<![CDATA[Bearings are a crucial mechanical component in machinery systems, they function to support and reduce friction between moving parts. Damage to bearings can lead to decreased machine efficiency and further damage to other components. This study aims to develop a bearing fault detection system using a Deep Learning approach. The dataset used was obtained from acoustic signal recordings of induction motor bearings under three different conditions: good condition, slightly damaged condition, and severely damaged condition. The signals are processed using the Short Time Fourier Transformation method to represent them as two-dimensional time-frequency-based spectrograms. The bearing conditions are classified by a Convolutional Neural Network model based on these spectrograms. The architecture used in this study consists of several convolutional and pooling layers for feature extraction and classification. The model training uses a dataset that has been split between training data and validation data. The training results show that the model can achieve a validation accuracy of up to 99%, with stable performance and no indication of overfitting or underfitting. The accuracy value reaches 0.99, with a precision of 1.00, recall of 1.00, and an F1-score of 1.00. The macro and weighted averages, each valued at 0.99, indicate that the model performs excellently across all classes. This study proves that the STFT and CNN methods are effective in detecting and classifying bearing faults using acoustic signals. This system has the potential to be implemented in industry as an efficient tool for preventive maintenance compared to conventional methods that rely on human hearing.]]>
