A new Bayesian network model for risk assessment based on cloud model, interval type-2 fuzzy sets and improved D-S evidence theory

Abstract

Traditional Bayesian network (BN) model is established by crisp sets and probabilities, and its effectiveness and applicability are restricted. In order to solve this problem, a new BN model for risk assessment based on cloud model (CM), interval type-2 fuzzy sets (IT2FSs) and the improved Dempster-Shafer (D-S) evidence theory was developed in the following route. Firstly, CM and IT2FSs were used to define a new solution for the digital characteristics, and the standard and index cloud membership functions for the IT2FSs-CM were established. Then, the index cloud membership function for the IT2FSs-CM was defined as the basic probability assignment (BPA) function, a new evidence fusion rule algorithm was defined, and the improved D-S evidence theory was proposed. Furthermore, the improved D-S evidence theory was used to transform the BPA function into the IT2FSs based prior probability, and a new BN model for risk assessment was developed. Finally, the BN model was applied to a nuclear power plant (NPP) construction project, its risk was assessed at level I (very low risk), and the results by IT2FSs-CM and Bayesian reasoning were compared with those by common methods. The findings show that the new BN model is effective and applicable in risk assessment.