.png){kind=icon}
.png){kind=icon}
Return
Multi-task machine learning for structural optimizations of lattice shells
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engstruct.2026.122654.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A multi-task reinforcement learning (MTRL) framework optimizes lattice shells by minimizing the product of structural volume and strain energy. • MTRL can be trained on one task and applied to multiple tasks. • MTRL shows generalizability across three optimization problems: shape, sizing, and combined size-shape optimization. • MTRL outperforms the Genetic Algorithm (GA) and Simulated Annealing (SA) with Wilcoxon test p-values of 0.0039 and 0.0117.
Keywords:
Multi-task reinforcement learning
Structural optimization
Lattice shells
Shape optimization
Sizing optimization