.png){kind=icon}
.png){kind=icon}
Return
Direct Mesh-Free Topology Optimization Using Random Feature Method
.png){kind=icon}
PRE
AI
DOI：10.1016/j.cad.2025.103939.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• By coupling a density field neural network with a physics-informed response network, RFMTO eliminates the discretization of design variables and the need for traditional finite element analysis, enabling direct structural topology optimization. • For the physical response network, we introduce RFM, which combines the advantages of classical numerical methods and deep learning. RFM offers spectral accuracy for smooth solutions while reducing computational time, significantly enhancing efficiency. • For the loss function of the density network, we incorporate pointwise density targets to guide the optimization process, preventing the optimized structure from converging to poor local minima in complex problems. • The framework has been extended beyond elasticity-related optimization problems to encompass thermodynamics-related problems as well.
Keywords:
density field neural network
physics-informed response network
structural topology optimization
spectral accuracy
thermodynamics-related problems
Journal
C
IF:
3.1
Papers:
3.1K
Citations:
6.4K