• 学术搜索
  • 科研智能体
    • Research Labs
    • AI 阅读
    • AI 文库
    • 深度研究
    • 学者亮点
  • 学术资源
    • AI2000
    • 期刊/会议
    • 学者库
    • 学术API
      MCP&Skill
    • 溯源树
    • 数据集
  • 知识沉淀
    • 学术空间
订阅小程序
旧版功能
aminer vip
开通会员低至0.73元/天
一次搞定AI科研
立即登录
  • English
  • 联系方式
    Journal of Guangdong University of Technology

    Journal of Guangdong University of Technology

    JournalISSN 1007-7162

    年发文量

    研究主题

    论文(4576)

    排序
    1A Knowledge-enhanced Large Language Model Intelligent System for Injection-molding Defect Diagnosis
    Hongjie Huang, Min’ao Zhong, Hongming Lai,Jianghao Lin, Zhigang Ren

    As one of the core processes in modern manufacturing, injection molding quality is critically dependent on dozens of process parameters, such as melt temperature and packing time. Common defects like flash, short shots, and silver streaks frequently arise during production. However, current industrial fault diagnosis heavily relies on the experiential knowledge of engineers and manual consultation of technical documentation. This approach is not only inefficient but also hinders the accumulation and transfer of specialized expertise. To address this issue, this paper integrates large language models and retrieval-augmented generation (RAG) technology to establish an efficient and reliable intelligent question-answering system for injection molding defect diagnosis for the first time. First, multi-source domain-specific corpora were collected to construct a high-quality instruction dataset, and the Qwen3-8B model was fine-tuned using low-rank adaptation (LoRA) for parameter efficiency, achieving domain adaptation for injection molding knowledge. Simultaneously, key modules were deeply enhanced and customized. To address the prevalence of complex documents in this field, the parsing strategy was transformed into a "layout-priority combined with regional optical character recognition" approach, significantly improving parsing accuracy and retrieval citation quality of industrial documents. Second, a text classifier was introduced as a pre-retrieval module to identify whether user queries pertain to the injection molding domain, thereby activating knowledge base retrieval only for professional inquiries and effectively boosting system response efficiency. Furthermore, an independent knowledge graph retrieval route was incorporated to achieve fused multi-source knowledge recall. Finally, the system was evaluated in detail using multi-dimensional metrics, including faithfulness, answer accuracy, context recall, recall-oriented understudy for gisting evaluation (ROUGE) , and BERTScore. Experimental results confirm that the proposed "fine-tuned model + enhanced retrieval" integrated system significantly outperforms traditional RAG systems and other baseline models across all evaluated metrics. This system offers an efficient and reliable intelligent solution for injection molding defect diagnosis, demonstrating substantial practical value and broad potential for industrial application.

    2026
    引用
    AI阅读
    加入学术空间
    2A Topology Optimization Algorithm for Minimizing Information Age in Two-hop Wireless Networks
    Bohao Chen,Qinglin Xie,Quan Chen,Yuan Chai

    Age of Information (AoI) has been proposed as a new metric to quantify data freshness from the point of destination. In this research, the problem of minimizing AoI in orthogonal frequency division multiplexing (OFDM) wireless networks is focused on. Since all existing works assume that the network topology is given and then propose the efficient scheduling algorithms on the basis of given topology, a study is focused on the work for AoI optimization with topology control in a two-hop OFDM-based wireless networks. Firstly, under the Poisson arrival service model, the mathematical expressions of AoI from the source node to the relay node and from the relay node to the aggregation node are derived, revealing the nonlinear relationship between AoI and network load rate. The lower bound of the average AoI of whole network is theoretically analyzed, and the optimal load rate is proved to be 0.457. Furthermore, a two-stage relay node selection algorithm is designed in the OFDM scenario, taking into account key parameters such as packet size, transmission distance, and channel rate, in which the optimal relay node is selected based on the load rate and theoretical optimal threshold. The simulation results show that the algorithm is significantly better than the baseline strategies under various network density, load, and data size conditions, with an average AoI reduction of 32 percentage point.

