Two-Level Multifidelity Algorithm With Dimension Reduction for Efficient Uncertainty Quantification of MWCNT Interconnects

Journal article

Surila Guglani, Sourajeet Roy
IEEE transactions on electromagnetic compatibility (Print), 2021

Cite

APA Click to copy
Guglani, S., & Roy, S. (2021). Two-Level Multifidelity Algorithm With Dimension Reduction for Efficient Uncertainty Quantification of MWCNT Interconnects. IEEE Transactions on Electromagnetic Compatibility (Print).

Chicago/Turabian Click to copy
Guglani, Surila, and Sourajeet Roy. “Two-Level Multifidelity Algorithm With Dimension Reduction for Efficient Uncertainty Quantification of MWCNT Interconnects.” IEEE transactions on electromagnetic compatibility (Print) (2021).

MLA Click to copy
Guglani, Surila, and Sourajeet Roy. “Two-Level Multifidelity Algorithm With Dimension Reduction for Efficient Uncertainty Quantification of MWCNT Interconnects.” IEEE Transactions on Electromagnetic Compatibility (Print), 2021.

BibTeX Click to copy

@article{surila2021a,
  title = {Two-Level Multifidelity Algorithm With Dimension Reduction for Efficient Uncertainty Quantification of MWCNT Interconnects},
  year = {2021},
  journal = {IEEE transactions on electromagnetic compatibility (Print)},
  author = {Guglani, Surila and Roy, Sourajeet}
}

Abstract

In this article, a polynomial chaos (PC) approach to quantify the uncertainty in transient simulation results of multiwalled carbon nanotube interconnect networks is presented. The proposed algorithm offers two distinct levels of numerical efficiency. The first level of efficiency comes from the use of a multifidelity formulation where the numerical expediency of a crude low-fidelity model of the network is combined with the accuracy of a high-fidelity model. At the second level, the crude low-fidelity model is further leveraged to perform dimension reduction. Consequently, the proposed algorithm offers substantially more speedup in the training of PC metamodels than existing techniques. The advantages of the proposed approach are validated using multiple numerical examples.