Search Results (5,498)

In this paper, we propose a novel few-shot optimization with Hybrid Euclidean Distance with Large Language Models (HED-LM) to improve example selection for sensor-based classification tasks. While few-shot prompting enables efficient inference with limited labeled data, its performance largely depends on the quality of selected examples. HED-LM addresses this challenge through a hybrid selection pipeline that filters candidate examples based on Euclidean distance and re-ranks them using contextual relevance scored by large language models (LLMs). To validate its effectiveness, we apply HED-LM to a fatigue detection task using accelerometer data characterized by overlapping patterns and high inter-subject variability. Unlike simpler tasks such as activity recognition, fatigue detection demands more nuanced example selection due to subtle differences in physiological signals. Our experiments show that HED-LM achieves a mean macro F1-score of 69.13 ± 10.71%, outperforming both random selection (59.30 ± 10.13%) and distance-only filtering (67.61 ± 11.39%). These represent relative improvements of 16.6% and 2.3%, respectively. The results confirm that combining numerical similarity with contextual relevance improves the robustness of few-shot prompting. Overall, HED-LM offers a practical solution to improve performance in real-world sensor-based learning tasks and shows potential for broader applications in healthcare monitoring, human activity recognition, and industrial safety scenarios. Full article

With the enhancement of safety performance requirements in the car manufacturing field, the quality standards for the sheet molding process have imposed higher demands. However, during the pressure-building phase of pump-controlled hydraulic die cushion systems, the combined effects of high-order dynamics, system uncertainties, and strong nonlinearities pose significant challenges to maintaining precise control and dynamic response performance of the blank holder force (BHF). To address these challenges, we propose an intelligent model predictive control (MPC) strategy that synergistically integrates an extended state observer (ESO) for disturbance compensation with parameters optimized by a genetic algorithm (GA). The mathematical model and state-space model of the system are established. Subsequently, the ESO is integrated with MPC to enable active compensation for internal and external disturbances. The GA is employed to optimize the controller parameters within the MPC framework. Finally, a simulation testbed for the pump-controlled hydraulic die cushion experimentally validates the process. Experimental results demonstrate that compared to MPC and conventional PID control, the proposed strategy achieves significant reductions in pressure overshoot (0.87% and 1.8% at 100 bar; 3.3% and 5.9% at 200 bar), pressure-building time (13.9% and 31.4% at 100 bar; 6.7% and 11.5% at 200 bar), and stroke length (10.5% and 32% at 100 bar; 11.5% and 28.1% at 200 bar). This validates its effectiveness in enhancing both control precision and dynamic response performance, providing a reliable solution for large-scale applications of pump-controlled hydraulic die cushions in high-dynamic stamping scenarios. Full article

This paper presents the design of a fault detection system (FDD) based on principal component analysis (PCA) to detect faults in the transient state of distillation processes. The FDD system detects faults due to changes in calorific power and pressure leaks that can occur during the heating of the mixture to be distilled (transient), mainly affecting the quality of the distilled product and the safety of the process and operators. The proposed FDD system is based on PCA with a T2 Hotelling statistical approach, considering data from a real distillation pilot plant process. The FDD system is evaluated with two fault scenarios, performing power changes and pressure leaks in the pilot plant reboiler during the transient state. Finally, the results of the FDD system are analyzed using Accuracy, Precision, Recall, and Specificity metrics to validate its performance. Full article

Modern computing systems increasingly rely on specialized hardware accelerators, such as Graphics Processing Units (GPUs), to meet growing computational demands. GPUs are essential for accelerating a wide range of applications, from machine learning and scientific computing to safety-critical domains like autonomous systems and aerospace. To enhance performance, modern GPUs integrate dedicated in-chip units, such as Tensor Cores(TCs), which are designed for efficient mixed-precision matrix operations. However, as semiconductor technologies scale down, reliability challenges emerge. Permanent hardware faults caused by aging, process variations, or environmental stress can lead to Silent Data Corruptions, which silently compromise computation results. In order to detect such faults, self-test libraries (STLs) are widely used, corresponding to suitably crafted pieces of code, able to activate faults and propagate their effects to visible points (e.g., the memory) and possibly signal their occurrence. This work introduces a structured method for generating STLs to detect permanent hardware faults that may arise in TCs. By leveraging the parallelism and regular structure of TCs, the method facilitates the creation of effective STLs for in-field fault detection without hardware modifications and with minimal requirements in terms of test time and memory. The proposed approach was validated on an NVIDIA GeForce RTX 3060 Ti GPU, installed in a Hewlett-Packard Z2 G5 workstation with an Intel Core i9-10800 CPU and 32 GB RAM, available at the Department of Control and Computer Engineering (DAUIN), Politecnico di Torino, Turin, Italy.This setup was used to address stuck-at faults in the arithmetic units of TCs. The results demonstrate that the methodology offers a practical, scalable, and non-intrusive solution for enhancing GPU reliability, applicable in both high-performance and safety-critical environments. Full article

