EXODUS/packages/seacas/doc-source/numbers/numbert.tex

\documentstyle[12pt,sequence]{sreport}
\input{my$tex:macro.tex}
\def\NUM{{\sf Numbers}}
\def\EXO{{\sf Exodus}}
\makeatletter
\def\listoffigures{\@restonecolfalse\if@twocolumn\@restonecoltrue\onecolumn
  \fi\filbreak\bigskip\leftline{\large\bf Figures}\markboth
   {FIGURES}{FIGURES}\addcontentsline{toc}{chapter}{Figures}
  \@starttoc{lof}
\if@restonecol\twocolumn\fi}
%
\def\listoftables{\@restonecolfalse\if@twocolumn\@restonecoltrue\onecolumn
  \fi\filbreak\bigskip\leftline{\large\bf Tables}\markboth
   {TABLES}{TABLES}\addcontentsline{toc}{chapter}{Tables}\@starttoc{lot}
\if@restonecol\twocolumn\fi}
\def\l@figure{\@dottedtocline{0}{0.0em}{1.5em}}
%
\def\chapter{
   \global\@topnum\z@            % Prevents figures from going at top of page.
   \secdef\@chapter\@schapter
}
\begin{document}
\sand88-0737
\UC-32
\title{NUMBERS: A Mass Properties Calculation Program for Two- and
Three-Dimensional\\Finite Element Models}
\author{Gregory D. Sjaardema}
\division{Applied Mechanics Division I}
\abstract{abstract}
\makesandtitle
\tableofcontents
\listoffigures
\clearpage
%
% Chapter 1: Introduction
%
\chapter{Introduction}

During the structural analysis of a body, it is often necessary or
helpful to know several properties or statistics of the finite element
description.  The \NUM\ code was written to provide an efficient
method for calculating these properties. Properties calculated
include: (1)~the volume and mass of each element block, (2)~the mass
(or area) moments of inertia, (3)~the location of the center of
gravity, (4)~the minimum, maximum, and average element volume (area
for two-dimensional bodies) for each element block, and (5)~the number
of elements in each element block.

%to verify
%accurate representation of the actual body and to determine the
%density of pseudo materials used to approximate the mass of
%non-structural components.
%of the body which can be used to determine the eccentric impact
%response (slapdown) of a deformable body~\cite{slapdown}.
%which is sometimes required to
%determine the orientation of a body subjected to impact events.
%which can be used to
%estimate the explicit integration time step and therefore, the
%computer time required for the analysis,

The initial impetus for development of this code was the structural
analysis of a nuclear waste shipping container subjected to several
hypothetical 30-foot drops onto a rigid surface at various
orientations.  However, the calculated properties are useful for many
different types of analyses.

The remainder of this report is organized as follows.  The numerical
formulation is described in the next chapter.  The following chapter
describes how the properties calculated by \NUM\ were used in the
analyses of a nuclear waste transportation container.  A procedure for
estimating the stable time step size is described in Chapter~4.
Chapter~5 describes the execution of the code, followed by the
Conclusions and Summary in Chapter~6.


%
% Numerical Formulation
%

\chapter{Numerical Formulation} A brief summary of the equations used
to calculate the mass properties is given in this chapter. The method
used to integrate these equations is not detailed since the numerical
algorithms can be found in most textbooks on finite element
methodology.  The moments of inertia are originally calculated about
the origin of the coordinate system and then transferred to the
centroid using the parallel-axis theorem.

