.png){kind=icon}
.png){kind=icon}
Return
Deep learning-driven preconditioned conjugate gradient method for finite element analysis
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2026.114980.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A sparse U-shaped neural network is developed for finite element preconditioning. • Condition-number-based loss is developed for conjugate gradient preconditioners. • A cache-assisted Lanczos method accelerates loss evaluation for sparse matrices. • The proposed method improves overall efficiency over conventional preconditioners. • Strong generalization is achieved for civil stiffness matrices with varying materials.
Keywords:
sparse neural network
preconditioned conjugate gradient
finite element analysis
condition number
cache-assisted Lanczos method
Journal
IF:
8
Papers:
5.4K
Citations:
3.5W