Objective  This study aims to investigate the dynamic modeling process of the relationship between bearing load and displacement using transfer learning techniques.

Method A multi-supported shafting system was selected as the research object. The key components of the modeling process were systematically analyzed, and an orthogonal experimental scheme was designed focusing on three critical factors: the accuracy of transferred knowledge, the adaptability of transfer strategies, and the precision of target data samples. The modeling research based on transfer learning was then conducted accordingly.

Results The results indicate that the importance ranking of the factors is as follows: accuracy of transferred knowledge > adaptability of transfer strategy > precision of target data samples. During the modeling process, priority should be given to the accuracy of transferred knowledge and the adaptability of transfer strategies to quickly align with the target domain. Furthermore, maintaining the target data sample error within 4% ensures the relative stability of the modeling process.

Conclusion This study provides theoretical support for constructing the relationship model between bearing load and displacement in engineering applications. It offers significant value in reducing testing costs and enhancing modeling efficiency.