Bearings play a crucial role as key components in electric motors widely used in industrial settings. When defects occur in these bearings, they can lead to serious motor failures, causing significant downtime for the entire system. Therefore, prompt detection of bearing defects is essential to minimize system downtime and improve production efficiency. Previous research has mainly focused on using convolutional neural network (CNN)-based and recurrent neural network (RNN)-based models for bearing fault diagnosis. However, these models often involve high computational complexity for learning and rely on high-dimensional data, such as time series data collected from sensors or image data generated from time series data. Employing these models can result in significant delays in the fault diagnosis process, making them less suitable for real-world industrial environments. In this study, we propose an innovative approach to reduce both training and inference time by utilizing time domain features (TDF) obtained through the transformation of time series current data. The TDF provides a low-dimensional data representation that can be used to facilitate model training and inference. Additionally, the proposed model uses linear layers based on residual structure and dense connectivity structure to improve performance and reduce model complexity. To evaluate the effectiveness of our proposed method, we conducted simulations using the Paderborn University bearing dataset for training and inference. In experimental results, we show that the proposed method outperforms conventional methods in model complexity and fault diagnosis accuracy.

Fault diagnosis in Computer Numerical Control (CNC) machines remains challenging due to severe class imbalance, scarcity of fault data, and distribution shifts across machines. This paper introduces Undersampling-based Few-shot Learning (UnderFSL), a simple yet effective framework that integrates strategic undersampling using Condensed Nearest Neighbor (U-CNN) with a Relation Network few-shot classifier. The proposed method first transforms raw 1D vibration signals into 2D Continuous Wavelet Transform (CWT) scalograms to capture time–frequency structure and then reduces the majority (normal) class using U-CNN, yielding a compact set of boundary-informative prototypes while alleviating imbalance. Finally, a Relation Network is trained in an episodic FSL regime on the balanced set to support cross-machine generalization. On the Bosch CNC machining benchmark under leave-one-machine-out validation, UnderFSL attains a macro F1-Score of 0.96, an accuracy of 0.96, a recall of 0.92, and a precision of 1.00, surpassing traditional and standard deep baselines. The results suggest that boundary-preserving undersampling combined with metric learning provides a robust and scalable path for industrial fault diagnosis when fault data are extremely limited.