Identification of key factors for elevator safety based on Bayesian complex network

As the core carrier of modern urban vertical transportation, elevator systems face safety issues primarily caused by cascading failures across multiple stages. This paper proposes an accident chain analysis framework based on Bayesian probabilistic complex networks. By integrating expert rules with historical data to construct a dynamic weight network, it quantifies the uncertainty of propagation paths and identifies critical vulnerable nodes. The Dirichlet prior is used to smooth low-frequency edge weights, combined with variational inference to optimize posterior parameters, and a multi-order conditional probability fusion algorithm is designed to simulate the incremental enhancement characteristics of fault propagation, optimizing Bayesian probabilistic weights. Experiments are conducted using national elevator accident cases to build a disaster chain network, combining PageRank and betweenness centrality for vulnerability analysis to identify key nodes, thereby ranking risk factors and determining high-risk propagation paths in accidents. Finally, a hierarchical blocking strategy is proposed to provide theoretical support for elevator safety prevention and control.