    2026
    引用
    AI阅读
    加入学术空间
    3A Point Cloud Completion Method Guided by Image Information of Missing Regions
    Xianhui Huang,Guangping Li

    View-guided point cloud completion (ViPC) enhances completion quality by incorporating image modality, yet existing methods often introduce background interference in cross-modal attention, underutilize image cues aligned with missing regions, and employ decoders with limited structural modeling capacity. To address these limitations, in this work, RGK (Reverse to Get the Key), a cross-modal guidance module composed of Reversed Cross Attention (RevCA) and a Missing Region Image-guided Completion Decoder (MRICD) is proposed. RevCA augments standard cross-attention through point-wise dynamic similarity gating and reversed attention redistribution to emphasize features associated with missing regions. MRICD further selects an auxiliary image token sequence based on RevCA and performs cross-attention-based interaction for accurate completion. Experiments on ShapeNet-ViPC show that integrating RGK into baseline networks consistently improves performance over baselines and competing methods, demonstrating its effectiveness in extracting key cross-modal features and restoring point cloud structures.

    2026
    引用
    AI阅读
    加入学术空间
    4A Research on Multimodal Method for Cooler Shelf Inventory Recognition
    Weiwei Zhang, Yujie Zhu, Bowen Wu,Zhijing Yang,Tianshui Chen

    This paper addresses the challenges of low efficiency in manual inventory counting and the weak generalization of existing visual methods in retail inventory management by proposing the cooler shelf inventory recognition (CSIR) framework. The framework takes multimodal inputs, encodes multi-angle images using a Vision Transformer, and aligns the resulting features to the latent space of the LLaMA (Large Language Model Meta AI) decoder through linear projection. A decoder-only language model is then constructed to enable end-to-end inventory information generation. This paper combines a domain-specific tokenizer that serializes shelf positions, product types, and inventory levels into discrete tokens to support autoregressive generation, and constructs a real-scenario dataset containing 17,000 samples covering complex conditions such as multi-view, reflective surfaces, and dense arrangements, with multidimensional evaluation metrics. Experimental results show that the proposed method achieves a tolerance-free overall accuracy of 70.17%, representing an approximately 10% improvement over the detection baseline, along with a 5.5-fold increase in inference efficiency. These results effectively reduce labor costs and inventory discrepancies, providing a scalable and reproducible reference solution for automated inventory management.

    2026
    引用
    AI阅读
    加入学术空间
    5Human-label-free Image Quality Assessment Via Bayesian Inference and Adaptive Transition Matrix
    Shenghao Chen, Wenchao Jiang, Tao Wu

    Blind image quality assessment (BIQA) models heavily depend on expensive human subjective ratings. To reduce this dependency, full-reference IQA (FR-IQA) metrics are used as surrogates for generating pseudo-labels. However, FR-IQA-based methods introduce substantial label noise that degrades assessment accuracy. In this research, an adaptive transition matrix learning-based BIQA (ATML-BIQA) method is proposed that learns from synthetically degraded images and multiple annotators without requiring human labels or suffering from noisy label effects. The method operates in four stages: first, distorted images are synthesized from high-quality references and paired randomly. Eight FR-IQA models then assign pseudo-binary labels indicating relative perceptual quality. Second, a CNN-based BIQA model is trained on these binary labels to generate Bayesian optimal labels. Third, an adaptive instance transition matrix (AITM) is designed to model the relationship between noisy pseudo-labels and optimal labels, capturing the correlation between noisy annotations and latent ground truth for effective label noise correction. Finally, the BIQA model is retrained with corrected labels to produce final quality scores. This research significantly reduces the reliance of BIQA on expensive manual annotations and overcomes the accuracy limitations of traditional proxy labels by adaptively modeling annotation noise. Experimental results demonstrate that while enhancing model generalization, this method provides an efficient and robust new paradigm for quality assessment tasks in unsupervised or weakly supervised environments.

    2026
    引用
    AI阅读
    加入学术空间
    立即登录,查看全部 4576 篇论文

    高被引作者

    作者引用发文
    DUAN Wei11571
    DENG Zhao11571
    Zhaokun Li11571
    ZhenFu Li11571
    Ruichu Cai4789
    Bingfeng Chen4764
    Jian Zhu4692
    Pei-yu Liu4681
    Weihua Li19831
    Qinruo Wang18922

    高产作者

    作者引用发文
    Weihua Li19831
    Renhuang Wang9025
    Shaohua Teng6525
    YongJian Liu15824
    Ying Chen3824
    XueQiang Yang4223
    Qinruo Wang18922
    Chaoyong Jin6122
    Renyuan Zhang10521
    ShengLin Zheng9320

    相关期刊

    Diagnostic Microbiology and Infectious Disease
    European Journal of Clinical Microbiology & Infectious Diseases
    Intervirology
    Le infezioni in medicina
    Annals of clinical microbiology and antimicrobials
    Infectious diseases
    Journal of Clinical Virology
    New Microbes and New Infections
    Future Virology
    Journal of Electronic Science and Technology