Digital Twin Modeling Method for Hierarchical Stiffened Plate Based on Transfer Learning

As the key load-bearing component of spacecraft, strength evaluation stiffened plate structures faces two challenges. On one hand, simulation results are sometimes inaccurate, due to simplification true loading conditions and modeling details. other data from sensors cannot provide full-field information structure, which may result in misjudgment structural state. To this end, a digital twin method multi-source fusion based on transfer learning is proposed paper. In learning, sensor utilized as source dataset target dataset, respectively. First, pre-trained deep neural network (DNN) model established dataset. Then, DNN fine-tuned using lower rate fewer training epochs. Finally, can be built, capable visualizing structure. verify effectiveness method, an experimental study hierarchical carried out. Compared with traditional high prediction accuracy efficiency demonstrating its potential for health monitoring spacecraft orbit.