课程文献中心 更多>>

关键词： Deep learning   Feature extraction   Fourier transform infrared spectroscopy   Gradient-weighted class activation mapping   Microplastic identification  

摘要： To address the low accuracy of spectral identification of mixed microplastic samples and the “black box” nature of deep learning models, this paper proposes an optimization method based on Gradient Weighted Class Activation Mapping (Grad-CAM). Model interpretability analysis using Grad-CAM reveals an imbalance in the feature extraction process—insufficient attention to the features of minor components in mixed samples. Based on these findings, multi-head attention and multi-region attention mechanisms are introduced to enhance the model's ability to extract global and local features. Considering the complex and diverse ATR-FTIR spectral features of microplastics and the gradient vanishing problem in deep learning models, Resblock residual connections are introduced on top of CNN, and a ResM2A-CNN model is designed. Experimental results show that the overall accuracy of the optimized model on the test set is improved from 95.67% to 99.92%. Notably, the ResM2A-CNN model exhibits strong classification ability for PA and its mixtures, outperforming other traditional models in various performance metrics.This study highlights the effectiveness of using interpretability analysis to guide model optimization, providing a breakthrough method for the rapid and accurate detection of microplastics. Furthermore, this research offers practical insights into modeling complex spectral data and optimizing deep learning models.

关键词： Symbolic regression   CAD-free   Lifting surface   Mass modeling   Volume computation   Deep neural networks  

摘要： Accurate structural mass estimation of lifting surfaces is required throughout aircraft design, yet existing methods cannot simultaneously provide closed-form geometric descriptors, physically consistent similarity scaling, and exact differentiability without repeated three-dimensional geometry instantiation. This work presents a symbolic-regression (SR) framework for segment-level mass estimation of solid, skin-laminated, shell, and wing-box lifting surfaces, using external surface area and enclosed volume as governing mass descriptors and quasi-analytical rib and spar volume expressions for internal contributions.

关键词： CAD   cinnamyl alcohol dehydrogenase   C4H   cinnamate 4-hydroxylase   CCoAOMT   caffeoyl-CoA O-methyltransferase   CCR   cinnamoyl-CoA reductase   CDS   coding sequences   CE   catechin equivalent   COMT   catechol-O-methyltransferase   4CL   4-coumarate-CoA ligase   DLR   dual-luciferase report   EMSA   electrophoretic mobility shift assay   ERF   ethylene response factor   GA   gallic acid   PAL   phenylalanine ammonia-lyase   PCR   polymerase chain reaction   P. digitatum   Penicillium digitatum   P. longanae   Phomopsis longanae Chi   POD   peroxidase   PPO   polyphenol oxidase   RT-qPCR   real-time quantitative PCR   SA   salicylic acid   SAR   systemic acquired resistance   TF   transcription factor   Y2H   yeast two-hybrid   Dimocarpus longan Lour   Salicylic acid   Disease resistance   Phenylpropanoid pathway   ERF transcription factor   Molecular mechanism  

摘要：

• •
  ● Salicylic acid (SA) treatment reduces disease occurrence in longan fruit.
• •
  ● SA treatment induces longan resistance by activating phenylpropanoid metabolism.
• •
  ● DlERF98 expression is upregulated by SA treatment.
• •
  ● DlERF98 targes phenylpropanoid metabolism-relates genes.
• •
  ● DlERF98 improves disease resistance in longans against Phomopsis longanae Chi.

关键词： 3D reconstruction   3D Gaussian Splatting   Computer-aided design (CAD)   Editable reconstruction   Industrial assembly   Assembly preview  

摘要：

• •
  Proposed a CAD-guided editable 3D Gaussian framework for assembly preview.
• •
  Used ALIKED-NN-assisted SfM to improve sparse reconstruction stability.
• •
  Built a part-level 3DGS pipeline with segmentation and point-cloud filtering.
• •
  Enabled editable assembly previews from CAD priors and real-world images.

关键词： BEM   Building Energy Modeling   CNN   Convolutional Neural Network   GCN   Graph Convolutional Network   XAI   Explainable Artificial Intelligence   CAM   Class Activation Mapping   EPC   Energy Performance Certificate   EUI   Energy Use Intensity   MLP   Multi-Layer Perceptron   RMSE   Root Mean Squared Error   CVRMSE   Coefficient of Variation of the Root Mean Squared Error   MAE   Mean Absolute Error   MLR   Multiple Linear Regression   RF   Random Forest   XGBoost   eXtreme Gradient Boosting   SHAP   SHapley Additive exPlanations   S/V   Surface-to-Volume Ratio   Energy modeling   Floor-plan morphology   Multimodal learning   Explainable diagnostics   Residential energy performance  

