Comparative Study of CNN and RNN for Motor fault Diagnosis Using Deep Learning
Dong-Jin Choi, Ji-Hoon Han, Sang-Uk Park, Sun-Ki Hong
Abstract
Electric motors are in a very important position in the industry. Therefore, Motor fault diagnosis is also very important. In this paper, motor fault diagnosis is carried out using CNN and RNN, the best-known method of supervised learning. Fault diagnosis algorithm is obtained using two algorithms and learning and verification are performed using motor vibration data. As a result, the CNN and RNN algorithms showed high accuracy, although the CNN algorithm showed higher accuracy than RNN. However, RNN is an algorithm with a Recurrent Neural Network structure. Therefore, compared with CNN, it has the advantage of being robust against changes in the surrounding environment of the motor. Therefore, both CNN and RNN show usable accuracy, and it is most important to select the appropriate algorithm for the situation.