Abstract: With the rapid expansion of the smart home appliance market, safety risks have become increasingly prominent. The integration of hardware and software, complex system architectures, and deep interactions in smart appliances contribute to diverse risk scenarios. Traditional risk assessment methods face challenges in multi-source data integration and multidimensional analysis. This study proposes a knowledge graph-integrated STPA-FMEA framework for smart appliance risk assessment, with a sweeping robot as a case study. The research establishes a unified risk assessment framework combining knowledge graphs, System-Theoretic Process Analysis (STPA), Failure Mode and Effects Analysis (FMEA). Three core contributions are presented: First, a domain-specific knowledge graph for sweeping robot risk assessment is constructed. Multi-source heterogeneous data are integrated using the seven-step ontology modeling method, with Protégé and Neo4j enabling visualization. Second, STPA-FMEA is applied for risk identification, leveraging Cypher queries and mapping rules to automate hierarchical control structure extraction and component failure mode analysis. A five-level risk assessment indicator system is developed, with risks quantified through an enhanced Risk Priority Number (RPN) method incorporating user impact and environmental factors. High-risk scenarios (e.g., mechanical entanglement, motor overheating) and critical components (e.g., batteries, charging docks, sensors) are identified, accompanied by targeted control strategies. Third, in the quantitative assessment phase of FMEA, an improved method for calculating the Risk Priority Number (RPN) was innovatively proposed, incorporating user impact and environmental impact coefficients. By integrating expert scoring and fuzzy evaluation methods, a comprehensive assessment of risk scenarios was conducted, yielding multiple RPN values. This approach enhances the alignment of evaluation results with actual conditions. In the results analysis section, the study examined traditional RPN, the influences of user and environmental factors, and improved RPN from multiple dimensions. This clarified high-risk scenarios and critical risk components, while revealing the differential impacts of various factors on risk scenarios. A thorough evaluation and analysis of potential risk scenarios in the robotic vacuum cleaner system were conducted, and comprehensive control strategies targeting high-risk scenarios were proposed from the perspective of control and feedback mechanisms. The framework identifies 34 risk scenarios, pinpoints critical components, and proposes mitigation strategies. Results demonstrate that knowledge graphs enable structured multi-source data integration, STPA-FMEA reveals systemic risk pathways through dual perspectives. This method not only provides a solid theoretical foundation for improving the safety and risk management of smart home appliances, but also demonstrates strong potential for broader application in the risk analysis of other intelligent systems and complex devices.