Fault Detection and Isolation Based on Improved Latent Dirichlet Allocation for Planetary Gearbox
Jingyuan Wang, Xianghua Wang, Xiangrong Wang, Jiegang Wang · 2022 IEEE International Conference on Unmanned Systems (ICUS) · 2022
A new fault diagnosis algorithm for planetary gearbox based on improved Latent Dirichlet Allocation (LDA) is proposed. The conventional LDA requires the number of latent topics as prior knowledge, which is usually not exactly known. To circumvent this issue, improved LDA is proposed by combining the Chinese Restaurant Process mixture model with LDA, avoiding the shortcomings of the conventional LDA. The improved LDA is then applied to planetary gearbox for fault detection and isolation. Finally, simulations are conducted to verify the effectiveness of the proposed method.