Rotary electrical machines can be classified into two basic types as Direct Current (DC) Machines and Alternating Current (AC) Machines. AC machines are also classified as induction machines and synchronous machines. When the basic classification of AC and DC is considered as rotating electrical machines, they can be classified in three categories as induction machines, synchronous machines and DC machines. By looking at the external appearance of a rotary electric machine, it may not be immediately clear which type it is in the first place. Label values are often not readable from photos. In remote control planning in factories, motor production facilities, it is sometimes necessary to understand what type it is from the picture or screen image of the rotating electrical machine. On the other hand, in some cases, the label values may not be legible even up close. In order to be a solution to such problems, in this study, triple classification of rotary electric machines has been carried out with the help of the convolutional neural network (CNN) model, which is one of the most preferred deep learning methods. For this, the pre-trained CNN model, which was trained and tested before, was redesigned and trained by changing the last three layers in accordance with the purpose of this study, with the help of the transfer learning (TL) technique. As a result of the training, it was seen that a very successful classification and detection was performed with a satisfactory accuracy rate, especially when considering multiple classifications.
Anahtar Kelimeler: Electrical Machines, Induction, DC, Synchronous, CNN, TL
|
