SPECIES DETECTION OF ROTARY ELECTRIC MACHINES WITH DEEP LEARNING FROM APPEARANCES


Creative Commons License

Noğay H. S.

7 th INTERNATIONAL CONGRESS ON ENGINEERING AND TECHNOLOGY MANAGEMENT, 16 April 2022, pp.488-495

  • Publication Type: Conference Paper / Full Text
  • Page Numbers: pp.488-495
  • Bursa Uludag University Affiliated: No

Abstract

Abstract: Rotary electric machines are widely used in industry, energy production, renewable energy sources and factories, mainly alternating current (AC) machines and direct current (DC) machines. Therefore, any scientific study that can be done on rotating electrical machines will contribute to industrialization. Although the rotating electric machines vary according to the manufacturers, the ones with the same power are very similar to each other in terms of external appearance. In systems that require remote control or monitoring with a screen, it is sometimes necessary to determine the type of rotating electrical machine immediately. Label values may not be read on the screens or the label may remain outside the camera dial while it is in operation. In this study, a deep learning (DL) method, a convolutional neural network (CNN) model, was implemented in order to classify and detect rotating electrical machines with only external appearance for such special cases. In the test results after the training of the model, success was achieved with an accuracy rate of 87.5%. It has been observed that the use of the proposed DL model in the classification and species detection of rotating electrical machines gives very successful results. Keywords: Rotary Electrical Machines, CNN, TL, DL, TL, Induction, DC, Synchronous