استخراج ویژگی بر اساس ثابت شبه انرژی لاپلاسین و گراف‌های مسیری به‌منظور تشخیص عیوب بلبرینگ‌

Conventional bearing fault diagnosis methods mainly rely on feature extraction from vibration signals in the time, frequency, or time–frequency domains. However, the majority of these methods face limitations when analyzing signals contaminated with severe noise as well as signals with a limited number of samples. In this study, a new method based on graph spectral theory is presented for bearing fault diagnosis. In this approach, the vibration signal is modeled as a path graph, and the resulting adjacency and Laplacian matrices provide distinctive information about the types of faults. In particular, the Laplacian-energy-like invariant is selected as the main feature to describe the complexity of the graph and detect different faults. The proposed method is evaluated using both simulated signals and experimental data. The results demonstrate that this approach outperforms other single-feature methods in terms of fault diagnosis accuracy and, in addition, exhibits robust performance under noisy conditions and with a limited number of samples.