The volume and the moments of inertia about the origin of the
coordinate system are given by the following integrals:
\begin{eqnarray}
 V      &=& \int_V dV                             \\
 I_x    &=& \int_V \rho\left(y^2 + z^2\right)\,dV \\
 I_y    &=& \int_V \rho\left(x^2 + z^2\right)\,dV \\
 I_z    &=& \int_V \rho\left(x^2 + y^2\right)\,dV \\
 I_{xy} &=& \int_V \rho xy\,dV                    \\
 I_{xz} &=& \int_V \rho xz\,dV                    \\
 I_{yz} &=& \int_V \rho yz\,dV
\end{eqnarray}

where $V$ is the volume, $\rho$ is the density, $I$ is the mass moment
of inertia, and the subscripts $x$, $y$, and $z$ denote the axes about
which the moment is measured.  The double subscripts indicate products
of inertia.  Note that the product of inertia with respect to any two
orthogonal axes is zero if either of the axes is an axis of symmetry.

\paragraph{Two-dimensional Axisymmetric Body}
For a two-dimensional axisymmetric body, the integrals are written in
cylindrical coordinates using the radius $r$ and the angle $\theta$,
where $x = r\cos\theta$ and $z = r\sin\theta$.  The $y$-axis is
assumed to be the axis of revolution.  The infinitesimal volume $dV$
is equal to $r\,dA\,d\theta$ where $dA$ is an infinitesimal area
equal to $dr\,dy$. Equations (1) through~(4) are rewritten as:
\begin{eqnarray}
V   &=& \int_{-\pi}^\pi\int_A r\,dA\,d\theta   \\
I_x = I_z &=&\int_{-\pi}^\pi\int_A
              \rho\left(y^2 + r^2\sin^2\theta\right)r\,dA\,d\theta\\
I_y &=& \int_{-\pi}^\pi\int_A \rho r^3\,dA\,d\theta
\end{eqnarray}
Performing the outer integration from $-\pi$ to $\pi$ gives:
\begin{eqnarray}
V   &=& 2\pi\int_A r\,dA   \\
I_y &=& \int_A 2\pi\rho r^3\,dA   \\
I_x = I_z &=& \int_A 2\pi\rho r y^2\,dA + \int_A \pi\rho r^3\,dA\\
    &=& \int_A 2\pi\rho r y^2\,dA + {I_y\over2}
\end{eqnarray}
The products of inertia $I_{xy}$, $I_{xz}$, and $I_{yz}$ are zero since
all three axes are symmetry axes.

\paragraph{Two-dimensional Planar Body}
For a two-dimensional planar body, area moments of inertia are
calculated; the depth in the $z$-direction is ignored. The integrals
simplify as follows:
\begin{eqnarray}
 I_x    &=& \int_A \rho y^2\,dA                   \\
 I_y    &=& \int_A \rho x^2\,dA                   \\
 I_z    &=& \int_A \rho\left(x^2 + y^2\right)\,dA = I_x + I_y \\
 I_{xy} &=& \int_A \rho xy\,dA
\end{eqnarray}
The products of inertia $I_{xz}$ and $I_{yz}$ are both zero.

\paragraph*{Evaluation of Integrals} The integrals are evaluated using
the isoparametric element formulation and Gaussian Quadrature.
Details of this process can be found in most books on finite element
methodology.  In \NUM, two options are available for integration of
the equations.  The two-dimensional equations can be integrated with
either 1- or 4-point quadrature, and the three-dimensional equations
with either 1- or 8-point quadrature.  The mass moments of inertia for
bodies with non-rectangular elements are calculated more accurately
with the higher-order integration; volumes and areas are integrated
exactly with either integration option.  The second-order quadrature
rule is also useful for calculating the section properties of a
non-standard shape.  For this purpose, the discretization of the body
is only refined enough to capture the essential details of the shape
since a structural analysis will not be performed.

\paragraph{Calculation of Centroid Location} The location of the
centroid is calculated using the first mass moments about each of the
three coordinate axes.  The mass moments are summed in each of the
three coordinate directions over the range of the elements.  The
centroid location is then given by the quotient of the mass moment
sums divided by the total mass of the body.  For a two-dimensional
axisymmetric body, $X_c$ and $Z_c$ (the $X$ and $Z$ coordinates of the
centroid) are zero; for a two-dimensional planar body,
$Z_c$ is zero.

