Residual useful lifetime prognosis and design of control systems
Résumé
This paper proposes to combine the deterministic behavior of a feedback control system with a stochastic degradation process. A modeling framework is suggested with an intrinsic deterioration caused by the control input and its variations. The applied command is seen as an environment covariate with states modifying the degradation level thanks to a Cox model. Consequently when degrading the actuator has a loss of eff ectiveness to carry out this command. Generally, an actuator is supposed to know several operational modes. For that purpose, two system missions are studied - slow and fast missions - with a Markov chain allowing to parameterize the sojourn times in the di fferent system states. A third order system is controlled in two ways i.e. a classical PID controller commonly used in industrial automation and an optimal control LQR. Both controllers are adjusted according to an expected residual useful lifetime. The system performances and its reliability are assessed aided with Monte Carlo simulations.