Two-dimensional pattern entropy for bearing fault diagnosis

Entropy-based methods have been widely applied in bearing fault diagnosis due to their excellent ability to detect the dynamic characteristics of nonlinear signals. However, existing methods rely on one-dimensional patterns lacking geometric characteristics, making it challenging to accurately capture the dynamic characteristics of signals, and their performance is sensitive to multiple parameters. To overcome the aforementioned limitations, a new entropy metric called two-dimensional pattern entropy (TDPE) was proposed, which requires only grid density as its single parameter. TDPE constructs a two-dimensional time-frequency difference map and partitions its boundaries to obtain two-dimensional patterns that contain geometric characteristics of time-frequency features, thereby enabling the dynamic characteristics of signals to be characterized more accurately and comprehensively. The results of the simulation experiments demonstrate that TDPE exhibits excellent performance in terms of length sensitivity, consistency in complexity quantification, and robustness to noise. In real-world bearing fault diagnosis experiments, TDPE achieves higher diagnostic reliability and stability compared to other entropy-based metrics, fully validating its superiority in extracting fault features from vibration signals.