基于峭度熵与分层极限学习机的动量轮轴承故障诊断研究

The momentum wheel is the key component of the satellite attitude control system, and its reliability is directly related to the life and safety of the whole satellite. As the core component of momentum wheel, bearing is prone to failure. Due to its unique structure and complex operating environment, the signal to noise ratio of monitoring signals is low, and early fault diagnosis is difficult. Aiming at this situation, a feature extraction method combining variational mode decomposition and kurtosis entropy was proposed to obtain the weak fault features of momentum wheel bearing monitoring signals and to establish the feature vectors. The layered extreme learning machine was introduced, and the structure and coding method were optimized for bearing fault identification. Finally, the proposed method was applied to the actual fault diagnosis. The comparison with the traditional ELM method shows that the proposed method has higher diagnostic accuracy (98.5%) in the fault diagnosis of momentum wheel bearings.