Accuracy Requirements Research Articles

A review of cellular metamaterials with thermal expansion coefficient that is negative, zero, or large positive

When designers face high accuracy requirements or large temperature changes, they need to consider the effects of thermal deformation and stresses. Therefore, materials that can adjust their coefficient of thermal expansion (CTE) are very attractive. While almost zero CTE materials minimize these problems, those with negative adjustable CTE can also offset the deformation in materials with positive CTE. In recent years, various structures with distinctive mechanisms exhibiting unusual coefficient of thermal expansion (CTE) have been proposed by researchers. A comprehensive review is required to encapsulate the current advancements, classify different mechanisms, and assess their effectiveness. This review paper will initially categorize all known materials with negative thermal expansion (NTE) properties. It then describes the continuous composite materials with NTE, including their mechanisms and applications. Following that, it examines architected cellular structures with negative CTE behavior made from positive CTE materials, which is the main focus of the paper. To evaluate the reliability of each mechanism, comparisons and discussions are made based on their negative CTE range, thermal stresses, stiffness, and strength. Additionally, the discussion includes production methods, measurement techniques, available materials, their limitations, and recent advancements.

Multiscale Computational Protocols for Accurate Residue Interactions at the Flexible Insulin-Receptor Interface.

The quantitative characterization of residue contributions to protein-protein binding across extensive flexible interfaces poses a significant challenge for biophysical computations. It is attributable to the inherent imperfections in the experimental structures themselves, as well as to the lack of reliable computational tools for the evaluation of all types of noncovalent interactions. This study leverages recent advancements in semiempirical quantum-mechanical and implicit solvent approaches embodied in the PM6-D3H4S/COSMO2 method for the development of a hierarchical computational protocols encompassing molecular dynamics, fragmentation, and virtual glycine scan techniques for the investigation of flexible protein-protein interactions. As a model, the binding of insulin to its receptor is selected, a complex and dynamic process that has been extensively studied experimentally. The interaction energies calculated at the PM6-D3H4S/COSMO2 level in ten molecular dynamics snapshots did not correlate with molecular mechanics/generalized Born interaction energies because only the former method is able to describe nonadditive effects. This became evident by the examination of the energetics in small-model dimers featuring all the present types of noncovalent interactions with respect to DFT-D3 calculations. The virtual glycine scan has identified 15 hotspot residues on insulin and 15 on the insulin receptor, and their contributions have been quantified using PM6-D3H4S/COSMO2. The accuracy and credibility of the approach are further supported by the fact that all the insulin hotspots have previously been detected by biochemical and structural methods. The modular nature of the protocol has enabled the formulation of several variants, each tailored to specific accuracy and efficiency requirements. The developed computational strategy is firmly rooted in general biophysical chemistry and is thus offered as a general tool for the quantification of interactions across relevant flexible protein-protein interfaces.

Cross-scale dimensional measurement of high-precision modules based on machine vision

Abstract High-precision modules are of paramount importance in various fields such as machinery, automotive, aerospace, and IT electronics. Specifically, magnetic plates, as a type of high-precision linear module, are frequently employed in the manufacture of linear motors. Given the stringent requirements for dimensional measurement accuracy of high-precision components, manual measurement methods are inadequate for quality inspection of finished parts. Consequently, machine vision technology is widely used for measuring the dimensions of high-precision parts. To achieve the desired measurement accuracy, often at micro-nano levels, and to ensure the measurability of the magnetic plate’s overall dimensions, which range from millimeters to centimeters, a cross-scale measurement approach is essential. This approach must take into account the effects of depth of field (DOF) and the workpiece depth. The selection of an appropriate edge contour fitting method is critical in determining measurement accuracy. To address these challenges, our team constructed a machine vision system based on a mathematical model of DOF to resolve measurement challenges caused by the magnetic plate’s DOF issues. Furthermore, to improve edge fitting accuracy, we utilized the least squares method in combination with the Tukey weight function. This statistical approach allows for robust fitting by minimizing the influence of outliers and enhancing the precision of edge detection. Finally, the overall size of the magnetic plate is obtained through cross-scale calculation. Our findings reveal that the single-hole error and margin error within the magnetic plate are confined within a range of ±10 μm, the overall size error is within the range of ±90 μm. This level of precision is crucial for ensuring the quality and performance of linear motors. Implementing this method significantly enhances product detection efficiency on actual production lines, providing substantial improvements in quality assurance processes.

