Sinir ağları kullanılarak motor arızalarının belirlenmesi

Since the frequent breakdown of induction motors is a case of industry, the detection of faults is of great importance in order to protect motors and not interrupt vital processes. In this thesis, vibration data from asynchronous motors which are under different loading conditions are classified by using convolutional neural networks. In order to create vibration data, different fault types were deliberately made on asynchronous motors and dataset was created by experimenting with different loading values. The one-dimensional vibration data is transformed into two-dimensional grayscale images and then three-dimensional color images using autocorrelation values, thereby allowing the convolutional neural networks to recognize the vibration data. Different motor faults can be distinguished easily thanks to the convolutional neural network which do not need any feature extraction method.