摘要： Early-stage energy modeling is crucial for optimizing residential building design. However, current approaches often face challenges related to design uncertainty, incomplete input parameters, and limited computational efficiency. This study develops a multimodal surrogate modeling framework for early-stage building energy modeling and diagnostics by integrating architectural floor-plan images with numerical energy-efficiency features. A convolutional neural network (CNN) is used to extract visual spatial features directly from floor-plan layouts, while a multi-layer perceptron (MLP) processes numerical design parameters. These two feature streams are then fused within a hybrid CNN–MLP architecture for energy use intensity prediction. The proposed model adopts a multi-scale convolutional structure without pooling operations, allowing it to preserve fine-grained spatial information in floor-plan images. Validation using more than 3,000 residential design samples shows that the proposed model outperforms single-modal machine learning baselines, including Random Forest and XGBoost, as well as multimodal deep learning models based on ViT and EfficientNet feature extractors. To enhance model interpretability and provide design-oriented feedback, SHAP and class activation mapping (CAM) techniques are incorporated. The results indicate that external boundaries, corner geometries, and geometric complexity play important roles in residential energy performance. Geometry-based metrics, including a modified compactness ratio (O/A) and an auxiliary external boundary ratio (aux_O/A), are further introduced to support spatial diagnostics. To examine model transferability, this study conducts a transfer learning experiment using residential floor-plan data from Birmingham. The results show that fine-tuning the model with a small subset of target-city samples can effectively correct cross-city prediction bias, suggesting the potential of the proposed framework for adaptation to dif

关键词： Industrial optical metrology   Multi-view stereo vision   CAD-guided measurement   Chamfer-edge extraction   Circular-holecontourreconstruction  

摘要： High-precision three-dimensional measurement of chamfered circular holes remains challenging in industrial environments due to chamfer interference, edge ambiguity, reflective surfaces, and multi-view reconstruction errors. To address these issues, this paper proposes a CAD-guided robotic multi-view stereo vision framework for automated 3D measurement and localization of chamfered circular holes. First, a photogrammetry-based multi-view pose estimation method is developed to establish accurate geometric relationships among different viewpoints. Then, an efficient viewpoint-planning strategy with virtual visual feedback is introduced to optimize robot poses and acquisition trajectories using CAD prior information. To achieve reliable chamfer-edge extraction, a multi-stage edge detection and layered clustering-based filtering framework is proposed. Furthermore, an ellipse matching method combining prior projection and multi-constraint verification is designed to improve correspondence robustness across multiple views. Finally, a unified multi-view 3D contour reconstruction and optimization method is proposed to jointly reconstruct hole geometry from all edge points, thereby reducing cumulative reconstruction errors and improving global geometric consistency. Experimental results demonstrate that the proposed framework achieves high accuracy, robustness, and efficiency under different chamfer angles, hole sizes, materials, and illumination conditions, showing strong potential for complex industrial hole measurement applications.

关键词： Computer-aided design   Generative artificial intelligence   3D model generation   Mechanical engineering   Design automation  

摘要： Computer-aided design (Cad) solid modeling lies at the core of product design in mechanical engineering, enabling designers to transform conceptual design ideas into precise, parametric, 3D Cad models. Automating solid modeling has the potential to accelerate engineering workflows, foster innovation, and lower barriers to design for non-expert users. Recently, increasing efforts are being devoted to leveraging learning-based methods to address this challenge. To the best of our knowledge, no existing survey provides a comprehensive overview of learning-based methods for 3D Cad model generation, nor addresses their applications in mechanical engineering. For that reason, we survey literature comprising state-of-the-art, learning-based methods for 3D Cad model generation, with particular emphasis on their relevance to mechanical engineering. Based on the type of input to the respective methods, we define three tasks and classify the literature accordingly: Cad Generation, Cad Reconstruction, and Cad Reverse Engineering. To provide a structured and comprehensive overview, we further organize the literature according to its types of output and the methodologies applied. Additionally, this survey highlights current research gaps, including limitations in fine-grained user control and geometric detail preservation, and discusses potential directions for future research. Finally, we present one potential use case for each of the aforementioned tasks in the mechanical engineering workflow, aiming to bridge the gap between existing methods and real-world applications. These use cases include design space exploration, quality control, and data enhancement.

关键词： CAD-CAM   Maryland   Survey   Digital technologies   Dentists  

摘要： This is a cross-sectional survey, investigating the adoption of Computer-Aided Design and Computer-Aided Manufacturing (CAD-CAM) technology among general dentists working in private practices. It explores the prevalence of restorative CAD-CAM utilization and describes self-reported perceptions of the technology.

