Fault diagnosis for satellite sensors and actuators using nonlinear geometric approach and adaptive observers

Abstract

Summary This paper presents a novel scheme for diagnosis of faults affecting sensors that measure the satellite attitude, body angular velocity, flywheel spin rates, and defects in control torques from reaction wheel motors. The proposed methodology uses adaptive observers to provide fault estimates that aid detection, isolation, and estimation of possible actuator and sensor faults. The adaptive observers do not need a priori information about fault internal models. A nonlinear geometric approach is used to avoid that aerodynamic disturbance torques have unwanted influence on the fault estimates. An augmented high‐fidelity spacecraft model is exploited during design and validation to replicate faults. This simulation model includes disturbance torques as experienced in low Earth orbits. This paper includes an analysis to assess robustness properties of the method with respect to parameter uncertainties and disturbances. The results document the efficacy of the suggested methodology.

An integrated fault estimation and fault tolerant control method using H∞‐based adaptive observers

“…According to this assumption, the generic input fault is supposed to be differentiable after its occurrence which is also considered in related works. 31,[35][36][37] In fact, it states that the given estimation accuracy is not guaranteed about the faults with impulsive behavior or at particular times that sudden changes occur. However, it should be noted that this assumption is not imposed a particular restriction and may include a wide range of practical defects.…”

“…Based on the mentioned theorem, the disturbance effects are attenuated by employing the H ∞ optimization developed in (18)- (20). It removes the local diffeomorphisms as required in the work of Baldi et al 19 Also, the robustness provided by the interconnected use of the SMO and adaptive observers in the work of Zhang et al 20 suffers from the chattering phenomenon and imposes additional nonlinear transformations.…”

“…The restriction given in Equation 6is convenient in most works in which the adaptive observers have been proposed. [37][38][39] According to lemma 3 in the work of Raufi et al , 38 it is proved that the Equation 6is satisfied if and only if rank(CE) = rank(E). The above condition states that E should be a full-column rank and the number of independent rows of the matrix C should not be less than the number of independent columns of the matrix E. In other words, it means that the number of independent measurements of the system should not be less than the size of input faults, which of course it has been considered in many observer schemes.…”

“…There exist some solutions in the literature to overcome this issue. In the work of Baldi et al 19 a joint use of AO and a geometric approach was proposed. Nevertheless, some local diffeomorphisms are necessary to obtain an isolated subsystem that is decoupled from the disturbances.…”

“…It can precisely identify the fault shape by attenuating disturbance effects without the SPR restriction or PE condition as emphasized in the related manuscripts. [14][15][16][17][18] It avoids the local diffeomorphisms or nonlinear transformations, which are necessary in the works of Baldi et al 19 or Zhang et al 20 Furthermore, due to direct estimation of fault by the developed method, it is not required to define auxiliary state variable and obtain augmented dynamic models as it can be seen in Wang et al 29 or Lan and Patton. 30,31 In the observer design synthesize, it is only assumed that the fault magnitude and its variations are bounded and a priori knowledge of the related upper bounds is not required.…”