Realtime Non-invasive Fault Diagnosis of Three-phase Induction Motor

The objective of this paper is to apply deep learning network running on an embedded system platform diagnose faults a three-phase electric motor by non-contact method based operating noise. To accomplish this, at first, should be designed and trained computer, then converted equivalent run the system. input data two-dimension spectrogram image noise emitted in four main cases, including normal operation, phase shift, loss bearing failure. execution time accuracy these structures will deployed three microcontrollers ESP32, ESP32-C3 nRF52840 determine suitable structure for real-time running. Experimental results show that proposed models could well both computer with highest accuracies are 99,7% 99,3%, respectively. In particular, preliminary remarkable recognition 1,7 seconds 72%, respectively associated realtime performance.