Search Results (4,191)

Liquid rocket engines (LREs) are mission-critical propulsion systems whose reliability directly affects launch safety and the repeated operation of reusable launch vehicles. Data-driven LRE fault diagnosis remains difficult because labeled fault samples are scarce, engine-to-engine variations cause distribution shifts, and existing evaluations frequently emphasize classification accuracy without controlling for input representation. This study establishes a reliability-oriented benchmark that evaluates seven representative machine-learning and deep-learning methods from four dimensions: diagnostic capability, sensor informativeness, cross-unit generalization, and data efficiency. The experiments use 1106 fault samples from the XJTU-REF dataset, covering nine fault categories, 19 sensor channels, and three simulated engine units. To ensure an apples-to-apples model comparison, all seven methods are first evaluated using the same 19-dimensional steady-state representation. Under this controlled setting, Random Forest achieves the highest accuracy of $89.15\%$ and Macro-F1 score of $88.74\%$, while the best deep model, 1D-CNN, reaches $86.42\%$. An additional representation-ablation experiment shows that concatenating the same steady-state features with learned sequence representations increases 1D-CNN accuracy to $87.58\%$, reducing its gap from Random Forest to $1.57$ percentage points. This result demonstrates that the large difference observed under the original heterogeneous-input setting is partly attributable to representation design rather than model family alone. In cross-unit evaluation, Random Forest obtains an average accuracy of $75.37\%$, whereas Transformer obtains $61.54\%$, indicating that engine-to-engine variation remains a major deployment challenge. The top five sensors contribute $47.62\%$ of the total Random Forest importance, and, with five labeled samples per fault category, KNN and Random Forest achieve $71.24\%$ and $70.82\%$ accuracy, respectively. These findings provide quantitative guidance for algorithm selection, sensor prioritization, and data-acquisition planning in reusable launch vehicle health-monitoring systems. Full article

Cardiovascular disease diagnosis requires accurate and timely analysis of electrocardiogram (ECG) signals to support reliable clinical decision-making. However, ECG signals are inherently non-stationary, exhibit substantial inter-patient variability, and may share similar morphological patterns across different cardiac disorders, making automated multi-class diagnosis challenging. This study proposes a multi-domain machine learning framework for automated ECG-based cardiac disease classification, integrating signal preprocessing, heartbeat segmentation, Variational Mode Decomposition (VMD), multi-domain feature extraction, minimum Redundancy Maximum Relevance (mRMR) feature selection, and hybrid CNN–Transformer learning. Experiments are conducted on the PTB-XL database using five diagnostic superclasses: NORM, MI, STTC, CD, and HYP. First, a fourth-order Butterworth band-pass filter (0.5–40 Hz) is applied to remove baseline wander and high-frequency noise, followed by adaptive Symlet-8 wavelet denoising with soft thresholding to suppress residual high-frequency fluctuations while preserving the P-wave, QRS complex, and T-wave morphology, after which the signal is z-score normalized. R-peaks are subsequently detected to segment standardized cardiac cycles. VMD is then employed to decompose the heartbeat signals into intrinsic modes, from which the most informative modes are retained using correlation-based mode selection. Temporal, statistical, spectral, and nonlinear features are extracted to capture complementary characteristics of cardiac electrical activity, while mRMR selects the eight most informative features by maximizing feature relevance and minimizing redundancy. The resulting representation is processed through a hybrid CNN–Transformer architecture, in which convolutional layers learn local morphological patterns and Transformer-based attention captures long-range dependencies within the cardiac feature representation. The proposed framework achieves 93.60% accuracy, 93.61% macro precision, 93.60% macro recall, 93.60% macro F1-score, and 98.40% macro specificity across the five diagnostic classes. Confusion-matrix analysis, receiver operating characteristic (ROC) analysis, comparative evaluation, and ablation experiments further demonstrate the discriminative capability and robustness of the proposed approach. These findings indicate that multi-domain biomedical feature learning combined with attention-based deep learning can provide an effective and robust strategy for automated ECG-based cardiac disease classification, highlighting the potential of machine learning for intelligent biomedical signal analysis and computer-aided clinical diagnosis. Full article

