International audienceThis paper deals with robust fault detection for non-linear systems. This problem is usually solved by designing an observable subsystem which is only affected by the fault and not by the control and disturbance inputs. However, such a subsystem may not exist so that the so-called fundamental problem of residual generation (FPRG) is not solvable. The aim of the present paper is to design a fault detection filter when the conditions for the existence of a solution to the non-linear FPRG are not satisfied. Our approach is made in a geometric context. Under some decoupling assumptions, the design of sliding mode observers allows us to reconstruct the disturbance inputs and then to generate an effective residual. An illustrative example is given throughout the paper

No abstract

In this paper, a sensorless output feedback controller is designed in order to drive the Induction Motor (IM) without the use of ux and speed sensors. Firstly, an observer that uses only the measured stator currents is synthesized to estimate the mechanical variables (speed and load torque) and the magnetic variables (uxes) by structurally taking into account the unobservability phenomena of the Sensorless IM (SIM) and the parametric uncertainties. Secondly, a current-based eld oriented sliding mode control, that uses the ux and the speed estimates given by the former observer is developed so as to steer the estimated speed and ux magnitude to the desired references. Since the observer error dynamic is independent from the known input control and depends on the IM parametric uncertainties, a kind of separation principle is introduced to guarantee the practical stability of the whole closed-loop system "observer-controller" ("O-C") according to observability and unobservability time variation. A signicant benchmark taking into account the unobservability phenomena of the SIM is presented to show the performances of the whole control scheme against experimental setup .

No abstract