Engineering reports : open access最新文献

Inspection of surface defects in industrial components is vital to ensuring product quality and operational safety. This study presents an improved YOLOv8s-based deep learning model designed to detect fatigue and linear cracks on part surfaces with high localization and classification accuracy. Unlike conventional manual inspection, the proposed system achieves real-time detection while maintaining low computational cost. A custom dataset comprising high-resolution images of concrete and metallic textures captured under diverse environmental conditions was used for training and evaluation. Data augmentation techniques such as Mosaic and Contrast Limited Adaptive Histogram Equalization (CLAHE) were employed to enhance generalization, and the model was trained for 100 epochs to ensure stable convergence. The enhanced YOLOv8s architecture integrates the Convolutional Block Attention Module (CBAM) and Ant Colony Optimization (ACO) for intelligent feature selection, resulting in improved learning efficiency and reduced overfitting. Experimental results show that the model achieves mAP@IoU = 0.5 of 0.9626 and [email protected]:0.95 of 0.7803, with validation box and class losses of 0.6849 and 0.5377, respectively. Precision and recall values of 0.926 and 0.923 demonstrate excellent detection completeness and low false positives. Comparative analysis with BsS-YOLO and Efficient YOLOv8-ES confirms the superior accuracy and efficiency of the proposed approach. Visual inspection further validates the model's robustness in identifying diverse crack patterns under challenging surface conditions. The improved YOLOv8s model thus offers a scalable, real-time, and accurate solution for automated defect detection in industrial applications. Future work will focus on multi-class defect recognition and deployment on lightweight edge-computing devices.

Suicide ranks as the 18th leading cause of death worldwide among young adults, claiming over 720,000 lives each year. Early detection of individuals at risk is essential for timely intervention. This study introduces a Triple-Layer Ensemble (TLE) model that predicts suicidal behavior using machine learning techniques. The proposed model combines Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN) to enhance prediction accuracy. Experimental results show that the TLE model surpasses individual classifiers and traditional ensemble methods, achieving 94.81% accuracy, 98.15% ROC-AUC, and a Matthews Correlation Coefficient (MCC) of 92.23%. To improve interpretability, Explainable AI (XAI) methods, Local Interpretable Model-Agnostic Explanations (LIME), and Shapley Additive Explanations (SHAP) highlight key predictors such as mental support, stress levels, and self-harm history. Additionally, a web-based platform incorporating the TLE model provides real-time suicide risk assessment, enabling healthcare professionals to implement personalized interventions. The proposed framework delivers high predictive performance with transparency and interpretability, offering a scalable solution for early suicide risk prediction and prevention.

The increasing adoption of cyber-physical systems (CPS) in Industry 4.0 has heightened vulnerability to cyber threats. This study proposes a machine learning–based intrusion detection framework, DBID-Net, to effectively identify and prevent attacks in CPS environments. The framework integrates advanced modules for data handling and threat analysis, employing AMMIS for data cleaning, ADASYN for data augmentation, CESF for feature extraction, and the Whale Hopper Optimization Algorithm (WHOA) for optimal feature selection. A bidirectional learning model is used for intrusion detection, with WHOA further enhancing scalability and adaptive attack mitigation. Experimental results show that DBID-Net achieves an F1-score of 0.988, a sensitivity of 0.984, a specificity of 0.983, and an accuracy of 0.9916. These findings demonstrate that DBID-Net offers a robust and scalable solution for securing CPS infrastructures against evolving cyber threats in Industry 4.0.

Modern power systems are significantly impacted by unpredictable load fluctuations, renewable energy integration, and an increasing number of outages—operational scenarios that have the potential to cause voltage instability and collapse. This necessitates near real-time monitoring and control by operators, enabling the identification of critical lines or vulnerable buses operating close to their stability limits. This study proposes a modified line voltage stability index (MLVSI) to enhance the accuracy and computational speed of voltage stability assessment (VSA) adapted to diverse operating scenarios. The index incorporates active and reactive power, the angular difference between sending end and receiving end bus voltages, and line impedance. Using the IEEE 57-bus test system, the proposed index was validated against existing stability indices—line stability index (Lmn), modern voltage stability index (MVSI), and novel collapse prediction index (NCPI)—by comparing accuracy and computation time under various operational scenarios. Results indicate that MLVSI provides higher accuracy and faster detection, particularly under extreme operating conditions. For example, compared with the NCPI, the MLVSI achieved a 4.8% improvement in accuracy and reduced the computation time by 0.07–0.103 s on different operating scenarios. The adaptability of MLVSI to diverse scenarios underscores its potential for broad application in the assessment of voltage stability.