\chapter{Example Usage of \NUM\ Properties}

Example uses of the properties calculated by \NUM\ in the analyses of
a nuclear waste shipping container are presented in this chapter.
During the development of the finite element models required for these
analyses, several properties were required to verify correctness of
the model, determine impact orientations, calculate pseudo densities,
and other uses. There was no method available for efficiently
calculating these properties; therefore, the \NUM\ code was written
specifically to calculate the needed properties.  The descriptions
below are not a complete list of the uses of the \NUM\ code; different
analyses would use the properties differently.

\paragraph*{Mass and Volume Calculation} Several finite element
models, both two-dimensional and three-dimensional, were required in
the shipping container structural analyses.  The mass and volume of
each of the materials in each of the models were calculated to ensure
that the body was being represented correctly.

\paragraph*{Calculation of Pseudo Density} The cargo in the container
was approximated by a pseudo material since the actual configuration
of the cargo was too complicated to model.  The structural properties
of the material were unimportant; however, the cargo was required to
have the correct mass.  \NUM\ was used to calculate the total volume
of the pseudo cargo material; the cargo mass was then divided by the
calculated volume to determine the density of the pseudo material.

\paragraph*{Centroid Location and Mass Moments of Inertia} The
location of the centroid of the container was  used to
determine the orientation of the container with respect to the rigid
surface for the stable top and bottom corner drops.  In a stable
corner drop, the body is oriented such that a vector from the centroid
to the impact point is normal to the rigid surface, and the
initial velocity of the body is directed along this vector.

The mass moments of inertia and the mass were used to calculate the
longitudinal radius of gyration.  The longitudinal radius of gyration
has been shown to be an important factor in the behavior of a body
during an eccentric impact~\cite{slapdown}. In an eccentric impact, a
normal to the impact surface at the impact point does not pass through
the center of mass of the body.  This impact orientation, in which
there is an initial impact at one end of the body followed by a
secondary impact at the opposite end of the body, is commonly called
{\em slapdown}.  For certain geometries, the secondary impact velocity
is greater than the initial impact velocity.

\paragraph*{Maximum and Minimum Elemental Quantities} The maximum,
minimum and average element volumes for each material were used for
information and validation of the finite element discretization.  The
minimum time factor was used to determine the approximate computer
time required for an analysis using the procedure described in
Chapter~\ref{s:time}.

All of these quantities were calculated using \NUM. It gives a concise
summary of the important model statistics that could not be easily
determined by other methods.  The above list is only a sample of the
possible uses of the properties calculated by \NUM; other uses include
calculating section properties for non-standard cross-sections and
providing properties for rigid-body dynamics problems.

\chapter{Estimation of Time Step Size and Required Computer
Time}\label{s:time}

\NUM\ calculates a quantity called the `minimum time factor' for each
material block.  This value and the material's physical properties can
be used to estimate the maximum stable time step for a transient
dynamics finite element analysis using explicit integration. Explicit
integration is used in most large-deformation, nonlinear, transient
dynamics finite element codes~\cite{pronto,dyna}. The analysis time
divided by the time step size is equal to the number of integration
steps required in the analysis. Multiplying this quantity by the
number of elements and the CPU time required per element integration
step gives a good estimate of the total computer time required for an
analysis.  The procedure described in this chapter has not yet been
fully investigated.  Modification of the calculated time step may be
necessary to improve the accuracy of the predictions; however, the
procedure provides an accurate initial estimate.

