An increasing number of industrial applications require compact motors delivering high torque and power with maximum efficiency. If properly designed, Permanent Magnet Synchronous Motor (PMSMs) are capable to satisfy this needs. To fully unlock such capabilities, knowing and controlling precisely the thermal state of the motor is fundamental. In fact, high temperatures can be critical as they may cause insulation melting or magnets demagnetization. Due to the rotation, the direct measurement of rotor's inner temperatures is an expensive and complex option, and we here estimate it from more accessible quantities. In particular we will show by experiments on an actual motor that estimations within few degrees of error can be obtained by feeding readings of external temperatures and of the motor electrical status into suitably trained neural architectures.
Temperature Sensors Virtualization in High Performance Electric Motors / Zanellini A.; Pellegrini S.; Nerone M.; Valic I.; Zauli M.; De Marchi L.; Matteazzi N.; Violi M.; Rovatti R.. - ELETTRONICO. - (2023), pp. 99-104. (Intervento presentato al convegno 3rd IEEE International Workshop on Metrology for Automotive, MetroAutomotive 2023 tenutosi a Palazzo Ducale of Modena, ita nel 2023) [10.1109/MetroAutomotive57488.2023.10219114].
Temperature Sensors Virtualization in High Performance Electric Motors
Zanellini A.;Nerone M.;Valic I.;Zauli M.;De Marchi L.;Rovatti R.
2023
Abstract
An increasing number of industrial applications require compact motors delivering high torque and power with maximum efficiency. If properly designed, Permanent Magnet Synchronous Motor (PMSMs) are capable to satisfy this needs. To fully unlock such capabilities, knowing and controlling precisely the thermal state of the motor is fundamental. In fact, high temperatures can be critical as they may cause insulation melting or magnets demagnetization. Due to the rotation, the direct measurement of rotor's inner temperatures is an expensive and complex option, and we here estimate it from more accessible quantities. In particular we will show by experiments on an actual motor that estimations within few degrees of error can be obtained by feeding readings of external temperatures and of the motor electrical status into suitably trained neural architectures.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.