Biomedical spectroscopy, including Raman, surface-enhanced Raman spectroscopy (SERS), infrared spectroscopy, and hyperspectral imaging, is increasingly combined with machine learning for disease classification, sample characterization, and biomarker-oriented analysis. However, high predictive performance does not establish whether model-relevant spectral features are biologically meaningful or whether highlighted regions can support reliable biochemical interpretation. Explainable artificial intelligence (XAI) methods, particularly SHAP and LIME, are increasingly used to identify influential wavenumbers, spectral bands, and wavelength intervals; yet feature importance is often interpreted too directly as biochemical or clinical evidence. This focused narrative review synthesizes SHAP, LIME, and related XAI methods across biomedical spectroscopy applications in cancer diagnostics, microbial identification, pharmaceutical analysis, and tissue or biofluid characterization. We examine key challenges, including correlated variables, peak overlap, preprocessing dependence, background choice, model dependence, and explanation instability. Beyond spectral attribution, we propose a five-step validation framework linking valid model development, explanation stability, region-level interpretation, biochemical plausibility, and independent analytical, biological, or clinical validation. The framework is intended to distinguish candidate spectral evidence from unstable or technically confounded explanations and to support reproducible, transparent, and clinically meaningful use of XAI in biomedical spectroscopy. Full article

This paper presents a significant evolution of the Identification of Active Anomalies (IAA) system, moving beyond previous descriptive frameworks by integrating an advanced machine learning pipeline for automated, real-time Structural Health Monitoring (SHM). Utilizing acoustic emission (AE), the upgraded IAA framework combines signal clustering, image recognition, and machine learning to monitor the structural condition of a highway overpass located near a major urban agglomeration. The monitoring results provide a reliable foundation for assessing structural health and implementing automated traffic control, which is essential to ensure safe operations. Unlike baseline implementations, this intelligent system extracts multi-parametric features using Principal Component Analysis (PCA) and transforms temporal wave streams into Continuous Wavelet Transform (CWT) scalograms. These visual representations are processed by a custom 14-layer Deep Convolutional Neural Network (CNN) combined with an unsupervised Self-Organizing Map (SOM) to eliminate operational noise and classify internal failures. AE signals recorded under service loads undergo multi-parametric analysis using pattern recognition techniques and are assigned to specific classes corresponding to active anomalies within the material or structure. Each class is linked to a distinct structural hazard level, ranging from safe operation to a critical loss of structural safety. Corresponding traffic control measures, including vehicle speed and weight restrictions, are dynamically introduced to maintain operational safety. To validate the scalability of the framework, this study synthesizes statistical data across a comprehensive fleet of 180 monitored bridge structures, backed by a predictive ARIMA time-series model that forecasts residual service life. The proposed methodology was experimentally validated on an A2 highway overpass, a vital component of the Łódź transport hub that facilitates north–south and east–west transit in Poland. The IAA system functions as a proactive diagnostic tool for infrastructure management agencies, preventing sudden, unforeseen structural failures. Ultimately, it enables the efficient and safe operation of a Smart City while ensuring that maintenance funds are rationally and optimally allocated. Full article

