This research introduces a diffusion-based method for denosing bearing signals and proposes a network architecture capable of fully restoring key frequency bands. The approach ensures accurate reconstruction of critical signal features for more precise bearing condition monitoring. This enhances the reliability of diagnostics by preserving essential signal characteristics. 

This research focuses on detecting data drift in bearing signals by generating feature map-based images using simulation signals. The study then applies the YOLO object detection algorithm to estimate the condition of the bearings. This approach enhances the accuracy and reliability of bearing status estimation under varying conditions.

This study estimates battery capacity by utilizing degradation information obtained through incremental capacity analysis and extracting state-of-charge invariant features using a deep learning model. The proposed method improves the accuracy of the capacity estimation algorithm even with signals obtained over a short period. This approach can be effectively applied in real-world battery usage, such as short-term charging, and contributes to the development of a robust battery management system. 

Rotating machinery faults can cause significant industrial disruptions, but traditional sensor-based diagnostics often require invasive methods. This study introduces a vision-based diagnostic algorithm that amplifies specific frequencies using time-frequency filtering, enabling accurate fault detection through measurable displacement. The approach was validated on a rotor kit testbed, effectively identifying faults with precise displacement analysis. 

This research leverages data from time-frequency representation (TFR) and applies whitening as a preprocessing step for unsupervised domain adaptation in speed estimation. The proposed method effectively handles continuous sequence input for improved speed prediction. The approach enhances accuracy in varying conditions without relying on labeled data.