Deep Learning for Adrenal Gland Segmentation: Comparing Accuracy and Efficiency Across Three Convolutional Neural Network Models

Adrenal glands are vital endocrine organs whose accurate segmentation on CT imaging presents significant challenges due to their small size and variable morphology. This study evaluates the efficacy of deep learning approaches for automatic adrenal gland segmentation from multiphase CT scans. We implemented three convolutional neural network architectures (U-Net, SegNet, and NablaNet) and assessed their performance on a dataset comprising 868 adrenal glands from contrast-enhanced abdominal CT scans. Performance was evaluated using the Dice similarity coefficient (DSC), alongside practical implementation metrics including training and deployment time. U-Net demonstrated superior segmentation performance (DSC: 0.630 ± 0.05 for right, 0.660 ± 0.06 for left adrenal glands) compared to NablaNet (DSC: 0.552 ± 0.08 for right, 0.550 ± 0.07 for left) and SegNet (DSC: 0.320 ± 0.10 for right, 0.335 ± 0.09 for left). While all models achieved high specificity, boundary delineation accuracy remained challenging. Our findings demonstrate the feasibility of deep learning-based adrenal gland segmentation while highlighting the persistent challenges in achieving the segmentation quality observed with larger abdominal organs. U-Net provides the optimal balance between accuracy and computational requirements, establishing a foundation for further refinement of AI-assisted adrenal imaging tools.

Multimodal anomaly detection in complex environments using video and audio fusion

Due to complex environmental conditions and varying noise levels, traditional models are limited in their effectiveness for detecting anomalies in video sequences. Aiming at the challenges of accuracy, robustness, and real-time processing requirements in the field of image and video processing, this study proposes an anomaly detection and recognition algorithm for video image data based on deep learning. The algorithm combines the innovative methods of spatio-temporal feature extraction and noise suppression, and aims to improve the processing performance, especially in complex environments, by introducing an improved Variable Auto Encoder (VAE) structure. The model named Spatio-Temporal Anomaly Detection Network (STADNet) captures the spatio-temporal features of video images through multi-scale Three-Dimensional (3D) convolution module and spatio-temporal attention mechanism. This approach improves the accuracy of anomaly detection. Multi-stream network architecture and cross-attention fusion mechanism are also adopted to comprehensively consider different factors such as color, texture, and motion, and further improve the robustness and generalization ability of the model. The experimental results show that compared with the existing models, the new model has obvious advantages in performance stability and real-time processing under different noise levels. Specifically, the AUC value of the proposed model is 0.95 on UCSD Ped2 dataset, which is about 10% higher than other models, and the AUC value on Avenue dataset is 0.93, which is about 12% higher. This study not only proposes an effective image and video processing scheme but also demonstrates wide practical potential, providing a new perspective and methodological basis for future research and application in related fields.

BCD-YOLO: a railway perimeter foreign body intrusion detection method based on Yolov8