Purpose: Early detection of glottic cancer is essential for better outcomes, but current diagnostic methods remain largely invasive, highlighting the need for reliable non-invasive alternatives. Acoustic voice signals offer a non-invasive screening input, but their clinical value depends on whether signal-processing features and learned models can distinguish malignancy from acoustically similar benign lesions rather than merely separate pathological from healthy voices. Despite growing research on acoustic analysis combined with machine learning (ML) techniques, no systematic synthesis of this evidence currently exists. Therefore, the present review aimed to identify acoustic features that differentiate early glottic malignancies from both benign vocal fold lesions and normal vocal function using various ML techniques. Method: Following PRISMA 2020 guidelines, a comprehensive literature search was conducted across Scopus, EBSCO, Ovid (Embase), and PubMed databases through November 2025. Eighteen studies met inclusion criteria, and PROBAST was used for quality appraisal of included studies. Data extraction focused on acoustic feature categories, ML techniques, classification performance metrics, and validation strategies. Results: Nine acoustic feature categories emerged, with cepstral coefficients (particularly MFCCs) most prevalent across 55.6% of studies. Binary pathological-versus-normal classification achieved consistently high accuracy (96–98.3%) across both traditional machine learning and deep learning approaches. However, malignant-versus-benign discrimination demonstrated substantially lower performance (81–87.88% accuracy, AUC 0.631–0.91), with voice-only models proving insufficient. Multimodal models incorporating clinical variables or laryngoscopic images achieved higher performance, but their results cannot be attributed to voice features alone. All 17 prediction studies were judged at high overall risk of bias, predominantly because of small effective sample sizes, model-selection optimism, possible data leakage, and limited independent validation. Conclusions: ML can detect broad voice pathology, but current evidence is insufficient to support stand-alone acoustic screening for early glottic cancer. Future studies require clearly defined cancer-specific targets, patient-level analysis, standardized acquisition, nested model development, calibration, and independent multicentervalidation. Full article

Online partial discharge monitoring is increasingly complementing periodic offline testing in medium-voltage switchgear, particularly through the use of embedded and edge-computing platforms. This review critically examines systems based on artificial intelligence and machine learning for partial discharge detection and classification, considering the complete diagnostic chain from sensing to field deployment. The analyzed literature is organized into five interdependent layers: sensors and analog front-ends, data acquisition and triggering architectures, phase-synchronized signal representations, machine learning models, and target embedded hardware. Sensing techniques based on high-frequency current transformers, transient earth voltage, and ultra-high-frequency sensors are compared in terms of bandwidth, sensitivity, installation requirements, and immunity to interference. Particular attention is given to phase-resolved partial discharge patterns, time–frequency representations, event-driven acquisition, hardware-assisted data reduction, and synchronization mechanisms. The analysis demonstrates that high classification accuracy obtained under offline laboratory conditions does not, by itself, indicate deployment readiness. Practical implementations must also satisfy constraints related to analog-to-digital converter bandwidth, buffering, memory usage, inference latency, energy consumption, quantization, thermal performance, and field noise. Lightweight neural networks, optimized object detectors, input dimensionality reduction, quantized inference, and multimodal data fusion are identified as promising development directions. However, current research remains limited by laboratory-scale validation, incomplete hardware reporting, and insufficient long-term field datasets. Full article

Maritime decarbonisation requires transparent use of verified emissions data without overstating what regulatory datasets can explain. This study examines how EU Monitoring, Reporting, and Verification (MRV) data can support descriptive benchmarking of container ship distance-normalised fuel consumption. Of 14,147 2024 MRV records, 2097 container ships were retained. Fuel consumption per nautical mile (FC/nm) is treated as an absolute distance-normalised fuel-use indicator, not a cargo-adjusted efficiency metric. Four algorithms were assessed under target-related diagnostic and reduced specifications. Diagnostic results represent target-related reconstruction, not genuine prediction. Training-only hyperparameter tuning yielded reduced-specification test R² values of 0.8958 for Gradient Boosting and 0.8888 for XGBoost. Random-Forest attribution in the reduced specification was dominated by emissions-derived predictors. A strict non-emissions Random Forest using only sea time and certification performed worse than the mean-value baseline (R² = −0.2890). Comparison with MRV-reported fuel consumption per transport work (mass), available for 95.1% of vessels, showed a strong inverse association (ρ = −0.822), confirming that FC/nm quartiles are not cargo-adjusted efficiency rankings. The study shows that explainable machine learning can transparently diagnose information structure in MRV data, while demonstrating that these variables alone cannot support causal, voyage-level, cross-vessel efficiency, or policy interpretations. Full article

