.png){kind=icon}
.png){kind=icon}
Return
Multiphysics-informed generative modeling for nonlinear transient material removal prediction in grinding
.png){kind=icon}
PRE
AI
J
J
F
B
L
Y
S
S
DOI：10.1007/s00170-026-18655-9.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Grinding is a key finishing process in precision manufacturing, where spatially non-uniform allowance and region-dependent removal objectives must be controlled simultaneously. However, because the material-removal process is not directly observable, existing models struggle to capture the coupled effects of structural parameters, process variables, and the resulting nonlinear transient responses. Conventional numerical simulations and purely data-driven approaches often fail to reconcile predictive accuracy with physical consistency. To address these issues, this paper proposes an integrated material removal modeling framework that couples multiphysics simulation with data-driven approaches. A transient multiphysics model is developed to elucidate the complex interaction between structural parameters and process variables that influence material removal behavior. To tackle the challenge of nonlinear transient prediction, this paper proposes a physics-informed generative adversarial network (PI-GANs), which improves the model's ability to represent time-varying removal behavior by integrating simulation outputs with experimental data through joint optimization. Under uniform allowance conditions, the standard deviation of material removal was reduced by 53.3%, and the maximum absolute deviation decreased by 80.4%. Under non-uniform allowance compensation conditions, the spatial standard deviation of the final thickness was reduced by 59.5%, and the maximum absolute deviation decreased by 56.9%. These results demonstrate that the framework improves material-removal prediction while maintaining physical consistency under complex operating conditions. The framework also offers a transferable route for physics-data hybrid modeling in grinding and related finishing processes.
Keywords:
Intelligent manufacturing
Multiphysics simulation
Physics-informed machine learning
Material removal prediction
Adaptive process optimization
Digital grinding system
Journal
T
IF:
0
Papers:
2.0K
Citations:
0
Organization
Cited Papers