Uncertainty of FEM Solutions Using a Nonlinear Least Squares Fit Method and a Design of Experiments Approach

Published

October 7, 2015

Author(s)

Jeffrey T. Fong, Nathanael A. Heckert, James J. Filliben, Pedro V. Marcal, Robert Rainsberger

Abstract

Uncertainty in COMSOL finite element simulations due to (a) mesh-induced truncation errors, and (b) model parameter uncertainties, is estimated using a nonlinear least squares logistic distribution fit method, and a design-of-experiments approach, respectively. Examples to illustrate both approaches are given using the COMSOL Structural Mechanics module (stress analysis of a wrench), the COMSOL RF module (application of an MRI RF coil design), and other general-purpose FEM software packages. Significance and limitations of both methods are presented and discussed.

Proceedings Title

Proceedings of 2015 COMSOL Users' Conference, Boston, MA, Oct. 7-9, 2015

Conference Dates

October 7-9, 2015

Conference Location

Boston, MA

Pub Type

Conferences

Download Paper

Local Download

Keywords

Design of Experiments, Finite Element Method, Logistic function, Nonlinear Least Squares Fit Method, Uncertainty Quantification.

Citation

Fong, J. , Heckert, N. , Filliben, J. , Marcal, P. and Rainsberger, R. (2015), Uncertainty of FEM Solutions Using a Nonlinear Least Squares Fit Method and a Design of Experiments Approach, Proceedings of 2015 COMSOL Users' Conference, Boston, MA, Oct. 7-9, 2015, Boston, MA, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=919592 (Accessed October 10, 2025)