Task scheduling for multiple humanoid robot manipulators in industrial and collaborative settings remains a significant challenge due to the complexity of coordination, resource sharing, and real-time decision-making. In this study, we propose a framework for modeling task scheduling for multiple humanoid robot manipulators by using the symbolic discrete controller synthesis technique. We encode the task scheduling problem as discrete events using parallel synchronous dataflow equations and apply our synthesis algorithms to manage the task scheduling of multiple humanoid robots via the resulting controller. The control objectives encompass the fundamental behaviors of the system, strict rules, and mutual exclusions over shared resources, categorized as the safety property, whereas the optimization objectives are directed toward maximizing the throughput of robot-processed products with optimal efficiency. The humanoid robots considered in this study consist of two pairs of six-degree-of-freedom (6-DOF) robot manipulators, and the inverse kinematics problem of the 6-DOF arms is addressed using metaheuristic approaches inspired by biomimetic principles. Our approach is experimentally validated, and the results demonstrate high accuracy and performance compared to other approaches reported in the literature. Our approach achieved an average efficiency improvement of 40% in 70-robot systems, 20% in 30-robot systems, and 10% in 10-robot systems in terms of production throughput compared to systems without a controller. Full article

Since the 1990s, cannabis has experienced a gradual easing of access restrictions, accompanied by the expansion of its legalization and commercialization. This shift has led to the proliferation of cannabis-based products, available as cosmetics, food supplements, and pharmaceutical dosage forms. Consequently, there has been a growing demand for reliable and reproducible extraction techniques alongside precise analytical methods for detecting and quantifying cannabinoids, both of which are essential for ensuring consumer safety and product quality. Given the variability in extraction and quantification techniques across laboratories, significant attention has recently been directed toward method validation. Validated methods ensure precise cannabinoid measurement in cannabis-based products, supporting compliance with dosage guidelines and legal limits. Thus, this review highlights recent advancements in these areas, with a particular focus on High-Performance Liquid Chromatography (HPLC) coupled with Ultraviolet (UV) detection, as it is considered the gold standard for cannabinoid analysis included in cannabis monographs present in several pharmacopeias. The research focused on studies published between January 2022 and December 2024, sourced from PubMed, Scopus, and Web of Science, that employed an HPLC-UV analytical technique for the detection of phytocannabinoids. Additionally, the review examines cannabinoid extraction techniques and the validation methodologies used by the authors in the selected papers. Notably, ultrasound extraction has emerged as the most widely utilized technique across various matrices, with Deep Eutectic Solvents (DESs) offering a promising, efficient, and environmentally friendly extraction alternative. Analytical chromatographic separations continue to be predominantly conducted using C18 reversed-phase columns. Nevertheless, in recent years, researchers have explored various stationary phases, particularly to achieve the enantioseparation of cannabinoids. Full article

Physical fatigue significantly impacts safety and performance across industrial, athletic, and medical domains, yet its detection remains challenging due to individual variability and limited generalizability of existing methods. This study introduces a multimodal fatigue detection system integrating surface electromyography (sEMG) and inertial measurement unit (IMU) signals, processed through a hybrid convolutional neural network–long short-term memory–attention (CNN-LSTM-Attention) model. Fatigue was induced in 35 healthy participants via step-up-and-down exercises, with gait data collected during natural walking before and after fatigue. The model leverages sEMG from the gastrocnemius lateralis and IMU-derived jerk signals from the tibialis anterior and rectus femoris to classify fatigue states. Evaluated using leave-one-subject-out cross-validation (LOSOCV), the system achieved an accuracy of 87.94% with bilateral EMG signals and a balanced recall of 87.94% for fatigued states using a combined IMU-EMG approach. These results highlight the system’s robustness for personalized fatigue monitoring, surpassing traditional subject-dependent methods by addressing inter-individual differences. Full article

Staphylococcal enterotoxins (SEs) are major contributors to foodborne intoxications. Reliable detection methods for SEs are essential to maintain food safety and protect public health. Since the heat-stable toxins also exert their toxic effect in the absence of the bacterium, reliance on DNA detection alone can be misleading: it does not allow for determining which specific toxins encoded by a given strain are produced and epidemiologically linked with a given outbreak. Commercially available diagnostic assays for SE detection are so far limited in sensitivity and specificity as well as in the range of targeted toxins (SEA–SEE), thus non-targeted SEs linked to foodborne illness remain undetected at the protein level. This study aimed to develop a highly sensitive and specific multiplex suspension immunoassay (SIA) for SEA to SEI. To this end, high-affinity monoclonal antibodies (mAbs) for the specific detection of the individual SEs were generated. When implemented in sandwich ELISAs and multiplex SIA, these mAbs demonstrated exceptional sensitivity with detection limits in the low picogram per millilitre range. When applied for the analysis of SE production in liquid cultures of a panel of 145 whole-genome sequenced strains of Staphylococcus spp. and Enterococcus faecalis, the novel multiplex SIA detected and differentiated the eight SEs with assay accuracies of 86.9–100%. Notably, the multiplex SIA covered one to four sequence variants for each of the individual SEs. Validation confirmed high recovery rates and reliable performance in three representative complex food matrices. The implementation of the novel mAbs in a multiplex SIA enabled, for the first time, simultaneous detection, differentiation, and quantification of multiple SEs from minimal sample volumes using Luminex^® technology. As a result, the multiplex SIA will help strengthen food safety protocols and public health response capabilities. Full article