This review critically examines the application of microwave-assisted technologies in gold mining and processing, highlighting their potential to improve extraction efficiency and environmental sustainability. The study focuses on the use of microwave irradiation in ore pretreatment, leaching enhancement, treatment of waste activated carbon, and synthesis of gold nanoparticles. Evidence from recent research demonstrates that microwave-assisted processes can significantly increase gold recovery rates, reduce processing times, and lower energy consumption compared to conventional techniques. For refractory ores, microwave pretreatment effectively improves mineral liberation and leaching kinetics, achieving extraction rates exceeding 90% in some cases. Additionally, the integration of microwave roasting with chemical additives such as NaOH and KOH has shown further enhancement in gold recovery. Despite these promising outcomes, challenges remain in terms of temperature control, process scalability, and optimization across different ore types. The review concludes by outlining key directions for future research, including the development of industrial-scale systems, comprehensive economic assessments, and the exploration of microwave applications in combination with alternative lixiviants. Overall, microwave-assisted technologies present a promising pathway toward more efficient and sustainable gold production.

Oily sludge, a hazardous waste generated during crude oil exploitation, gathering, and transportation, with crude oil constituting one of its primary sources of pollution. Reclamation of hydrocarbon components within oily sludge constitutes a pivotal step toward its resource utilization. However, the presence of emulsifying surfactants and adhesion of crude oil to solid surfaces foster stable systems, which impede oil recovery and necessitate demulsification as a prerequisite for effective reclamation. HTCO, which generates abundant free radicals, represents a promising strategy for destabilizing the stabilized oil–water interface. This study employed single-factor and orthogonal array experiments to systematically investigate the effects of five key operational parameters on crude oil recovery efficiency: catalyst concentration (0–200 mg L⁻¹), reaction temperature (50°C–250°C), reaction time (5–25 min), oxidation coefficient (1.0–3.0), and solid-to-liquid ratio (1:20–1:4). After HTCO treatment of oily sludge, optimal conditions were identified as follows: catalyst concentration at 50 mg L⁻¹, temperature at 200°C, reaction time of 10 min, oxidation coefficient of 1, and solid-to-liquid ratio at 1:5. Under these parameters, the oil recovery efficiency reached 89.46%. HTCO synergistically breaks highly emulsified oil sludge efficiently with high recovery rates, enhances crude oil quality, reduces downstream costs, and delivers environmental benefits.

This study evaluates audio source separation with Fast Independent Component Analysis (FastICA) in a fully specified, reproducible pipeline and benchmarks it against Principal Component Analysis (PCA) and Nonnegative Matrix Factorization (NMF). Three conversational recordings collected at the Department of Statistics, Pabna University of Science and Technology were canonicalized to 48 kHz WAV 96.63 s each, converted to mono, trimmed at 25 dB, RMS-normalized, and time-aligned. Sources were mixed with a fixed 3 × 3 row-normalized Gaussian matrix. FastICA used parallel updates with a logcosh nonlinearity and unit-variance whitening; PCA served as a multichannel linear baseline; NMF operated on a mono short-time Fourier transform with KL divergence and NNDSVDA initialization, followed by soft masking and inverse transform. Performance was computed with BSS Eval after best-permutation and scale alignment and summarized over 10 runs as mean ± SD. FastICA achieved Signal-to-Distortion Ratio (SDR) 53.51 ± 0.07 dB, Signal-to-Interference Ratio (SIR) 53.52 ± 0.07 dB, and Signal-to-Artifact Ratio (SAR) 79.58 ± 0.00 dB, well above a 20 dB high-quality SDR threshold. PCA yielded SDR 2.79 ± 0.00 dB, SIR 2.79 ± 0.00 dB, SAR 80.64 ± 0.00 dB; NMF produced SDR −2.26 ± 0.00 dB, SIR 0.41 ± 0.00 dB, SAR 4.80 ± 0.00 dB. Waveform and spectrogram visualizations, together with descriptive, high-order, and entropy statistics, corroborate these outcomes. The results establish FastICA as an effective classical approach for audio source separation and provide a transparent reference pipeline for comparative studies using SDR, SIR, and SAR.

