Research on an Intelligent Training System for Highway Construction Safety Based on Cloud-Edge-End Collaboration and Knowledge Graph

Abstract

The highway construction environment is complex, with high personnel mobility and numerous potential hazards. Traditional safety training models commonly suffer from issues such as being overly formalized, having dull content, lacking specificity, and missing a closed loop for effectiveness evaluation, making it difficult to effectively enhance the safety awareness and skills of practitioners. To address the above problems, this paper proposes and develops a "Cloud-Edge-End Integrated Multimedia Safety Training Management System," similar to the system adopted in Shaanxi Province's highway construction projects. This system is based on a "cloud-edge-end" collaborative architecture and deeply integrates knowledge graph technology, AI recommendation algorithms, and IoT hardware technologies. On the cloud side, a highway construction safety knowledge graph covering the five-dimensional elements of "personnel, machinery, materials, methods, and environment" is constructed, and a personalized assessment and course recommendation engine is developed based on the Transformer model; this achievement is supported by research on the construction and application of a multimodal safety knowledge graph for construction based on large language models, while the application of BIM technology in highway construction safety management also provides a technical background for this work. On the edge side, a portable intelligent training toolbox integrating identity verification, multimedia teaching, and paperless assessment is developed, and seamless collection of training process data is achieved through a PPT plugin. On the terminal side, a WeChat mini-program for trainees is developed, supporting pre-/post-training assessments, personalized learning path recommendations, and interactive learning. This study elaborates in detail on the system's overall architecture, key technologies (including automatic knowledge graph construction, AOV-optimal learning path generation, and edge device integration), and business processes. The system has been successfully applied in the Hu-Zhou-Mei Highway project in Shaanxi Province. Practice demonstrates that the system effectively achieves precise, dynamic, and closed-loop management of safety training, significantly improving training efficiency and outcomes, and provides an innovative technical model and practical reference for safety production management in the field of transportation infrastructure construction.