Compressed-air leaks create persistent parasitic demand, but machine-level condition monitoring is difficult because air consumption changes strongly with operating mode. LEADI (Leak Energy Anomaly Detection Index) was developed as an operating-mode-aware procedure that evaluates the deviation of directly measured flow rate from a local reference baseline derived from a stable post-repair condition with maintained pressure and low within-window variability. The method was developed on days 1–5 and evaluated on held-out days 6–7 from two one-week campaigns conducted before and after implementation of the prescribed corrective actions. With 60 min windows, LEADI flagged 19/19 evaluable pre-repair and 0/17 post-repair windows, with diagnostic coverage of 39.6% and 35.4%, respectively. A simple fifth-percentile flow comparator without operating-mode selection flagged 47/48 versus 1/48 windows. This shows that the low-flow region itself contains strong discriminatory information for separating the two periods. The role of the operating-mode layer is to restrict engineering interpretation to pre-specified eligible operating conditions. The flow-rate difference within the diagnostic operating condition was 122.0 L/min (95% CI 114.6–132.3). Over a common 168 h basis, measured volume decreased by 1370.8 m³ (28.90%), while a separate check normalized by pressurized time gave 28.12%. Because the specific energy consumption of the compressor station was not measured, the energy effect is reported only as a scenario for the same 168 h. The results support the applicability of LEADI as a selective decision-support layer for the investigated asset and the two observed conditions, without establishing universal leak detection or causal attribution of the observed change to individual defects. Full article

The rapid digital transformation of the retail industry has generated large-scale, high-frequency data; yet many retailers still rely on siloed systems that decouple demand forecasting from operational inventory management. This separation often leads to structural inventory imbalances, including simultaneous stockouts and overstock situations. To address this, we propose a unified, data-driven framework that integrates advanced sales forecasting with a diagnostic inventory health system, bridging predictive analytics with diagnostic decision support for proactive inventory-risk assessment. Utilizing a real-world dataset of approximately 500,000 product–store–day records from a Chinese e-commerce company, we evaluate six forecasting models across statistical, machine learning, and Deep Learning (DL) architectures. Results indicate that DL models achieved the lowest RMSE and MAPE values, whereas RF and XGBoost produced more favorable MASE values. Forecasting performance varied by evaluation criterion, with no model consistently outperforming all others. However, LSTM achieved the lowest testing RMSE (0.3996) and MAPE (21.63%). Building upon these predictive outputs, we introduce an inventory health diagnosis framework based on empirically calibrated thresholds for the Inventory Turnover Ratio (ITR) and Excess Inventory Rate (EIR), with the thresholds calibrated on the training and calibration period and evaluated on an independent temporal validation period. Application of this framework to the independent diagnostic validation period shows that 61.1% of SKU–store observations were classified as Potential Risk or Critical, including 16.3% classified as Critical, indicating inventory misalignment requiring managerial attention. By providing an interpretable and scalable system for proactive risk detection, this research provides a practical framework to support retailers in transitioning from demand forecasting to inventory diagnostic decision support. Full article

This narrative review focuses on the artificial intelligence (AI) approaches to the diagnosis and diagnostic support of COVID-19 in a clinical and regulatory setting. Although initial research found high internal accuracy, most had bias, little extra validation and poor real-world applicability. Artificial intelligence strategies include imaging (CT, CXR and ultrasound), clinical data, and multimodal systems based on machine learning, deep learning and new transformer models. Performances are very sensitive to the quality of data, labeling and representativeness. Key challenges include dataset shift, confounding, and fairness. Recent innovations in reporting standards and regulatory frameworks are geared towards enhancing reliability. The external validation, integration with clinical and provisional impact on patient outcomes and healthcare workflows should be prioritized in future work. Full article