The stable time step size for the central difference operator commonly
used in transient dynamic analysis codes is given by~\cite{cook}
\begin{equation}
\Delta t \leq {2\over \omega_{\max}} \left(\sqrt{1+\epsilon^2}-\epsilon\right)
\end{equation}
where $\omega_{\max}$ is the maximum frequency of the mesh, and
$\epsilon$
is the fraction of critical damping in the highest element mode.
Flanagan and Belytschko~\cite{flanagan} have derived simple formulas
for bounding the maximum eigenvalues for the uniform strain eight-node
hexahedron and the four-node quadrilateral which can be used to
provide conservative estimates of the maximum frequency.  The maximum
frequency estimate for a rectangular quadrilateral element is
\begin{equation}
\hat\omega_{\max}^2 = {4(\lambda + 2\mu)\over\rho}\left({1\over s_1^2} +
   {1\over s_2^2}\right)
\end{equation}
where $s_1$ and $s_2$ are the lengths of the sides of the rectangle,
$\lambda$ and $\mu$ are Lame's constants, $\rho$ is the density, and
$\hat\omega_{\max}$ is the predicted value for the maximum frequency.
Similarly, for a rectangular parallelepiped hexahedron,
\begin{equation}
\hat\omega_{\max}^2 = {4(\lambda + 2\mu)\over\rho}\left({1\over s_1^2} +
   {1\over s_2^2}+{1\over s_3^2}\right)
\end{equation}

Substituting the maximum frequency equation into the stable time step
equation and neglecting damping gives the following estimate for the
stable time step size:
\begin{equation}
\Delta\hat t\le \sqrt{\frac{\rho}{\lambda+2\mu}}
      \left(\sum_{i=1}^{n_D}\frac{1}{s_i^2}\right)^{-1/2}
\end{equation}
where $\Delta\hat t$ is the estimate of the stable time step, and
$n_D$ is the number of dimensions.  The first quantity on the
right-hand-side of the inequality is the inverse of the dilitational
wave speed.  \NUM\ calculates the minimum value of
\begin{equation}
      \left(\sum_{i=1}^{n_D}\frac{1}{s_i^2}\right)^{-1/2}
\end{equation}
for each material block.  This quantity is called the `Minimum T
Factor' in the \NUM\ output.

The computer time required to perform the analysis can be approximated
using the estimated time step size.  Most explicit transient dynamics
codes output the average CPU time required to perform the calculations
for one element for one time step. Although this quantity varies for
different constitutive models and the number of contact surfaces
(slidelines), the average value is usually relatively constant and
well known.  The CPU time per millisecond of analysis time can be
estimated using the formula
\begin{equation}
{CPU\over ms} = \left(1\times10^{-3} \over \Delta\hat t\right)
                ({\sf Speed})({\sf NUMEL})
\end{equation}
where {\sf NUMEL} is the number of elements and {\sf Speed} is the CPU
time per element per timestep.  Since there are often other charges
associated with a computer analysis ({\it e.g.} memory or I/O
charges), this quantity will usually have to be modified; however,
after a few analyses, a rule of thumb can be developed which will
enable the analyst to very closely estimate {\it a priori} the
computer time required to perform an analysis.  An accurate estimate
of computer time is usually required for efficient batch job submittal
since there are often a limited number of highly-used job classes.
The capability to closely estimate the correct timelimit can reduce
the analyst's time spent waiting for a job to finish or the time
required to restart a job that was originally submitted with
insufficient time.

\chapter{Execution of the \NUM\ Program} The \NUM\ program is written
in standard FORTRAN; all system dependent coding is handled by the
SUPES library package~\cite{supes}.  The code is currently available
on the Engineering Sciences VAXCluster and both of the Cray computers
at Sandia National Laboratories.  The code reads the body definition
from a file written in the \EXO~\cite{EXODUS} format. The user is
prompted for the order of quadrature to be used to evaluate the
integrals and, for two-dimensional bodies, the type of mesh---either
axisymmetric (default) or planar.  The code then prompts for the
density of each material or element block. An optional material label
can be entered after the density. This label will be echoed to the
output to identify each material block.

