Published November 2019 | Version v1
Journal article

Quaternion and octonion-based finite element analysis methods for computing multiple first order derivatives

  • 1. Universidad EAFIT, Medellín (Colombia)
  • 2. University of Texas at San Antonio (United States)
  • 3. Angelo State University, San Angelo, TX (United States)

Description

Highlights: • Quaternion- and octonion-based finite element methods were developed. • Multiple accurate first order derivatives computed in a single run. • Extends the complex Taylor series expansion method to any order Cayley-Dickson algebra. • The method was implemented through the addition of imaginary nodes/degrees of freedom. • The method was implemented within the commercial finite element code Abaqus. -- Abstract: The complex Taylor series expansion method for computing accurate first order derivatives is extended in this work to quaternion, octonion and any order Cayley-Dickson algebra. The advantage of this new approach is that highly accurate multiple first order derivatives can be obtained in a single analysis. Quaternion and octonion-based finite element analysis methods were developed in order to compute up to three (quaternion) and up to seven (octonion) first order derivatives of shape, material properties, and/or loading conditions in a single analysis. The traditional finite element formulation was modified such that each degree-of-freedom was augmented with three or seven additional imaginary nodes. The quaternion and octonion-based methods were integrated within the Abaqus commercial finite element code through a user element subroutine. Numerical examples are presented for thermal conductivity and linear elasticity; however, the methodology is general. The results indicate that the quaternion and octonion-based methods provide derivatives of the same high accuracy as the complex finite element method but are significantly more efficient. A Fortran code to solve a simple seven variable quaternion example is given in the Appendix.

Availability note (English)

Available from http://dx.doi.org/10.1016/j.jcp.2019.07.030

Additional details

Identifiers

DOI
10.1016/j.jcp.2019.07.030;
PII
S0021999119305157;

Publishing Information

Journal Title
Journal of Computational Physics (Print)
Journal Volume
397
Journal Page Range
vp.
ISSN
0021-9991
CODEN
JCTPAH

Optional Information

Copyright
Copyright (c) 2019 Elsevier Inc. All rights reserved.