Hindawi Publishing Corporation
Advances in Mechanical Engineering
Volume 2013, Article ID 906120, 16 pages
http://dx.doi.org/10.1155/2013/906120
Research Article
Assessing Numerical Error in Structural Dynamics Using
Energy Balance
Rabindranath Andujar,
1
Jaume Roset,
1
and Vojko Kilar
2
1
Department of Applied Physics, Polytechnic University of Catalonia, Campus Diagonal Nord, Building B5,
C/Jordi Girona 1-3, 08034 Barcelona, Spain
2
Faculty of Architecture, University of Ljubljana, Zoisova cesta 12, 1000 Ljubljana, Slovenia
Correspondence should be addressed to Vojko Kilar; vojko.kilar@fa.uni-lj.si
Received 30 May 2013; Accepted 23 September 2013
Academic Editor: Mehdi Ahmadian
Copyright © 2013 Rabindranath Andujar et al. Tis is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
Tis work applies the variational principles of Lagrange and Hamilton to the assessment of numerical methods of linear structural
analysis. Diferent numerical methods are used to simulate the behaviour of three structural confgurations and benchmarked in
their computation of the Lagrangian action integral over time. According to the principle of energy conservation, the diference
at each time step between the kinetic and the strain energies must equal the work done by the external forces. By computing this
diference, the degree of accuracy of each combination of numerical methods can be assessed. Moreover, it is ofen difcult to
perceive numerical instabilities due to the inherent complexities of the modelled structures. By means of the proposed procedure,
these complexities can be globally controlled and visualized in a straightforward way. Te paper presents the variational principles
to be considered for the collection and computation of the energy-related parameters (kinetic, strain, dissipative, and external
work). It then introduces a systematic framework within which the numerical methods can be compared in a qualitative as well as
in a quantitative manner. Finally, a series of numerical experiments is conducted using three simple 2D models subjected to the
efect of four diferent dynamic loadings.
1. Introduction
1.1. Targets and Interest of Our Research. Variational mechan-
ics date back as far as the Eighteenth Century, when Leibniz,
Euler, Maupertuis, and Lagrange devised the calculus of vari-
ations and the principles of least action. Tis methodology
of treating physical phenomena is based on the notion that
everything in nature tends to a state of minimal energy [1].
In structural engineering practice, there is a preference
to use forces and accelerations rather than energy concepts.
Unfortunately, this approach ofen limits our understanding
of the phenomena, as, for example, in the case of earthquakes,
damage is a function of the square of the velocity, and not so
much of the acceleration [2].
In parallel we will deal with a systematic treatment of the
numerical methods which have proliferated since the 1950s
with the ever-increasing power of computers. Tis ceaseless
growth in numbers and terminology has given place to a
cumbersome mix of mathematics, physics, and computer
science that is ofen difcult to grasp. We have used our work
to propose a possible categorization according to the physical
qualities which they represent instead of according to their
mathematical properties.
1.2. Variational Mechanics. According to the principles of
variational mechanics [3], the diference between kinetic
energy and strain energy in a structural system equals the
applied work due to external forces. In this way, by computing
the energy scalars and carefully accounting for this diference
at each time step, one should be able to infer the degree of
accuracy of a simulation [4].
Te correct values should not in any case diverge much
from zero, and deviations from this value would give us an
idea about how accurate and stable a method is.
1.3. Numerical Methods for Structural Analysis. In previous
works published by the authors [5, 6], it was shown how the
vast amount of existing numerical methods can be grouped