随着施工机械化水平的提升，塔式起重机（简称塔机）在工程建设中的应用日益广泛，但塔机安全事故频发的问题却未得到有效改善。大量研究表明，吊装作业人员的操作行为是影响吊装作业安全的关键因素。因此，有必要聚焦吊装作业人员的操作行为，围绕吊装不安全事件的识别和预警开展研究。
本研究通过分析塔机安全事故，厘清影响吊装作业安全的关键因素，对相关标准规范中的典型吊装不安全事件进行梳理。为加强吊装作业人员和管理人员对吊装作业状况的感知能力，本研究提出塔机吊装作业数字孪生虚体构建方法，梳理建模关键要素。并在实验室条件下，以缩尺塔机为实验对象，验证吊装作业数字孪生虚体构建方法的有效性，实现由实向虚的映射。在此基础上，本研究以塔机操作使用相关标准规范为主要知识来源，构建吊装不安全事件本体模型，从而建立知识图谱，实现吊装不安全事件的语义化和结构化。并将知识图谱中的实体和关系转化为计算机可理解的逻辑语言，在吊装作业数字孪生中嵌入吊装不安全事件预警方法，实现由虚向实的反馈，以规范吊装作业人员的操作行为。
本研究利用实验室缩尺塔机对所构建的吊装作业数字孪生进行实验验证。通过物理数据与孪生数据的比较，发现物理数据与孪生数据平均偏差最高不超过 7%，误差相对稳定，说明吊装作业数字孪生虚体能够有效反映吊装作业物理实体的运动状况。在预警实验中， 正确识别整体吊装事件的准确率约为 91.2%，正确识别整体吊装不安全事件的灵敏度约为 94.1%，错误识别整体吊装安全事件的误警率约为11.8%，证明本研究所构建的预警方法能有效对吊装不安全事件进行识别和预警。
本研究所构建的基于数字孪生的塔机吊装不安全事件预警方法，不仅为数字孪生和知识图谱在吊装作业中的应用提供理论基础，也对减少塔机安全事故、提高吊装作业安全管理水平、促进建筑业智能化发展具有重要意义。

关键词：塔式起重机；数字孪生；吊装作业；安全预警；知识图谱

With the advancement of construction mechanization, tower cranes have become increasingly prevalent in construction projects, resulting in a growing number of tower crane-related accidents. A substantial body of research indicates that the operational behavior of tower crane-related construction personnel is a critical factor influencing the hoisting safety. Therefore, it is necessary to focus on the operational behavior of tower crane-related construction personnel and conduct research on the monitoring and warning of unsafe hoisting events.
The research analyzes tower crane accidents, identifies key factors affecting hoisting safety, and reviews typical unsafe hoisting events in relevant standards and specifications. To enhance the perception abilities of tower crane-related construction personnel and management regarding crane operations, the research introduces a method to construct a virtual entity for tower crane hoisting operations, combs through the key elements of modeling, and verifies the effectiveness of the method by taking the scaled-down tower crane as an experimental object under laboratory conditions to realize the mapping from the physical to the virtual. Based on this foundation, the research primarily relies on the relevant standard specifications for tower crane operation to construct an ontology model of unsafe hoisting events. Consequently, a knowledge graph of unsafe hoisting events is established, achieving semantic and structural organization of these events. The entities and relationships within the knowledge graph are transformed into computable logical languages. The warning method for unsafe hoisting events is embedded into the digital twin of hoisting operations. This enables feedback from the virtual to the real, standardizing the operational behaviors of tower crane-related construction personnel.
This research conducts experimental verification of the constructed digital twin of hoisting operations using a scaled-down tower crane in the laboratory. By comparing the physical data with the digital twin data, it is found that the average deviation between the physical data and the digital twin data does not exceed 7%, and the error is relatively stable. This indicates that the digital twin of hoisting operations can effectively reflect the motion status of the physical entity involved in the hoisting operations. In the experiment on the warning of unsafe hoisting events, the accuracy rate is approximately 91.2%, the sensitivity is about 94.1%, and the false alarm rate is about 11.8%. This demonstrates that the warning method developed in this research can effectively identify and alert unsafe hoisting events.
The early warning method for unsafe hoisting events of tower cranes based on digital twins not only provides a theoretical foundation for the application of digital twins and knowledge graphs in hoisting, but also holds significant importance in reducing tower crane safety accidents, enhancing the safety management level of hoisting, and promoting the intelligent development of the construction industry.