Aircraft System State Recognition and Fault Prediction Based on a Test Diagnostic Model

Abstract

The existing testability models for fault prognosis of aircraft systems limit the implementation of prognosis and health management systems. This paper develops a test diagnosis modeling method and relevant algorithms to support dynamic testing and to evaluate fault prognostic ability during aircraft system design. According to the system principles and the complex function structure of aircraft systems, a test diagnostic model is established by integrating testing and prognostic information with a test diagnostic skeleton model using multi-signal flow. New test indexes are identified to assess the testability and prognostic ability of aircraft systems. Relevant state recognition and fault prediction algorithms are established by fusing the improved particle swarm optimization algorithm and Hidden Semi-Markov Model. The feasibility and validity of the test diagnostic modeling method and relevant algorithms are verified in an aircraft’s engine bleed air system. Training and test show that the model can support analysis and estimation, and the algorithms can ensure accurate results after training the HSMM using improved PSO algorithm.