Wear state assessment of external gear pump based on system-level hybrid digital twin

Abstract

Modeling technology is both the core and the difficulty of digital twin. In response to this challenge, a digital twin framework based on dynamic model is proposed and applied to fault prognostic and health management (PHM) of fuel control system. Taking the wear failure of external gear pump as an example, the system-level hybrid digital twin model is introduced. Firstly, based on the first principle, the dynamic model of the fuel control system is established to realize the mirror image of the physical entity, and the fault sensitive parameters that need to be updated in the physical model are defined according to the prior knowledge. Then, combined with deep reinforcement learning, an autonomous learning strategy of virtual entity is constructed, achieving updating and iteration of dynamic model. Finally, an interpretable gear pump wear assessment model is built, and the wear state is evaluated by the distance of the model output vector. From the results, the proposed method can ensure the real-time consistency of virtual entity and physical entity under different working conditions, and realize the wear assessment of external gear pump.