Abstract The railway perimeter is a key area for ensuring railway operational safety, and the intrusion of foreign objects seriously threatens the railway traffic safety. Considering the false and missed detections caused by the small size objects and the intricate background, this study proposed a detection approach called BCD-YOLO, based on YOLOv8. First, an enhanced dynamic sparse attention module was developed. It captures fine-grained details in the feature map and strengthens image context correlation, while reducing computational complexity. Second, the enhanced feature fusion module (CARAFE) was improved and T-CARAFE was constructed to replace the nearest neighbor linear interpolation method in YOLOv8, reducing the problem of feature loss during upsampling to improve the model’s capability of detecting small targets and feature catching capability. Finally, the DeepSORT object tracking algorithm was combined, which considered its trajectory and dwell time to determine whether an alarm is required, filtering out the effects of irrelevant objects in complex environments, and reducing the false alarm rate. The tracking ability of DeepSORT was further improved by introducing a shift window (Swin Transformer). In the experiment, the dataset defined a railway perimeter to simulate the actual railway perimeter environment, and only intruding objects within this perimeter were excluded. The results show that BCD-YOLO, compared with the original YOLOv8, has a precision improvement of 5.22%, recall rate improvement of 1.62%,mAP@0.5 improvement of 2.35%, and 105.5 frames per second. Combined with DeepSORT, the false detection rate is reduced by 16%, which satisfies the requirements of both accuracy and speed in railway object detection and effectively improves the problem of false and missed detections of small objects. Therefore, BCD-YOLO shows strong potential for deployment in intelligent railway monitoring systems, improving detection accuracy and reliability in complex environments while enhancing public safety by reducing intrusion-related risks.

Validation of the YuWell YE990 medical automatic electronic blood pressure monitor according to the Association for the Advancement of Medical Instrumentation/European Society of Hypertension/International Organization for Standardization Universal Standard (ISO 81060-2:2018/Amd.1:2020).

To validate the accuracy of the YuWell YE990 automated oscillometric upper-arm medical blood pressure (BP) monitor in adults according to the Association for the Advancement of Medical Instrumentation/European Society of Hypertension/International Organization for Standardization (AAMI/ESH/ISO) Universal standards (ISO 81060-2:2018 and Amendment 1:2020). Participants were recruited to meet the age, sex, BP, and cuff distribution criteria of the AAMI/ESH/ISO standards. BP was measured using a mercury sphygmomanometer (reference device) and YE990 (test device) following a same-arm sequential protocol with two trained observers and a supervisor. Bland-Altman plots were used to assess agreement, and scatter plots were used to assess the performance across varying arm sizes. Ninety-nine subjects were recruited and 85 were analyzed. The YE990 passed the accuracy requirements of the AAMI/ESH/ISO standards, with mean differences of 0.4 ± 6.31 mmHg for systolic BP and -0.9 ± 6.28 mmHg for diastolic BP for validation criterion 1. For validation criterion 2, the SD of the average BP difference between the test device and reference BP per subject was 5.00/5.70 mmHg (systolic/diastolic). YuWell YE990 meets the AAMI/ESH/ISO universal accuracy standard (ISO 81060-2:2018+Amd.1:2020) and is recommended for clinical use.

Research on scraper conveyor load prediction method based on wavelet transform and BP neural network

Scraper conveyor load prediction is crucial to realize the cooperative speed regulation of coal mining machine and scraper conveyor. In the synthesized mining face, due to the uncertainty of the coal fall, the load of the scraper conveyor fluctuates due to the change of the coal load, which shows a strong nonlinearity and non-smoothness, leading to the difficulty of prediction. To solve this problem, this paper proposes a BP neural network model combined with wavelet transform for scraper conveyor current prediction. By studying the mapping relationship between motor load and current based on the BP neural network algorithm, and taking the scraper conveyor current as the input condition, wavelet decomposition and data reconstruction of historical current data are carried out, and time series prediction is performed on the original data samples and reconstructed data samples, respectively. The simulation results show that the reconstructed BP neural network model using wavelet decomposition has higher prediction accuracy, in which the root mean square error is reduced by 13.26%, the average absolute error is reduced by 14.19%, and the percentage error is reduced by 17.43%. The model meets the accuracy requirements of scraper conveyor load prediction, and can provide theoretical basis for cooperative speed regulation of coal mining machine and scraper conveyor.