Robotic pancreaticoduodenectomy (RPD) has emerged as a viable alternative to open and laparoscopic approaches, offering potential advantages in precision and dexterity. However, its complexity and lack of standardization remain as barriers to widespread adoption. We present a step-by-step surgical approach to RPD, emphasizing key technical strategies to enhance safety, efficiency, and reproducibility. Our technique is structured into defined surgical steps, facilitating learning curve optimization and intraoperative consistency. Key refinements include an optimized trocar placement, the strategic suspension of vascular structures, and specific reconstructive techniques to reduce the operative time and improve surgical ergonomics. These improvements may contribute to a reduction in perioperative morbidity and procedural standardization. Standardizing RPD through defined surgical steps and structured learning pathways may improve its feasibility, safety, and broader adoption. Further studies are needed to validate these strategies in high-volume centers. Full article

In high-altitude cold regions, significant diurnal and seasonal temperature variations intensify freeze-thaw damage to rocks, critically impacting slope stability. This study examines a Xinjiang mine slope to assess freeze-thaw effects through laboratory experiments on three lithologies under varying freeze-thaw cycles. Mechanical parameters were determined via the Hoek–Brown criterion, and FLAC3D simulations analyzed stress-deformation characteristics and safety factor trends, validated against field monitoring. After 90 cycles, the results show progressive degradation: uniaxial compressive strength declined by 29.7–45.8%, elastic modulus by 42.7–63.3%, Poisson’s ratio by 16.0–42.1%, cohesion by 71.7–77.1%, internal friction angle by ~52.0%, and tensile strength by 79.3–83.6%. The slope safety factor decreased exponentially (44.5% reduction). Both simulations and monitoring revealed “step-like” displacement growth, with minor discrepancies attributed to modeling assumptions. These findings provide critical insights for safe mining operations in cold regions, highlighting the severe mechanical deterioration induced by freeze-thaw cycles and its implications for slope stability. Full article

The increasing popularity of edible insects as a sustainable food source necessitates stringent safety measures to monitor pesticide contamination. This study aimed to assess and enhance a QuEChERS-based extraction method coupled with gas chromatography–tandem mass spectrometry (GC-MS/MS) for the quantification of pesticide residues in edible insects (bamboo caterpillars, house crickets, silkworm pupae, giant water bugs, and grasshoppers) by combining multiple individual insect specimens into a single, homogenized sample—five replicates were tested. The method was optimized by evaluating various extraction parameters and showed strong linearity for all 47 target pesticides, with correlation coefficients (R²) ranging from 0.9940 to 0.9999. The limits of detection (LODs) varied between 1 and 10 µg/kg, while the limits of quantification (LOQs) ranged from 10 to 15 µg/kg. Recovery studies conducted at three fortification levels (10, 100, and 500 µg/kg) revealed recoveries ranging from 64.54% to 122.12%, that over 97.87% of the pesticides exhibited satisfactory recoveries within the range of 70–120%, and relative standard deviations (RSDs) below 20%, between 1.86% and 6.02%. Matrix effects (%MEs) range from −33.01% to 24.04%, and to those that experienced no effect. More than 94% of the analytes showed minimal ion suppression or enhancement. These results conform to the SANTE guidelines for monitoring pesticide residues in edible insects, enhancing food safety standards and safeguarding consumer protection. Full article

The search for natural antimicrobials has intensified with rising food safety demands. This study evaluated 23 probiotic strains, identifying Bifidobacterium sp. strain TF04 as a potent inhibitor against pathogens, with inhibition zone diameters of 12.85 ± 0.12 mm (Escherichia coli), 14.85 ± 0.10 mm (Staphylococcus aureus), and 17.50 ± 0.23 mm (Staphylococcus epidermidis). Preliminary analysis shows that the main antibacterial compounds produced by TF04 in the process of bacterial growth inhibition are antibacterial active proteins. TF04 exhibits optimal bacteriostatic activity within the pH range of 2–4, with a notable decline in effectiveness as the pH value increases. At the same time, the bacteriostat produced by TF04 showed strong thermal stability and ultraviolet stability. TF04 demonstrated excellent probiotic potential: surviving acidic (pH 2.0, >45% viability) and bile conditions (3% bile salts, >55% survival). It showed strong auto-aggregation (40.10%) and hydrophobicity (>30%), indicating gut colonization potential, along with notable antioxidant capacity. Safety was confirmed by absent hemolytic and gelatinase activities. These properties position TF04 as a promising multifunctional candidate for food preservation, combining antimicrobial efficacy with probiotic benefits. Further studies will purify its bioactive compounds and validate applications in food systems. Full article

