Abstract: In the field of Industrial Internet of Things, device entity discovery constitutes a critical component of device management. Compared to intelligent sensors, the discovery process of non-intelligent sensors is particularly complex due to their lack of inherent communication protocols, making efficient and accurate identification of non-intelligent devices a pressing technical challenge. The paper proposes a knowledge graph and large language model (LLM)-based efficient recognition scheme for non-intelligent sensors. The solution first constructs a comprehensive knowledge graph by extracting attribute values from non-intelligent sensors. Subsequently, the knowledge graph information is fed into a large language model for fine-tuning, with optimal model parameters obtained through experimental optimization. Test results demonstrate that the method achieves a recognition accuracy of 96%, significantly enhancing identification efficiency. The proposed scheme is recognized as applicable to large-scale IoT application scenarios due to its high efficiency and accuracy, demonstrating broad application value across various Internet of Things environments.

Key words: Keywords: Internet of Things, knowledge graph, large model fine-tuning, non-smart sensors, entity discovery

摘要： 在工业物联网领域，设备实体发现构成了设备管理的关键组成部分，相较于智能传感器，非智能传感器由于缺少内在的通信协议，其发现过程显得尤为复杂，如何高效且准确地识别非智能设备，成为了亟待解决的技术难题。本文提出了一种基于知识图谱和大模型的非智能传感器高效识别方案。该方案首先通过提取非智能传感器的属性值，构建了一个全面的知识图谱。随后，将知识图谱信息输入至大模型中进行微调，通过一系列实验优化，得到了最佳模型参数。实验结果表明，本方法的识别准确率高达96%，显著提升了识别效率。本文提出的方案因其高效性和准确性，被认为适用于大规模物联网应用场景，具有广泛的物联网场景应用价值。

关键词: 物联网, 知识图谱, 大模型微调, 非智能传感器, 实体发现