.png){kind=icon}
.png){kind=icon}
Return
Welding robot digital twin and precision degradation prediction based on physics-informed neural networks
.png){kind=icon}
PRE
AI
S
J
C
W
C
DOI：10.1016/j.jmsy.2026.04.024.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• The article proposed a concise robot accuracy degradation evaluation model to describe the joint synchronous belt degradation. • Propose a small sample transfer learning technique based on physical modal loss, which guides the training process of neural networks through physics knowledge. • The article proposes a multi-source feature enhancement structure based on cross attention mechanism, which allows the model to simultaneously focus on the correlation between laboratory data and field data. • The proposed method was thoroughly validated in a real manufacturing system, and the performance advantages of the designed module were verified through ablation experiments and baseline comparisons.
Keywords:
robot accuracy degradation
physics-informed neural networks
transfer learning
feature enhancement
digital twin
Journal
IF:
14.2
Papers:
2.6K
Citations:
1.6W
Organization
Cited Papers