<p>Rotating machinery health monitoring is crucial for the early detection of bearing defects. Every defect in a machine generates a unique vibration signal (vibration signature), which makes vibration analysis a key diagnostic tool for detecting faults and ensuring a long lifetime for rotating machinery. This paper presents a bearing fault diagnosis method based on machine learning (ML) and deep learning (DL) approaches. The proposed method comprises four main steps: signal fragmentation, feature extraction, feature processing, and fault classification. Firstly, the raw time-series sensor data are fragmented into short segments within which the faults are detected. Then, a computationally efficient feature extraction process is implemented by selecting low-dimensional features in the time domain. The combination of a short fragmentation strategy and the low dimensionality of features makes the system more suitable for real-time monitoring and implementation on hardware in manufacturing settings. SMOTE (synthetic minority oversampling technique) is adopted to increase the number of training samples to enhance the performance of classifiers. The proposed method is model-agnostic, as demonstrated by the ability to train four ML classifiers (KNN, SVM, LR, and RF) and a custom convolutional neural network (CNN). The proposed method was evaluated on the Case Western Reserve University dataset. Experimental results demonstrate that the proposed method classifies ten different bearing faults with an accuracy of 99.7% using the proposed CNN model. Furthermore, the study presented a feature importance analysis to identify the contribution of each feature and the most contributing features in the classifier’s decision. In addition, a comparative analysis was conducted, which revealed that the proposed method outperforms other state-of-the-art models for bearing fault diagnosis.</p>