Comparative analysis of traditional machine learning Vs deep learning for sleep stage classification

Sleep stage classification is crucial for diagnosing sleep disorders and understanding sleep physiology. This study presents a comprehensive comparison between traditional machine learning algorithms and deep learning architectures using EEG recordings from the Physionet database. We extract 23 time and frequency domain features from each 30-second EEG segment and evaluate their performance across SVM, Random Forest, k-NN, and Gradient Boosting against CNN, LSTM, and hybrid CNN-LSTM models with attention mechanisms. Our results demonstrate that while traditional approaches achieve 82.4% accuracy with significant interpretability advantages, deep learning models reach 89.7% accuracy but require substantially more computational resources. The CNN-LSTM architecture with attention mechanisms performs best across all sleep stages, particularly for discriminating between similar stages like S1 and REM. However, traditional Random Forest classifiers offer competitive performance for resource-constrained applications with only 15% longer inference time. This comparative framework provides valuable insights for researchers and clinicians selecting appropriate methodologies for sleep analysis based on their specific requirements for accuracy, interpretability, and computational efficiency.

A deep learning framework with seismic-frequency-band constraints for thin-reservoir characterization

The characterization of the spatial structures of thin-layer sand bodies is the foundation for detailed reservoir description and physical property estimation. This requires using the geophysical inversion technique to make full use of the subsurface sedimentation information found in well logging and seismic data for comprehensive evaluation. However, the traditional model-based inversion method is limited by the frequency bandwidth of seismic data, and the resolution of the inversion results cannot meet the accuracy requirements of thin-layer reservoir characterization. In this paper, we introduce a convolutional model and seismic-frequency-band constraints into a deep learning framework to create a high-resolution inversion framework under physical constraints, and then test the proposed technical scheme in the model and on field data. We construct a 3D thin-layer sand body model with specific geological implications and carry out inversion tests based on sparse spike inversion, geostatistical inversion, and a deep learning framework. The experimental results demonstrate the feasibility of the deep learning framework and reveal its limitations in characterizing the spatial distribution of sand bodies. After introducing the convolutional model and seismic-frequency-band constraints, the inversion results more clearly distinguish the spatial distribution and superposition relationships of different sand bodies. The results obtained based on the field data fully confirm the effectiveness and applicability of the proposed method.

基于物联网技术的猪舍环境监测与综合评价系统研究

To address the issues of low precision and high cost in pig house environmental monitoring, an Internet of Things (IoT)-based pig house environmental monitoring system is proposed in this study. The system utilizes ESP32 as the main control chip of each sensor node, which is constructed in a star topology structure, realizing data transmission by using wireless communication technology. By incorporating the median filtering function and the Kalman filtering algorithm to perform data fusion of the same type of environmental parameters, the accuracy of the data is ensured. Based on the environmental suitability requirements of pigs, an evaluation index system for pig house environmental suitability is established, and comprehensive weight calculations are carried out by combining the entropy weight method and the improved analytic hierarchy process. Simulation experimental results demonstrate that the error of the temperature data after Kalman filtering fusion is merely 0.12%, fulfilling the monitoring accuracy requirements. The system can precisely evaluate the environmental suitability of pig houses, and it is applicable for monitoring pig house environments.

Accuracy of Large Language Models When Answering Clinical Research Questions: Systematic Review and Network Meta-Analysis.