Background: Hospitalizations among patients with Alzheimer’s disease (AD) carry substantial mortality risk, but length of stay (LOS) is time-dependent and may reflect heterogeneous inpatient trajectories. We examined unadjusted and adjusted LOS–mortality patterns and compared admission-only versus inpatient-course prediction using explainable machine learning. Methods: Using the full 2017 Nationwide Readmissions Database (NRD), we identified hospitalizations among adults aged ≥60 years with an ICD-10-CM G30.x AD code in any diagnosis position. Records with missing in-hospital mortality status were excluded. LOS was summarized in clinically interpretable bins and modeled using restricted cubic splines. Model A excluded explicit inpatient-course measures, whereas Model B added LOS, procedure count, and total charges. Performance was evaluated using patient-grouped 5-fold out-of-fold validation and summarized by AUROC and AUPRC; SHAP was used for interpretation. Results: Among 249,507 AD hospitalizations, 12,666 in-hospital deaths occurred (5.08%; weighted mortality 4.97%). Unadjusted mortality was highest at LOS 0–1 day (13.00%), lowest at 4–6 days (3.47%), and increased to 7.77% at ≥22 days. After multivariable adjustment, LOS remained strongly nonlinear, but adjusted predicted mortality declined across the modeled LOS range. Model A achieved AUROC/AUPRC of 0.780/0.180, whereas Model B improved to 0.828/0.329. Sepsis, diagnostic burden, acute kidney injury, age, stroke, and pneumonia were stable predictors; LOS and procedure burden added prognostic information in Model B. Conclusions: The crude LOS–mortality pattern was U-shaped, whereas the adjusted pattern suggests that the late-stay increase in unadjusted mortality is partly explained by patient complexity and evolving inpatient-course factors. Admission-only prediction provides meaningful early risk stratification, while inpatient-course information improves prognostic assessment as hospitalization evolves. Full article

The analysis of transient regimes is important for the reliable monitoring and diagnostics of electromechanical systems, since dynamic changes in load have a significant impact on the electrical and mechanical characteristics of electric motors. In this study, an approach for classification of dynamic operating regimes is proposed using ten synthetically generated load profiles representing linear, nonlinear, periodic, stochastic, and combined loads. An experimental setup was developed for implementing controlled load effects and recording electrical and mechanical parameters of DC and induction electric motors. To determine the most informative characteristics, the ReliefF, SFCPP, and FSNCA methods were applied, followed by dimensionality reduction and classification using statistical and machine-learning approaches. The selected feature vectors were evaluated by stratified 5-fold cross-validation and validated using Wilcoxon, Friedman, and permutation tests. The results obtained show that a limited set of electrical and electromechanical parameters contains sufficient information to reliably distinguish between different dynamic operating modes. The best-performing model achieved classification accuracy above 98%, confirming the effectiveness of the proposed framework. The proposed approach provides a reproducible methodology for generating representative datasets, assessing the informativeness of the features, and supporting intelligent systems for monitoring the condition of electric drives. The methodology is applicable both in laboratory conditions and in the development of digital twins and predictive maintenance systems. Full article

This paper raises a methodological question: How can we assess machine-made Old English when there is no parallel reference text and the standard metrics do not fit the task? We propose a pipeline with five measuring layers plus two compliance components, including lexical attestation with form linking, frequency-profile diagnostics, character-level comparison, word-embedding geometry under a verified mapping and dependency parsing, with bootstrap confidence intervals around the main contrasts. We apply the pipeline to a machine-made version of Gregory’s Dialogues: 4217 sentences, one for each sentence of the Old English original, generated under hard constraints. The unattested residue is two word types and 0.003% of tokens. The frequency profile diverges from the original by 0.008, less than the original diverges from the background corpus. At character level, in embedding space and in parsed syntax, the generated text stands at the same distance from the Dictionary of Old English Corpus as the original itself does. We propose an overall metric G, the geometric mean of seven bounded components, which scores the text at 0.991 with a confidence interval of [0.991, 0.992]. Two blind detection experiments with expert judges place the index externally: roughly two-thirds of generated sentences pass as authentic to specialists, so the divergence the pipeline measures is real at corpus scale but not available to sentence-by-sentence reading. The main contribution is a reusable evaluation method for historical language generation, together with a single interpretable score that subsumes the partial metrics without hiding them. Full article

