Risk Assessment of Seaplane Operation Safety Using Bayesian Network

Abstract

Seaplanes have become popular tourism and transportation tools with the ability of take-off and land on water. Recent seaplane accidents are highlighting the need for safety analysis of the seaplane operation process, which includes the sequential stages of water-taxiing, take-off, flight, and landing. This paper proposes a novel approach to modeling the risk of seaplane operation safety using a Bayesian network (BN). The rough risk factors that may cause seaplane accidents are identified by historical data, literature review, and interviews with experts. Based on the identification result, a risk evaluation indicator system is constructed and screened by the Delphi method. The structure of the proposed BN is derived from the indicator system. The parameter of the BN is obtained by expert experience and parameter learning from statistical data. The BN model is validated with an out-of-sample test demonstrating nearly 95% prediction accuracy of the accident severity level. The model is then applied to conduct diagnosis inference and sensitivity analysis to identify the key risk factors for seaplane operation accidents. The result shows that the four most critical risk factors are mental barrier, mechanical failure, visibility, and improper emergency disposal. It provides an early warning to take appropriate preventive and mitigative measures to enhance the overall safety of the seaplane operation process.