAR properties with spatial context to more effectively characterize urban vegetation structures. This feature enhances vegetation classification by estimating whether laser beams tend to traverse or terminate within a local neighborhood, independent of the specific return order of individual beams, thereby characterizing vegetation’s volumetric permeability. This makes transmissivity highly interpretable, unlike other echo-based statistical features. Transmissivity was evaluated alongside 52 conventional features using three different feature-importance measures across two distinct urban LiDAR datasets from Berlin and Hessigheim 3D (both datasets are from Germany). Transmissivity consistently ranked among the most influential features in both datasets across multiple scales and achieved the highest average gain (Berlin: 0.465; Hessigheim: 0.286) and the second highest mean absolute Shapley value (Berlin: 2.066; Hessigheim: 0.824). Permutation importance confirmed that transmissivity has the strongest impact on the mean decrease in F1-score (Berlin: 0.53; Hessigheim: 0.34) if used in uncorrelated feature subsets. Our findings support that echo-enriched point clouds allow efficient and accurate monitoring of urban vegetation. The feature is simple to compute and does not require additional data beyond standard multi-return LiDAR attributes. Full article

Software-Defined Networking (SDN) separates the control and data planes, introducing a logically centralized controller that is itself a high-value attack target. Despite growing interest in SDN intrusion detection, publicly available datasets either restrict evaluation to binary normal-vs-DDoS classification or lack control-plane telemetry, leaving multi-class detection of SDN-architectural attacks without a dedicated benchmark. This work presents LAN-SDN-NIDS, a publicly available, multi-class flow-level dataset of 1,125,059 records generated in a fully containerized Containernet/OpenDaylight testbed across five standard network topologies. Each flow record combines 29 traffic-level features with 11 control-plane-aware metrics—including Packet-In and Flow-Mod counts and first-seen delay. The dataset covers five attack classes in two categories: three that exploit SDN control-plane mechanisms (link fabrication, host injection, and port hijack) alongside DDoS and port scan, plus normal traffic. An XGBoost classifier trained on the full feature set achieved a macro F1 of 0.94; an ablation study showed that removing OpenFlow features causes link fabrication F1 to collapse from 0.97 to 0.19, indicating that control-plane telemetry is decisive for detecting SDN-architectural attacks under the conditions evaluated. A UMAP embedding is consistent with class separability, except for a structural overlap between host injection and normal traffic attributable to their shared ARP protocol. Full article

Raman spectroscopy combined with machine learning offers a rapid, label-free approach for bacterial identification, but robust translation remains challenged by spectral variability, biological heterogeneity, and limited model interpretability. Here, we present an integrated evaluation of an optimized Spectral Transformer (ST) framework for Raman-based bacterial classification benchmarked against a systematically optimized one-dimensional convolutional neural network (1D-CNN). The comparison was performed using a curated 36-class dataset comprising 15 Gram-negative bacterial entries, 15 Gram-positive bacterial entries, one non-bacterial microorganism, and five background/reference classes, enabling evaluation of both species-level and fine-grained bacterial classification. Under 15 dB noise-augmented evaluation, the ST achieved 80.6% ± 0.3% accuracy and a Matthews correlation coefficient (MCC) of 0.801 ± 0.003, outperforming the 1D-CNN baseline with 72.9% ± 0.3% accuracy and an MCC of 0.721 ± 0.003. Integrated Gradients analysis combined with attention map visualization enabled multi-level model interpretation, revealing that the ST’s improved robustness correlates with more bounded attribution patterns during misclassification, whereas the 1D-CNN’s feature attribution becomes scattered under noise perturbation. Importantly, this interpretability-driven analysis identified model-specific failure modes in the baseline architecture, including an over-reliance on non-specific spectral regions under noise, which can inform future data collection strategies and guide refinements to experimental protocols. These results demonstrate that attention-based spectral modeling improves Raman-based bacterial classification under noise-perturbed conditions while enabling multi-level interpretability that bridges model understanding with actionable feedback on experimental design and data quality requirements. Full article

