A hybrid information fusion approach to safety risk perception using sensor data under uncertainty

Abstract

This paper presents a hybrid information fusion approach that integrates the cloud model and the D–S evidence theory to perceiving safety risks using sensor data under uncertainty. The cloud model provides an uncertain transforming tool between qualitative concepts and their quantitative expressions and uses the measurement of correlation to construct Basic Probability Assignments. An improved evidence aggregation strategy that combines the Dempster’ rule and the weighted mean rule is developed to get rid of counter-intuitive dilemma existing in a combination of high-conflict evidence. A three-layer information fusion framework consisting of sensor fusion, factor fusion, and area fusion is proposed to synthesize multi-source information to get the final fusion results. The developed cloud D–S approach is applied to the assessment of the safety of a real tailings dam in operation in China as a case study. Data information acquired from 28 monitoring sensors is fused in a continuous manner in order to obtain the overall safety level of the tailings dam. Results indicate that the developed approach is capable of achieving multi-layer information fusion and identifying global sensitivities of input factors under uncertainty. The developed approach proves to perform a strong robustness and fault-tolerant capacity, and can be used by practitioners in the industry as a decision tool to perceive and anticipate the safety risks in tailings dams.