Background: CT image guidance and navigation, although routinely used in complex endovascular procedures, is an unexplored territory in evolving vascular robotic procedures. In robotic surgery, it promises the better localization of vasculature, the optimization of port placement, less inadvertent tissue damage, and increased patient safety during the dissection of retroperitoneal structures. However, unknown tissue displacement resulting from induced pneumoperitoneum and positional changes compared to the preoperative CT scan can pose significant limitations to the reliability of image guidance. We aimed to study the displacement of retroperitoneal organs and vasculature due to factors such as increased intra-abdominal pressure (IAP) due to CO₂ insufflation and patient positioning (PP) using intraoperative CT imaging in a cadaveric model. Methods: A thawed, fresh-frozen human cadaveric model was positioned according to simulated procedural workflows. Intra-arterial, contrast-enhanced CT scans were performed after the insertion of four laparoscopic ports in the abdomen. CT scans were performed with 0–5–15–25 mmHg IAPs in supine, left lateral decubitus, right lateral decubitus, Trendelenburg, and reverse Trendelenburg positions. Euclidean distances between fixed anatomical bony and retroperitoneal vascular landmarks were measured and compared across different CT scans. Results: Comparing the effects of various IAPs to the baseline (zero IAP) in the same PP, an average displacement for retroperitoneal vascular landmarks ranged from 0.6 to 3.0 mm (SD 1.0–2.8 mm). When changing the PPs while maintaining the same IAP, the average displacement of the retroperitoneal vasculature ranged from 2.0 to 15.0 mm (SD 1.7–7.2 mm). Conclusions: Our preliminary imaging findings from a single cadaveric model suggest minimal (~3 mm maximum) target vasculature displacement in the retroperitoneum due to elevated IAP in supine position and higher displacement due to changes in patient positioning. Similar imaging studies are needed to quantify procedural workflow-specific and anatomy-specific deformation, which would be invaluable in developing and validating advanced tissue deformation models, facilitating the routine applicability and usefulness of CT image guidance for target delineation during robotic vascular procedures. Full article

Traditional reliability methods using fixed removal plans often overlook withdrawal randomness, leading to biased estimates for asymmetric data. This study advances classical and Bayesian frameworks for the inverse Chen distribution, which is suited for modeling asymmetric data under adaptive progressively Type-II censoring with binomial removals. Here, removals post-failure follow a dynamic binomial process, enhancing a more realistic approach for reliability studies. Maximum likelihood estimates are computed numerically, with confidence intervals derived asymptotically. Bayesian approaches employ gamma priors, symmetric squared error loss, and posterior sampling for estimates and credible intervals. A simulation study validates the methods, while two asymmetric real-world applications demonstrate practicality: (1) analyzing diamond sizes from South-West Africa, capturing skewed geological distributions, and (2) modeling failure times of airborne communication transceivers, vital for aviation safety. The flexibility of the inverse Chen in handling asymmetric data addresses the limitations of symmetric assumptions, offering precise reliability tools for complex scenarios. This integration of adaptive censoring and asymmetric distributions advances reliability analysis, providing robust solutions where traditional approaches falter. Full article

This paper introduces TU-DAT, a novel, freely downloadable computer vision dataset for analyzing traffic accidents using roadside cameras. TU-DAT addresses the lack of public datasets for training and evaluating models focused on automatic detection and prediction of road anomalies. It comprises approximately 280 real-world and simulated videos, collected from traffic CCTV footage, news reports, and high-fidelity simulations generated using BeamNG.drive. This hybrid composition captures aggressive driving behaviors—such as tailgating, weaving, and speeding—under diverse environmental conditions. It includes spatiotemporal annotations and structured metadata such as vehicle trajectories, collision types, and road conditions. These features enable robust model training for anomaly detection, spatial reasoning, and vision–language model (VLM) enhancement. TU-DAT has already been utilized in experiments demonstrating improved performance of hybrid deep learning- and logic-based reasoning frameworks, validating its practical utility for real-time traffic monitoring, autonomous vehicle safety, and driver behavior analysis. The dataset serves as a valuable resource for researchers, engineers, and policymakers aiming to develop intelligent transportation systems that proactively reduce road accidents. Full article