Background/Objectives: Endocrine resistance due to activating mutations in the estrogen receptor alpha (ERα), especially Y537S, is still a big challenge in the treatment of hormone receptor-positive breast cancer, and there is a need to develop new small-molecule inhibitors that can overcome ligand-independent receptor activation. We used an integrated machine learning and molecular dynamics (MD)-based virtual screening (VS) approach to identify potential computationally prioritized hits of ERα Y537S in this study. Methods: The dataset (~3353 compounds) was characterized using molecular fingerprints and graphically represented by PCA and t-SNE to show structurally different clusters linked to potency (pIC50). To guarantee robust downstream model training, the diagnosis of structural and influential outliers was completed through the application of rigorous data quality control methods, including the Williams plot, Mahalanobis distance, and Cook’s distance diagnostics. This was followed by multiple benchmarks of machine learning classifiers (random forest, SVM, KNN, and CNN) for activity classification, yielding the highest AUC of 0.95 for the random forest classifier. Interpretable structure–activity insights were gained by the identification of key substructures of the fingerprint by SHAP and feature importance analysis that influence the predicted potency. Three lead candidates (Hit-1, Hit-2, and Hit-3) were screened and tested with a 200 ns all-atom MD simulation in comparison with a reference control to evaluate the binding stability. Comprehensive trajectory analysis, such as PCA, FEL, RDF, salt bridges, and DCCM, was completed following the structure-based identification of the most stable ligands. Results: Three candidates were identified as Hit-1, Hit-2, and Hit-3. Hit-3 was found to be bound in the most stable binding pose and had the most similar conformational behavior to the control, whereas Hit-2 showed transient pose instability associated with increased anti-correlated domain-level motions and an alternate high-salt-bridge conformational state. As no computational metric can be an absolute measure of experimental affinity, Hit-1 has the most favorable calculated end-point binding-energy estimate, while Hit-3 showed conformational stability in MD simulations. Conclusions: The computationally prioritized candidates can be subjected to further experimental testing, as the conducted docking, MD simulations, and end-point free energy analyses cannot establish overall compound binding, cell-based activity, pharmacological mechanism, or therapeutic efficacy. Full article

This study examines how accounting professionals’ self-reported forensic-auditing capabilities relate to their attitudinal readiness to recognize suspicious anti-money-laundering (AML) indicators, using a transparent hybrid design that combines classical psychometrics, structural equation modeling (SEM), and leakage-controlled explainable machine learning (ML). A cross-sectional survey collected 700 valid responses from public accountants in Lima, Peru, with a 30-item, five-point Likert instrument (19 forensic-auditing items in three dimensions and 11 AML-recognition items). Reliability was high (Cronbach’s α = 0.935; McDonald’s ω = 0.935), but average variance extracted was below 0.50 in every block (0.329–0.449), and Fornell–Larcker testing showed that skills-and-knowledge and AML recognition were not discriminantly distinct (r = 0.673 > √AVE = 0.651/0.649). Responses showed a pronounced ceiling (51% of answers were the maximum), and 86 respondents (12.3%) answered all 30 items identically; removing them lowered the forensic-auditing–AML association from r = 0.731 to 0.630 and explained variance from 53.8% to 40.3%. Skills-and-knowledge remained the strongest predictor in SEM and HC3-robust regression (β = 0.478 and 0.425). Under a leakage-free protocol, ensemble models reached ROC-AUC ≈ 0.86 on held-out data, but threshold tuning did not improve F1 test, and item-level attributions were unstable (Spearman ρ = 0.28). Forensic-auditing capabilities are positively associated with declared AML-recognition readiness, driven by applied skills and knowledge; the evidence is attitudinal and correlational, and should not be read as real detection capability. Because professional experience, seniority, sector, and prior AML training were not measured, the reported associations may be partly confounded by unobserved professional background, and the dominance of skills-and-knowledge is therefore advanced as tentative, pending resolution of the skills-and-knowledge/AML-recognition discriminant-validity overlap. A second, procedural contribution is that the study reports the data-quality screening, the failed validity tests, and the explanation-stability diagnostics that survey-based forensic-accounting research rarely makes visible. Full article