关键词： Anthocyanins   MAPK Signaling Pathway   JNK/p38 Inhibition   ERK1/2 Activation   Inflammaging   Chronic Diseases   Clinical Trials   4-HHE   4-Hydroxyhexenal   Aβ   Amyloid beta-protein   Amyloid β-protein   ACNs   AD   Alzheimer‘s disease   ADP   Adenosine diphosphate   AGEs   Advanced glycation end products   AI   Artificial intelligence   ALI   Acute lung injury   AMPK   AMP-activated protein kinase   An-NPs   Anthocyanin-loaded nanoparticles   AP-1   Activator protein-1   ApoA-I   Apolipoprotein A-I   ApoB   Apolipoprotein B   APP   Amyloid precursor protein   ARDS   Acute respiratory distress syndrome   ARPE-19   Human retinal pigment epithelial cell line   ASK1   Apoptosis signal-regulating kinase 1   BACE1   Beta-secretase 1   Bax   Bcl-2-associated X protein   BBT   Black soybean testa extract   Bcl-2   B-cell lymphoma 2   BDNF   Brain-derived neurotrophic factor   C3G   Cyanidin-3-glucoside   C35G   Cyanidin-3-diglucoside-5-glucoside   C3OG   Cyanidin-3-O-glucoside   C3OR   Cyanidin-3-O-rutinoside   CAD   Coronary artery disease   CAT   Catalase   CBE   Chinese bayberry extract   CCL2   Chemokine ligand 2   CDRE   Cyanidin- and delphinidin-rich extract   CEC   Cholesterol efflux capacity   CERAD   Consortium to Establish a Registry for Alzheimer ‘s Disease   C/EBPα   CCAAT/enhancer-binding protein alpha   COMT   Catechol-O-methyltransferase   COX-2   Cyclooxygenase-2   CREB   cAMP response element-binding protein   CRP   C-reactive protein   CVD   Cardiovascular disease   CXCL5   Chemokine (C-X-C motif) ligand 5   CXCL7   Chemokine (C-X-C motif) ligand 7   CXCL8   Chemokine (C-X-C motif) ligand 8   D3G   Delphinidin-3-glucoside   D3OR   Delphinidin-3-O-rutinoside   DHA   Docosahexaenoic acid   DUSP1   Dual-specificity phosphatase 1 (also known as MKP-1)   DTC   direct-to-consumer   eNOS   Endothelial nitric oxide synthase   ECM   Extracellular matrix   EGFR   Epidermal growth factor receptor   ERK   Extracellular signal-regulated kinase   EU   European Union   FAK   Focal adhesion kinase   FBG   Fasting blood glucose   GLUT4   Glucose transporter type 4   GlycA   Glycoprotein acetyls   GPx   Glutathione peroxidase   GSH   Glutathione   GSK3β   Glycogen synthase kinase 3 beta   HA   Hyaluronic acid   HbA1c   Glycated hemoglobin   HDL-C   High-density lipoprote  

摘要：

• •
  Anthocyanins (ACNs) achieves health promoting effect through MAPK pathway.
• •
  Health promoting effect of ACNs was validated from basic research to clinical trial.
• •
  Structure-activity relationship between ACNs and health benefit effects was unveiled.
• •
  Some health benefit products based on ACNs have been developed.
• •
  Future studies focus on pharmacokinetic advance of ACNs in drug-delivery approach

关键词： Alzheimer's disease   CHA   DNA breathing   Lateral flow   miR-106b-5p  

摘要： The early, non-invasive diagnosis of Alzheimer's disease (AD) remains a significant clinical challenge. MicroRNAs carried by L1CAM-immunocaptured small extracellular vesicles (sEVs) in blood, particularly miR-106b-5p, have emerged as promising candidate biomarkers for AD. However, their detection requires methods that are sensitive, specific, stable, and suitable for point-of-care (POC) applications. Catalytic hairpin assembly (CHA) combined with lateral flow immunoassay (LFIA) offers a potential POC solution, but its practical application is limited by high background leakage caused by “DNA breathing” and poor probe stability. Herein, we propose an integrated POC platform that combines immunomagnetic enrichment of L1CAM-positive sEVs with a novel “Fold-Adsorb-Block-Immobilize-Lyophilize” (FABIL) process for CHA probe stabilization. The FABIL process utilizes magnetic graphene oxide (MGO) to adsorb and physically separate hairpin probes (H1 and H2), thereby suppressing nonspecific hybridization and generating lyophilized CHA microspheres with room-temperature stability. When integrated into a lateral flow test strip, the FABIL-NE-miR platform achieves a visual detection limit of 10 fM and a liquid-phase limit of detection of 1.14 fM, overcoming the traditional trade-off between low background and high sensitivity. Clinical validation using 100 serum samples, including 32 AD cases and 68 non-AD controls, demonstrated that FABIL-NE-miR outperformed traditional low-background CHA systems, with an area under the ROC curve (AUC) of 0.9563, a sensitivity of 90.63%, and a specificity of 94.12%. Overall, the FABIL-NE-miR platform enables rapid visual detection of low-abundance AD-associated miRNAs from L1CAM-immunocaptured sEV fractions. By addressing key challenges such as background leakage and probe instability, this study provides a practical tool for AD screening and supports the feasibility of the FABIL strategy as a universal platform for next-generation nucleic acid