CLASSIFICATION OF GROOVE SHAPE IN ROTARY ELECTRIC MACHINES WITH CONVOLUTIONAL NEURAL NETWORK


Creative Commons License

Noğay H. S.

9 th INTERNATIONAL CONGRESS ON ENGINEERING, ARCHITECTURE AND DESIGN, İstanbul, Turkey, 14 May 2022, pp.1050-1059

  • Publication Type: Conference Paper / Full Text
  • City: İstanbul
  • Country: Turkey
  • Page Numbers: pp.1050-1059
  • Bursa Uludag University Affiliated: Yes

Abstract

Abstract: Deciding on the groove shape is an important step in the design process of rotating electrical machines. Since the rotary electric machines are designed with the help of package programs, the groove shape is automatically selected. However, groove leakages fields must be taken into account when deciding on the shape of the groove. A deep learning model that can both take into account the groove leakages and help the groove shape decision be made in the fastest way and also classify the groove shapes can facilitate and speed up the work of the designers. In this study, the convolutional neural networks (CNN) model, which is very popular among deep learning (DL) methods, was designed and implemented. In order to ensure the success of the model, increase its reliability, and ensure its generalizability, the pre-trained CNN model was rearranged and applied in accordance with the purpose of this study with the help of a transfer learning technique (TL). As a result, groove shape classification and detection were performed with 100% accuracy with the CNN model, and it was proven that the CNN model could positively affect the design process of rotary electric machines. Keywords: Rotary Electrical Machines, CNN, TL, DL