Input-output hidden Markov model to manage the remaining useful life of system under missing data
Résumé
This paper proposes a statistical model for diagnostic and prognostic system health by using the sensor data. Sometimes sensor misreads the observation for various reasons which contains one or more holes in the measured data and sensor saturation. The main contribution in this paper is to estimate and manage the remaining useful life (RUL) of the system considering multiple operating conditions under missing data. A recursive technique based on Input-Output Hidden Markov Model is proposed in this article for identifying the missing measurements or sensor saturation (ROSS) then predict the system failure at given operating conditions. An optimization performs based on the production speed that controls the operating conditions to manage the RUL. An example is given where the model parameters are estimated from the data set that has about 13% of missing data. The well-known algorithms are adapted in the model training and application.