Wavelet packet decomposition combined with gravitational search for robust fault diagnosis in harmonic drive

With the increasing deployment of industrial automation, Harmonic Drives have become critical transmission components in robotic manipulators requiring high-precision motion. Nevertheless, long-term operation frequently induces bearing degradation and gear wear, leading to performance deterioration and reduced service life. To address this challenge, this study proposes a hybrid fault diagnosis framework integrating Wavelet Packet Decomposition and the Gravitational Search Algorithm (WPD-GSA). Multi-domain features are extracted from vibration and stator current signals, and GSA is employed to select the most fault-sensitive components. Experimental validation demonstrates that the proposed WPD-GSA method achieves accurate fault classification, early fault detection, and strong robustness under noisy conditions, while maintaining high computational efficiency.