Abstract:

To effectively identify and control quality and safety risk of intelligent consumer products, a KG for intelligent consumer product quality and safety risk was constructed based on LLMs. Complex network analysis methods were integrated to systematically analyze risk distribution characteristics and propagation mechanisms. Data on intelligent consumer products were collected through multiple channels; the quality and safety risk issues in product accident cases were sorted out. Based on safety theories and the system structure of intelligent products, a multi-level and extensible knowledge ontology for quality and safety risks was constructed. Under the constraints of this ontology, LLMs were utilized to achieve automated knowledge extraction, and a KG containing 16 611 nodes and 32 178 edges was constructed. The KG was mapped into a risk network. Based on complex network theory, centrality indicators of three key node types—hazard factors, accident injuries, and safety events—were calculated to identify critical nodes and propagation paths in risk transmission. The research results show that among various intelligent consumer products, intelligent home products account for the highest proportion of risk entities; physical hazards and information hazards are the main risk types; property damage and physical injury are the primary accident consequences. Typical direct propagation paths and cascading propagation paths can be identified by the risk network.

Key words: large language models(LLMs), knowledge graphs(KG), intelligent consumer products, quality and safety risk, complex network