An application of a recently proposed fault detection and isolation (FDI) design methodology for linear parameter varying (LPV) systems is presented which concerns the robust detection of stator windings faults of an electrical induction motor regulated by a speed controller. On the basis of a detailed nonlinear mathematical model of the motor, it is shown how, based on a judicious convex interpolation of a family of linearized models, a quasi-linear parameter varying (quasi-LPV) approximation capable to catch most of the nonlinearities of the model can be achieved and can directly be used for synthesizing LPV-FDI observers. The design methodology consists in solving a multi-objective convex linear matrix inequalities optimization problem where disturbance rejection and fault sensitivity are traded-off in suitable frequency regions. The resulting diagnostic observer is gain-scheduled and uses a set of motor variables, assumed measurable online, as a scheduling vector. The effectiveness of the LPV-FDI framework is illustrated in a final numerical example.
LPV FDI SCHEMES FOR INDUCTION MOTORS
631Notice that in the quasi-LPV approach, each scheduling term is allowed to be a suitable function of the state and input, that is i .t / D i .x.t /; u.t //. The term 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 Time [s] J r1 (t) J th1 =0.2514 J r1,LPV (t) J r1,LIN (t) 0 1 2 3 4 5 6 7 8 9 10 0 0.1 0.2 0.3 0.4 Time [s] J r2 (t) J th2 =0.056 J r2,LPV (t) J r2,LIN (t) 0 1 2 3 4 5 6 7 8 9 10 0 0.1 0.2 0.3 0.4 Time [s] J r3 (t) J th3 =0.02512 J r3,LPV (t) J r3,LIN (t) Figure 10. First Experiment: rotor angular speed sensor fault (f 1 .t/). 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 Time [s] J r1 (t) J th1 =0.2514 J r1,LPV (t) J r1,LIN (t) 0 1 2 3 4 5 6 7 8 9 10 0 0.1 0.2 0.3 0.4 Time [s] J r2 (t) J th2 =0.056 J r2,LPV (t) J r2,LIN (t) 0 1 2 3 4 5 6 7 8 9 10 0 0.1 0.2 0.3 0.4 Time [s] J r3 (t) J th3 =0.02512 J r3,LPV (t) J r3,LIN (t)