Large language models (LLMs) have flourished and gradually become an important research and application direction in the medical field. However, due to the high degree of specialization, complexity, and specificity of medicine, which results in extremely high accuracy requirements, controversy remains about whether LLMs can be used in the medical field. More studies have evaluated the performance of various types of LLMs in medicine, but the conclusions are inconsistent. This study uses a network meta-analysis (NMA) to assess the accuracy of LLMs when answering clinical research questions to provide high-level evidence-based evidence for its future development and application in the medical field. In this systematic review and NMA, we searched PubMed, Embase, Web of Science, and Scopus from inception until October 14, 2024. Studies on the accuracy of LLMs when answering clinical research questions were included and screened by reading published reports. The systematic review and NMA were conducted to compare the accuracy of different LLMs when answering clinical research questions, including objective questions, open-ended questions, top 1 diagnosis, top 3 diagnosis, top 5 diagnosis, and triage and classification. The NMA was performed using Bayesian frequency theory methods. Indirect intercomparisons between programs were performed using a grading scale. A larger surface under the cumulative ranking curve (SUCRA) value indicates a higher ranking of the corresponding LLM accuracy. The systematic review and NMA examined 168 articles encompassing 35,896 questions and 3063 clinical cases. Of the 168 studies, 40 (23.8%) were considered to have a low risk of bias, 128 (76.2%) had a moderate risk, and none were rated as having a high risk. ChatGPT-4o (SUCRA=0.9207) demonstrated strong performance in terms of accuracy for objective questions, followed by Aeyeconsult (SUCRA=0.9187) and ChatGPT-4 (SUCRA=0.8087). ChatGPT-4 (SUCRA=0.8708) excelled at answering open-ended questions. In terms of accuracy for top 1 diagnosis and top 3 diagnosis of clinical cases, human experts (SUCRA=0.9001 and SUCRA=0.7126, respectively) ranked the highest, while Claude 3 Opus (SUCRA=0.9672) performed well at the top 5 diagnosis. Gemini (SUCRA=0.9649) had the highest rated SUCRA value for accuracy in the area of triage and classification. Our study indicates that ChatGPT-4o has an advantage when answering objective questions. For open-ended questions, ChatGPT-4 may be more credible. Humans are more accurate at the top 1 diagnosis and top 3 diagnosis. Claude 3 Opus performs better at the top 5 diagnosis, while for triage and classification, Gemini is more advantageous. This analysis offers valuable insights for clinicians and medical practitioners, empowering them to effectively leverage LLMs for improved decision-making in learning, diagnosis, and management of various clinical scenarios. PROSPERO CRD42024558245; https://www.crd.york.ac.uk/PROSPERO/view/CRD42024558245.

Insights into binary neutron star merger simulations: a multi-code comparison

Abstract Gravitational Wave (GW) signals from binary neutron star (BNS) mergers provide critical insights into the properties of matter under extreme conditions. Due to the scarcity of observational data, numerical relativity (NR) simulations are indispensable for exploring these phenomena, without replacing the need for observational confirmation. However, simulating BNS mergers is a formidable challenge, and ensuring the consistency, reliability or convergence, especially in the post-merger, remains a work in progress. In this paper we assess the performance of current BNS merger simulations by analyzing open-source GW waveforms from five leading NR codes – SACRA, BAM, THC, Whisky, and SpEC. We focus on the accuracy of these simulations and on the effect of the equation of state on waveform predictions. We first check if different codes give similar results for similar initial data, then apply two methods to calculate convergence and quantify discretization errors. Lastly, we perform a thorough investigation into the effect of tidal interactions on key frequencies in the GW spectrum. We introduce a novel quasi-universal relation for the transient post-merger time, enhancing our understanding of remnant dynamics in this region. This detailed analysis clarifies agreements and discrepancies between these leading NR codes, and highlights necessary improvements for the advanced accuracy requirements of future GW detectors.

Impact Analysis of Temperature Effects on the Performance of the Pick-Up Ion Analyzer

In deep-space exploration, Pickup Ion Analyzers (PUIAs) operate under varying thermal environments in orbit, where thermally induced stress–deformation coupling may severely degrade their performance and long-term stability. To address temperature field analysis for in-orbit PUIAs, in this study, we propose a coupled simulation framework integrating external heat flux, parallel temperature field calculation, and thermoelastic deformation analysis, establishing a systematic link from thermal inputs to performance analysis. Based on external heat flux results, a parallel LU decomposition algorithm reduced the computational time from 11.8 h to 2.9 h for rapid temperature field solutions. At 38 astronomical units (AUs), the instrument’s temperature distribution ranged from −45 °C to 51.13 °C, with simulation errors compared to COMSOL simulations meeting engineering accuracy requirements. Maximum thermoelastic deformation induced by thermal gradients reached 0.110 mm. Performance degradation due to deformation in key metrics—including ion energy resolution, angular resolution, detection field-of-view, geometric factor, and mass resolution—was below 7.2%. This research improves the computational efficiency of the temperature field and systematically quantifies temperature effects on PUIA performance in deep-space environments, and the proposed methodology could provide technical support for optimizing on-orbit thermal management strategies.

