This paper is concerned with a nonlinear, computationally efficient method for trending component health. While conceptually simple, the goal of component trending is to reduce spurious noise in the measured component health and to estimate the remaining useful life (RUL). The need for lower computational effort allows this to be done on an embedded system. This would be important for display on a cockpit multi-function display. Additionally, we describe a new method for state smoothing. This is a forward-backward technique with no computational overhead associated with updating the plant noise (associated with a Kalman Smoother), significantly reducing the number of operations needed. Finally, a comparison is made between the precision of the nonlinear state estimation vs. a linear state estimation in the computation of the RUL. The precision was quantified using mean relative of the prognostic.
展开▼