The output of the code consists of a summary of the time the code was
run, the machine it was run on, the \EXO\ filename, and the title read
from the file. The number of nodes, elements, dimension (two- or
three-dimensional), and quadrature order are printed next. The output
then lists the density, volume, mass, and optional material label for
each of the material blocks, followed by the centroid location and the
mass or area moments of inertia.  The final section lists the minimum,
maximum, and average element volume or area for each material block;
and the corresponding minimum time factor. Figures 1 and~2 show
example input files for a two-dimensional axisymmetric and planar
model, respectively; Figure~3 is the input for a three-dimensional
model.  The corresponding output files are shown in Figures 4
through~6.

%
% conclusions and summary
%

\chapter{Summary and Conclusions} The code \NUM\ was written to
provide an efficient method for calculating several important
properties of a finite element model that cannot be easily determined
by other methods.  The code calculates the mass and volume of each
material in the model; the centroid location and mass moments of
inertia of the total body; the minimum, maximum, and average element
volume for each material; and the minimum time step factor. Some of
the obvious uses of this code include: model validation, density
calculation of pseudo materials required for proper mass
representation, section property calculation, determination of impact
orientations with respect to the center of gravity, estimation of
eccentric impact response, estimation of the computer time
required for an analysis, and determination of the element with the
minimum time step size.

Other features that may be useful and could be added in the future
include the calculation of element distortion parameters which would
be useful for validating automatically generated finite element
discretizations, and the calculation of the mass properties of the
deformed body using the displacements calculated during the analysis.
This last option could be used to calculate the deformed volume of
materials with nonlinear volumetric behavior such as the consolidation
of crushed salt and the crushup of foam used as an energy absorber.
\clearpage
\begin{figure}
\verbatiminput{2di}
\caption{Two-Dimensional Axisymmetric Input Example}
\end{figure}
\begin{figure}
\verbatiminput{2dip}
\caption{Two-Dimensional Planar Input Example}
\end{figure}
\begin{figure}
\verbatiminput{3di}
\caption{Three-Dimensional Input Example}
\end{figure}
\begin{figure}
\verbatiminput{2d}
\caption{Two-Dimensional Axisymmetric Output Example}
\end{figure}
\begin{figure}
\verbatiminput{2dp}
\caption{Two-Dimensional Planar Output Example}
\end{figure}
\begin{figure}
\verbatiminput{3d}
\caption{Three-Dimensional Output Example}
\end{figure}
\begin{thebibliography}{99}

\bibitem{EXODUS} Mills-Curran, W. C., Gilkey, A. P., and Flanagan, D.
P., {\em EXODUS: A Finite Element File Format for Pre- and
Post-Processing}, SAND87-2997, \SANDIA, in preparation.

\bibitem{slapdown} Sjaardema, G. D., and Wellman, G. W., {\em
Numerical and Analytical Methods for Approximating the Eccentric
Impact Response (Slapdown) of Deformable Bodies}, SAND88-0616,
\SANDIA, in print.

\bibitem{pronto} Taylor, L. M., and Flanagan, D. P., {\em PRONTO 2D, A
Two-Dimensional Transient Solid Dynamics Program}, SAND86--0594,
\SANDIA, March 1987.

\bibitem{dyna} Hallquist, J. O., {\it User's Manual for DYNA3D and
DYNAP: Nonlinear Dynamic Analysis of Solids in Three Dimensions},
UCID--19156, Lawrence Livermore National Laboratory, University of
California, Livermore, California, July 1981.

\bibitem{cook} Cook, R. D., {\em Concepts and Applications of Finite
Element Analysis}, Second Edition, John Wiley and Sons, 1981.

\bibitem{flanagan} Flanagan, D. P., and Belystchko, T., {\em
Eigenvalues and Stable Time Steps for the Uniform Strain Hexahedron
and Quadrilateral}, Journal of Applied Mechanics, 84-APM-5,
Transactions of the ASME, 1984.

\bibitem{supes} Flanagan, D. P., Mills-Curran, W. C., and Taylor, L.
M., {\em SUPES--A Software Utilities Package for the Engineering
Sciences}, SAND86-0911, \SANDIA, 1986.

\end{thebibliography}
\end{document}