Energy efficient and robust node localization in WSNs using LSTM optimized DV hop framework to mitigate multihop localization errors

Wireless Sensor Networks (WSNs) are distributed sensor nodes that sense data from their surroundings and relay it to a central network for processing and analysis. Sensor localization is a crucial technique in WSNs, enabling precise positions of target nodes based on environmental signal perception. However, achieving high accuracy in node localization remains a challenge. This study introduces an improved DV-Hop positioning algorithm that integrates Long Short-Term Memory (OLSTM-DVHop) networks to enhance node position predictions. The algorithm processes original data through filtering, analysis, and feature extraction to improve predicted node positions. The study analyzed errors using a standard DV-Hop algorithm and designed a robust architecture for WSN positioning. Simulation experiments validated the proposed improvements, aligning with the algorithm’s accuracy requirements. The proposed error correction mechanism addresses uneven error distribution in the DV-Hop algorithm, adjusting the positions of nodes with significant deviations, reducing errors, and enhancing the positioning process’s reliability and accuracy. The effectiveness of the proposed algorithm is evaluated by comparing it with other localization algorithms across different terrain types. The improved DV-Hop algorithm significantly reduces localization errors and offers superior accuracy, outperforming other algorithms in various experimental scenarios.

Spatiotemporal UNet for multi-field spatiotemporal series generation

ABSTRACT Building multi-field spatiotemporal virtual environments is of great importance for industrial applications. However, due to the calculation complexity and the lack of consideration for spatiotemporal correlation, existing methods cannot meet the real-time and accuracy requirements. In this paper, we propose a novel Spatiotemporal UNet based on the three-dimensional convolutional neural network for the generation of multi-field spatiotemporal series. The MultiScale Attention Head is proposed to learn the multiscale spatiotemporal information. The Time InfoFusion module is proposed to mine the temporal correlation from spatiotemporal series. Moreover, considering the long-term periodicity and short-term stochasticity of spatiotemporal series, we propose the absolute time encoding strategy and spatiotemporal moving averages mean square error to optimize network learning. Experiments on three different regions of different physical fields show that the proposed method can generate virtual spatiotemporal environments in different tasks, with the RMSE decreasing 0.96–49.59% in 3-day lead time task, 0.16–49.97% in 5-day lead time task, and the MAE decreased up to 54.33% in 3-day lead time task and 55.30% in 5-day lead time task. The inference speed of Spatiotemporal UNet is 4.66 times of ConvLSTM, indicating its accuracy and real-time performance.

Accuracy index selection during optimization of particle accelerator ring tunnel network design based on Monte Carlo method

Abstract With advancements in science and technology, more countries are investing in fourth-generation light sources. The scale of these sources continues to expand, the requirements for alignment accuracy are also higher. With the increasing requirements of particle accelerators for alignment accuracy, the optimal design of tunnel control networks has become a core research issue. This study focuses on the 480-meter storage ring of the Hefei Advanced Light Facility (HALF). We systematically analyze the impact of laser tracker measurement distance and network density on the overall measurement accuracy of the control network based on the principles of three-dimensional measurement and network adjustment with software SpatialAnalyzer (SA). Using Monte Carlo simulation, three kinds of control network accuracy evaluation indexes (true value deviation, uncertainty, error) are compared. When using uncertainty and true value deviation as accuracy indicators, it is recommended to prioritize widening the laser tracker measurement range to optimize the layout of high-precision three-dimensional tunnel control networks and further refine network density. The results indicate that as the number of repeated measurements at control points increases, both deviation and uncertainty metrics improve, initially at a rapid rate that then gradually slows. Notably, the improvement in uncertainty is more pronounced than in deviation. After optimization, the HALF storage ring control network achieves a mean true value deviation of 66 μm and a mean uncertainty of 236 μm. The maximum deviation from the true value and uncertainty can reach 118 μm and 453 μm. This study provides a methodological framework for optimizing control networks in future particle accelerators and large-scale precision metrology applications.

