Fault Mechanism Assisted Intelligent Fault Diagnosis Method Based on Improved Generative Adversarial Network for Two-Stage Gearbox

Artificial intelligence and digital twin technology show the excellent ability in the fault prognostic and health management of machinery equipment. However, interaction between the data driven and physical informed is a challenge for modeling the digital twin framework. In response to this limitation, a fault mechanism assisted intelligent fault diagnosis method based on improved generative adversarial network for digital twin two-stage gearbox is proposed. Firstly, the twenty-six degree-of-freedom dynamic model is established to realize the virtual entity of two-stage gearbox. The initial parameters of dynamic model are determined by the prior knowledge based on the fault mechanism of gear. Then, the improved generative adversarial network is designed to achieve the interaction of virtual entity and physical entity. The measured signal from the physical entity is used to train the proposed model for minimizing the difference of simulation signal from the dynamic model. The loss of proposed model is introduced by combining the physics knowledge of gear dynamic mechanism and observations from the sensor. Finally, the proposed method is applied to simulate the vibration signal of gear fault under different conditions that is consistent with the measured signal, which can be applied for the fault diagnosis of gear. The two-stage gearbox fault experiments with missing sample are used to verify the effectiveness of the proposed methods.