Research on Safety Analysis Method of Functional Integrated Avionics Systems

Abstract

Confronted with difficulties in the safety status analysis of avionics systems attributed to functional integration, a new safety analysis method based on a denoising-improved deep belief network, a deep learning algorithm, is proposed, realizing the mapping of system status parameters to overall system safety status within the context of multiple dimensions. By setting the aircraft environment surveillance system (AESS) as an instance, the engineering data was utilized to verify the feasibility of the method, yielding high classification performance. Comparative experiments with original DBN, DSAE, and NN demonstrated that the proposed method can achieve safety status identification directly from complex original data, and possesses strong classification robustness.