Enhanced Dual Carry Approximate Adder with Error Reduction Unit for High-Performance Multiplier and In-Memory Computing

The Dual Carry Approximate Adder (DCAA) is proposed as an advanced 8-bit approximate adder featuring dual carry-out and carry-in full adders (FAs) along with an Error Reduction Unit (ERU) to enhance accuracy. The 8-bit adder is partitioned into upper and lower 4-bit blocks, connected via a dual carry-out full adder and a dual carry-in full adder. To minimize impact on the critical path, an ERU is designed for efficient error correction. Four variants of the DCAA are provided, allowing users to select the most suitable design based on their specific power, area, and accuracy requirements. The DCAA achieves a 78% reduction in Mean Error Distance (MED) while maintaining high computational speed and efficiency. When applied to Wallace Tree multipliers, it reduces delay by 32% compared to ripple carry adders (RCAs), and in in-memory computing (IMC) architectures, it significantly improves accuracy with minimal delay overhead. Experimental results demonstrate that the DCAA offers a well-balanced trade-off between accuracy, speed, and resource efficiency, making it suitable for high-performance, error-tolerant applications. Compared to existing approximate adders, DCAA exhibits superior error correction capabilities while achieving significantly lower delay. Furthermore, its efficient hardware implementation enables seamless integration into various computing paradigms, including AI accelerators and neuromorphic processors. Additionally, the scalability of the design allows for flexible adaptation to different bit-widths, making it a versatile solution for next-generation computing architectures.

High precision control moment gyroscope fault diagnosis via joint attention mechanism

The fault of one of the key systems in artificial satellites, the Control Moment Gyroscope (CMG), can lead to significant economic losses and irreparable consequences. Therefore, it is crucial to diagnose its faults promptly. Traditional fault diagnosis methods, however, face challenges such as local feature traps and difficulty in feature extraction when dealing with CMG vibration signals, making it hard to meet the requirements for accuracy and robustness. Hence, it is essential to design a high-accuracy model to assess the health status of CMG on time. To address these issues, a fault diagnosis method that combines the Joint Attention Mechanism (JAM) with one-dimensional dilated convolutional networks and residual connections is proposed. The method efficiently learns feature information through the JAM, effectively addressing the time-varying characteristics of vibration signals and focusing more on fault-related features. The influence of rotational speed on the model is overcome to some extent through JAM. The three rotational speeds are mixed as datasets, and the model achieves high accuracy. The proposed method significantly enhances the accuracy and robustness of CMG fault diagnosis. Experimental results on a self-collected dataset demonstrate that the proposed method achieves excellent accuracy (98.14%) and robustness in CMG fault diagnosis.

Efficient Monte Carlo Particle Transport Simulation for Hybrid Geometries Using an Intrusive Approach

Unstructured mesh (UM) is regarded as an attractive alternative to constructive solid geometry (CSG) in Monte Carlo (MC) simulations due to its adaptability to complex geometries and inherent advantages in conducting multiphysics coupling analysis with finite element codes. However, there generally exists a gap between practical solid models and transport-ready geometries due to the omission of background materials such as air and coolant. This necessitates additional effort to create computer-aided design models for background materials and to generate UM representations meeting accuracy requirements. Such limitations can restrict the applicability of MC simulations, particularly when background materials exhibit significant scale variations and narrow slots. Furthermore, the memory requirements and computational costs associated with UM-based MC simulations for large-scale problems may pose feasibility challenges. To mitigate these issues, this paper presents an intrusive approach to MC simulations within hybrid geometry. In the hybrid geometry, complex solid entities are represented using UM, while background materials or regular-shape geometries are modeled using CSG. This approach allows for arbitrary intersections between these two geometric representations by explicitly processing the topological relationships between CSG and UM objects during geometric initialization and particle tracking. The hybrid geometry–based particle tracking capability has been implemented in the MC code NECP-MCX. Validation studies were conducted against benchmarks, including the KRUSTY (Kilowatt Reactor Using Stirling Technology) component critical configurations and the ATR (Advanced Test Reactor) critical experiment. Numerical results underscore potential challenges encountered by the pure UM approach while demonstrating that hybrid geometry–based MC simulations can achieve results that closely align with experimental measurements in acceptable efficiencies.