Journal
IIE TRANSACTIONS
Volume 47, Issue 2, Pages 141-152Publisher
TAYLOR & FRANCIS INC
DOI: 10.1080/0740817X.2014.917777
Keywords
Gaussian processes; latent variables; carbon nanotubes; Buckypaper; multi-fidelity analysis
Funding
- NSF [CMMI-1000088, DMS-1208952]
- AFOSR DDDAS [FA9550-13-1-0075]
- King Abdullah University of Science and Technology [KUS-CI-016-04]
- Division Of Mathematical Sciences
- Direct For Mathematical & Physical Scien [1208952] Funding Source: National Science Foundation
Ask authors/readers for more resources
Buckypapers are thin sheets produced from Carbon NanoTubes (CNTs) that effectively transfer the exceptional mechanical properties of CNTs to bulk materials. To accomplish a sensible tradeoff between effectiveness and efficiency in predicting the mechanical properties of CNT buckypapers, a multi-fidelity analysis appears necessary, combining costly but high-fidelity physical experiment outputs with affordable but low-fidelity Finite Element Analysis (FEA)-based simulation responses. Unlike the existing multi-fidelity analysis reported in the literature, not all of the input variables in the FEA simulation code are observable in the physical experiments; the unobservable ones are the latent variables in our multi-fidelity analysis. This article presents a formulation for multi-fidelity analysis problems involving latent variables and further develops a solution procedure based on nonlinear optimization. In a broad sense, this latent variable-involved multi-fidelity analysis falls under the category of non-isometric matching problems. The performance of the proposed method is compared with both a single-fidelity analysis and the existing multi-fidelity analysis without considering latent variables, and the superiority of the new method is demonstrated, especially when we perform extrapolation.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available