A Torque Estimator for a Switched Reluctance Motor using an Orthogonal Neural Network
Résumé
Due to their simple construction, switched reluctance motors are very bad torque actuator. A precise knowledge of the produced torque in real time is necessary in order to design a high-performance speed-controlled motor drive. As the torque is difficult and expensive to measure, a method for estimating the torque of a switched reluctance motor at standstill is described. The non-linearity caused by the magnetic saturation and the rotor position is modeled with an artificial neural network allowing the method to be based on reference currents and measures of the position and currents only. The estimation objective is defined in terms of a reference trajectory that the neural network model output should match as closely as possible. The internal architecture of the network is obtained by using a serial of orthogonal functions. The internal parameters are deduced by using direct minimization of the error estimation. The obtained estimation is compared with the real torque. They are also compared with results given by other neural network architectures.
Domaines
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |