A BN-based risk assessment model of natural gas pipelines integrating knowledge graph and DEMATEL

Abstract

As one of the most essential process facilities, the natural gas pipeline may be affected by various hazards and result in frequent accidents. Although Bayesian Network (BN) and other methods have been applied for risk assessment of gas pipelines, most attempts rely on experts instead of data. In this paper, a novel risk assessment model integrating Knowledge Graph (KG), Decision-Making Trial and Evaluation Laboratory (DEMATEL), and BN is proposed to analyze gas pipeline accidents in a data-driven way to minimize the reliance on experts for the current BN-based approach. First, the KG is used to extract and illustrate the causal network from accident reports on the Internet instead of a limited number of experts. Then, DEMATEL is applied to quantify the complex correlations in the causal network to simplify the topology structure and convert it into a BN structure. Moreover, by conducting BN analysis, a probabilistic causation model of gas pipeline accidents is established to identify critical hazards, predict potential consequences and optimize risk reduction strategies. The proposed model can more objectively support the safety management and risk reduction of natural gas pipelines and other process installations in the digital age.