Diagnostic of an Aircraft by Different Dynamics of Operating Conditions

Abstract

This paper proposes an Input-Output Hidden Markov Model (IOHMM) to describe how the diagnostic of aircraft gas turbine engines can be estimated under multiple operating conditions. The PHM data challenge 2008 is used to model the system under the uncertainties of model parameters, operating dynamics, and the initial health situation. In this paper, multiple inputs are considered respective to the operating conditions. The thermodynamic simulation model generated the data of all sensors as a function of variations of flow and the efficiency of the modules concerned. The exponential rate of flow variation and efficiency loss was established in each data set, starting at a randomly selected initial deterioration set point. Well- known algorithms dedicated to Hidden Markov Model (HMM) are adapted to train different versions of IOHMM with the operating conditions. Finally, the best version of the model based on the most appropriate operating conditions for the system operation is selected to perform the diagnostic of the engine. The proposed method is validated by the cross-validation method to provide confidence over the model performance.