.png){kind=icon}
.png){kind=icon}
Return
A novel reliability-based design optimization method through instance-based transfer learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.cma.2024.117388.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The RBDO optimization process consists of two main steps: iterative updating of design points and repeated reliability analysis at these different design points. A large number of performance function calls are usually necessary for each reliability analysis, and it involves with the repeated reliability analysis at different design pints, leading to potentially prohibitive computational cost when dealing with challenging performance function evaluations. In order to overcome the issues caused by repeated reliability analysis, this paper proposes an RBDO method through instance- based transfer learning. The proposed method converts repeated reliability analysis at different design points into a serial of different target domain tasks through transfer learning, which provides a new perspective for solving RBDO problems. Firstly, it carries out the repeated reliability analysis with high accuracy by transferring the reliability knowledge from the source domain to the target domains. Subsequently, the gradient information is obtained using the first- order score function method and design point is updated. The method utilizes instance-based transfer learning to achieve repeated reliability analysis, enhancing the performance of RBDO methods, particularly in computational efficiency. The effectiveness of the proposed method is demonstrated through three numerical examples and one engineering application.
Keywords:
Reliability-based design optimization
Reliability analysis
Transfer learning
Journal
IF:
7.3
Papers:
1.3W
Citations:
5.6W