Grocery product recognition faces unique challenges in distinguishing visually similar items across hierarchical categories while maintaining computational efficiency. Though powerful in image classification, traditional vision transformers (ViTs) struggle with specialized retail datasets due to their high parameter counts and inadequate local feature extraction for fine-grained distinctions. We present HARVEST, a lightweight transformer architecture that addresses these limitations through five key components: (1) shifted patch tokenization, which enhances local feature capture via overlapping diagonal patches; (2) local information enhancer, which injects spatial awareness into patch embeddings; and (3) hierarchical attention, an integrated module that dynamically unites locality-enhanced attention, cross-level attention, and progressive classification heads to effectively fuse multiscale features across hierarchical levels. Evaluated on the Hierarchical Grocery Store dataset, HARVEST achieves 98.73% coarse-grained and 97.06% fine-grained accuracy with only 2.66M parameters–82.7% fewer than conventional models. This performance stems from its ability to resolve critical retail recognition challenges: distinguishing near-identical packaging variants (e.g., juice flavors differing by subtle color gradients) and capturing hierarchical relationships between product categories (e.g., apples $��\to �$ varieties) through progressive classification heads. The architecture's efficiency and accuracy advance automated shelf monitoring and inventory systems without compromising computational practicality. HARVEST achieves real-time performance, defined as inference latency below 10 ms per image on an NVIDIA T4 GPU, with an average throughput of 146.6 images per second (batch size 1), thereby facilitating seamless assistive grocery classification.

This study explores the detailed effects of speed, road adhesion, road slope, and dynamic maneuvers on the lateral stability of vehicles equipped with four-wheel steering (4WS) systems, also referred to as all-wheel steering (AWS) system. The research uses CarSim simulation software to comprehensively evaluate vehicle performance under various speeds (30–90 km/h), road slope (10°–30°) and adhesion coefficients (0.1–0.85) during lane changes and turning maneuvers. The results obtained reveal that 4WS significantly enhances stability and handling at moderate speeds and favorable adhesion conditions. However, high-speed operations, particularly on low adhesion surfaces or with a steep road slope, increase instability and safety risks. Practical implications also emphasize the importance of maintaining safe speeds, particularly under adverse road conditions, and decreasing speed during sharp turns to counteract centrifugal forces. Finally, these findings highlight the substantial role of cautious driving and adherence to speed limits in improving overall safety for 4WS vehicles.

The synergistic corrosion inhibition of a unique molecular combination, hexadecanoic acid-methyl ester, terpineol, and 1,5,9-undecatriene-2,6,10-trimethyl-(Z) present in Sunshine Ligustrum (yellow bush) extract was investigated on Fe (110) crystal surfaces in acidic medium. The study employed density functional theory (DFT), Monte Carlo (MC) simulations, radial distribution function (RDF), weight loss (WL), electrochemical impedance spectroscopy (EIS), potentiodynamic polarization (PDP), FE-SEM, and EDX analyses. The orbital matching principle (OMP) revealed that neutral 1,5,9-undecatriene-2,6,10-trimethyl-(Z) and terpineol donate electrons from their HOMO to the 3d-orbital of iron, due to their small energy gaps (0.530 and 0.552 eV, respectively). The protonated hexadecanoic acid-methyl ester and terpineol revealed higher HOMO (−0.229 and −0.239 eV); this indicates initial adsorption by physisorption followed by chemisorption. MC analysis showed that the three phytochemicals were synergistically adsorbed at low-energy sites. The RDF suggests the phytochemicals will bond with steel by chemisorption. The raw extract demonstrated high inhibition efficiencies of 77.53% by WL at 40 ppm and approximately 90% by EIS and Tafel plots at 60 ppm. Tafel plots indicate that yellow bush is a mixed-type (ambiodic) inhibitor. EIS analysis showed that the charge-transfer resistance (R_ct) increased significantly from 8.91 ± 6.6E-2 to 88.91 ± 7.4E-1Ωcm² with increasing inhibitor concentration, while the double-layer capacitance (C_dl) decreased from 716.47 ± 5.3 to 357.4 ± 3.1μFcm². FE-SEM and EDX confirmed the formation of a protective layer that effectively inhibited acidic corrosion. The results strongly suggest that these three phytochemicals warrant further screening before industry deployment as effective, low-dosage green corrosion inhibitors for the control of acidic corrosion of API 5 L X65 steel.