Simulation of a Neurofuzzy High Speed Estimation Applied to Magnetic Bearing Systems

The tendency to build light and high-speed rotors, to achieve a high performance of the machine, has as consequence a possible interaction between the rotor and its stator. During this interaction, the friction between the rotor and its stator, and the high energy of the former, can produce very complicated dynamical behavior. During contact the high energy of the rotor, which is dissipated by the frictional force, can severely damage both parts [1]. Estimating the rotor speed is an important task in conventional asynchronous motors. The elimination of the mechanical sensor as an encoder or a tachometer reduces the cost, improves the precision and the reliability of the whole system. Magnetic bearings systems always use a velocity sensor. In the case of failure of this sensor, rotor speed estimation could be used, in order to bring the rotor to still stand in a safety away, avoiding the situation mentioned above, therefore, the contact between the rotor and its retainer bearings. Several techniques are available in the literature for estimating the rotor speed of asynchronous motors. Recently, soft computing techniques are carried out to cover this proposal. The adaptive neuro-fuzzy inference system (ANFIS) technique has been successfully applied to estimating the speed of a conventional asynchronous motor. This work presents a simulation of a neuro-fuzzy speed estimator for a high speed asynchronous motor within a magnetic bearing system. As mentioned, the reliability of the system can be increased.