Customer churn is a critical challenge for subscription-based digital platforms, as customer activity patterns change dynamically over time. Traditional churn modeling methods fail to account for sequential dependencies across customer interaction histories. This paper presents Attn-ChurnNet, a novel attention-based Transformer architecture that effectively predicts customer churn by modeling sequential customer contact histories. The proposed methodology leverages multi-head self-attention with sinusoidal positional encodings and Pre-LN residual connections to highlight key interaction sequences and interpret the temporal dynamics of customer activity in subscription platforms. Experiments are conducted on the large-scale WSDM–KKBox Customer Churn Prediction dataset using a temporal train/validate/test split, incorporating transaction records, usage logs, and customer demographic information. Comprehensive comparison against established baselines—Logistic Regression (LR), Random Forest (RF), XGBoost, and Gated Recurrent Unit (GRU)—demonstrates that Attn-ChurnNet achieves a macro-averaged classification accuracy of 97.03% (±0.41%), precision of 95% (±0.5%), recall of 96% (±0.6%), F1-score of 95.50% (±0.5%), AUC of 0.98 (±0.004), Average Precision of 0.963, and log-loss of 0.15 (±0.007) under five-fold stratified cross-validation, outperforming all competing approaches with statistical significance ($p<0.01$, McNemar’s test). A comprehensive two-part ablation study (22 variants), calibration analysis (ECE = 0.031; $T_s^{*}$ = 1.08), attention entropy analysis with Jensen–Shannon divergence and two-sample t-test ($t=18.4$, $p<0.001$), Integrated Gradients attribution, conformal prediction (91.4% coverage, 88% singleton efficiency), precision–recall analysis, and computational complexity evaluation further validate the model’s design and production readiness. Full article

Due to the explosive growth in cloud-native infrastructures, the attack surface has dramatically increased in modern enterprise identity systems, where privilege escalation has become a major security risk. Conventional rule-based intrusion detection systems fall short in identifying multi-hop privilege inheritance paths and lateral movements over heterogeneous and dynamic identity graphs. This study introduces PEGraphSec-Net, a graph-theoretical framework for detecting privilege-escalation-relevant identity behavior, modeling cloud identity interactions as dynamic heterogeneous graphs of users, services, roles, tokens, and workloads. The core contribution of this framework is a graph-based detection pipeline—an Identity Relationship Graph Constructor, a Privilege-Escalation Path Encoder, and a Temporal Graph Attention Detection layer—evaluated on privilege-escalation-relevant attack categories using a documented proxy identity-graph construction derived from the UNSW-NB15 network-traffic benchmark, and benchmarked against six non-graph tabular classifiers (CNN, LightGBM, XGBoost, Random Forest, SVM, and MLP) trained under identical preprocessing; this pipeline achieves 98.78% accuracy, a weighted F1-score of 0.98692 (macro F1-score of 0.91828), and an AUC of 1.000 on the held-out test partition. PEGraphSec-Net is further benchmarked against three graph neural network baselines (GCN, GAT, and GraphSAGE) trained on the identical identity-graph topology and node attributes; all three substantially underperform PEGraphSec-Net (best case, GraphSAGE: 63.66% accuracy, 0.239 macro F1-score), indicating that a large share of PEGraphSec-Net’s performance derives from its explicit privilege-path encoding and temporal attention mechanisms rather than from the graph topology alone. An Adaptive Containment and Isolation Engine and a Mitigation Policy Reinforcement Optimizer are further proposed as risk-scoring and reward-driven policy-learning components, whose contribution is validated through module-wise ablation on classification performance; live containment action and reinforcement-learning-specific evaluation are left for future validation. The term “privilege escalation” is used throughout to denote the evaluated proxy attack categories (Exploits, Backdoor/Backdoors, and Reconnaissance) under a documented, decade-old (2015) network-intrusion benchmark, rather than production cloud-native IAM behavior, for which native-dataset validation remains an open direction. SHAP-based interpretability analysis links the model’s top-ranked traffic-level features back to the identity-graph risk, role, and trust-transition attributes they populate, evidencing that the learned representation captures semantically meaningful identity-behavior patterns within this proxy setting. Full article

Oats (Avena sativa L.) are an analytically demanding cereal. They carry a lipid fraction of roughly 5–9% together with high lipase and lipoxygenase activity, a protein fraction dominated by globulins, a high-molecular-weight (1→3), (1→4)-β-D-glucan, and avenanthramides that occur in no other grain. Each of these fractions changes during kilning, milling, storage, and fermentation, and each change affects a quality attribute of commercial consequence. Fourier-transform infrared (FT-IR) spectroscopy, interpreted with chemometric models, offers a rapid and non-destructive route to several of these attributes from a single measurement. This review draws the oat-specific literature together. It covers the identification, classification and authentication of oat material, including: the separation of oats from gluten-containing cereals, the discrimination of cultivars and of commercial milled forms, and the grading of grain; the prediction of quality parameters and the compositional analysis of moisture, β-glucan, protein, starch and lipid; the molecular changes that accompany processing and storage—β-glucan depolymerisation, lipid oxidation and hydrolysis, thermally induced protein aggregation, avenanthramide degradation and starch retrogradation—and the spectral features by which sound and deteriorated oat material can be discriminated; and the portable and handheld instruments now bringing the measurement closer to the production line. Reported calibrations are frequently accurate but rest on limited sample sets, and multi-site validation together with shared reference spectra remains the principal requirement before routine industrial use. Full article