Coyote
Coyote
Coyote
DEfl3 003Q22
COYOTES
1511
O c t o b e r 1982
I C E
PORTIONS OF THIS REPORT ARE ILLEGIBLE. It hat bttn reproduced from the best aviTiiSiF" Mpy to ptrmit the broadest possible avail ability. ABSTRACT COYOTE is a finite element computer program designed for the solution of two-dimensional, nonlinear heat conduction problems. ematical basis used to develop the code is described. complete user instructions are presented. The theoretical and math Program capabilities and
This work was sponsored by the U.S. Department of Energy under contract DK-AC04-76DP00789. A U.S. Department of Energy facility.
, ; r
m :
A'
TABLE OF CONTENTS
PREFACE INTRODUCTION FORMULATION OF THE BASIC EQUATIONS Continuum Equations Finite Element Equations Discrete Boundary Conditions SOLUTION OF THE FINITE ELEMENT EQUATIONS Transient Analysis Steady State Analysis PROGRAM DESCRIPTION Organization Mesh Generation Element Library Boundary and Initial Condi I ions Equation Solution Rezoning Heat Flux Calculations Plotting INPUT GUIDE Header Card SETUP Command Card FORHKF Command Card OUTPUT Command Card ZIPP Command Card MEATFLUX Command Card REZONE Command Card : ; i . : . ' 2! 22 :>: 2! 2D 20 H'i . ' ,I HI 34 3C 37 IS8 s:i 54 Hi; ; , . , r ; i . : :;
-
PLOT Command Card RESTART Command Card Program Termination Card Input Deck Structure User Supplied Subroutines Initial Conditions Error Messages Computer Requirements and Control Cards EXAMPLE PROBLEMS Grid Generation Heat Conduction in a Steel Bar One-Dimensional Heat Conduction in a Cylinder Heat Conduction in a Finned Radiator Heated Salt Block Heat Conduction in a Slotted Bar APPENDIX A: APPENDIX B: REFERENCES CONSISTENT UNITS TIME STEP ESTIMATION
'!'
liC'i-1^ I'iiKe 12 Hi
:i J 11 i:-. ! 1 .-l-.i-i; i ('.,!-, I ! - i I ( ' ( i n I i-'i I lo.nir.'] I v.-lV I ' d \ i . : l r 'Chi i arri lie: ('.-! i h e . C:\vci Pl'nl i: . :i'i 1 - : !.- 2 a " i b e r i n j j I' i i i i i i n i ' 47
- i : .'! . ! . -!'.<:.:,
w : ! n Hi-:-- i a r t u i l h i'sir
, ;
! i. < ( . - - - K . | : : '. . r ! - :
l i br n u r i ;IPR fM
' ! ; : ! ,!
ki'irw.l!
18 It) 20 21 22 22
.M.-sl
I'l.
1
!':!.!.;<!!! ! ' ! ! i on
I'i'i.M.r. 2
I s.-tr:-] 1 < I : . : I -v. 2 - - 1 i ii a Ii-ifim I ii:i, C\- I i n ' i c r I.i'.t 1 1 i .-.I i -.!,I \.t:-:i.l. !'i-i.I.
Mia 1 I l e a t
Confine I i on
i(l(i 109
:-v I ' l o t - -! M i n i p l i
1 1ST OF FIGTI'FS
iconti
I'titH.'
Example Problem 4Heat Conduction in a Finned Radiator input I,.!biinR--ExnripIe Probl m 4 (ontour Plots Example Problem 4 Example Problem 5Heated Sail Block ft
110 112 111! 115 116 119 121 122 123 124 124
Example Probler RHeat Conduction in a Slotted Bar Input ListingExample Problem 6 Temperature Profi1es--Examplo Problem 6 Contour PlotExample Problem G Time Histories, Special Points--Example Problem 6
-fi
-T
PRKFACK
At ihe time of release of t . h e first V(?rsinn of COYO'IT in r:id-]97S. i; not anticipated that the code would receive the heavy usage that i t . r;;rn:-' . enjoys. In response to user needs, COYOTK has undergone several minor \J>. rt However, continued roqUPSIS for addi f iona I cun^'
ities combined with significant changes in computer hardware dictated i l : ( ..< for a major upgrading of the code. version of the COYOTK program. In an effort to make the program more flexible and, therefore, niore . ( . > ' a number of capabilities and improvements have been added to COYUTI-.. Th- >\u The present report describes this l;)\>-~-
ratic, quadrilateral, and t ri aninil f t r finite elements have been supp ! erien: . : the addition of linear quad r i ! ; , t era ! s and triangles to the element libra:These elements are computationally inexpensive and are particularly useful ; simulating boundary value problems that contain discontinuous or nearly dis continuous behavior, e.g., change of phase. The time integration schcr.e 1 : 1 implicit i 1. >
Improved
flexibility in the modeli ng of problems has been achieved by -vnanding the parameter lists on user supplied subroutines. This allows both material pi variables. for
user convenience commands have been added to permit editing of the printed put as well as the definition of special computation/output points wnli.ii '>. rnesh . The plot capabi 1 i t i os of 1 he code have also been expanded . Final i y . i
f'OYOTK program has been made operational on the Cray Research S-l computer. This version of COYOTK allows problems of substantially larger sixe 10 be analyzed.
J
Unfortunately, the additions to COVOTF n: .-r"; ; i in; <:d <c,m<- sin:i! I chanros in the input and data syntax for the code. The m ' * version ~ f COVOTr. i-.. h.M-c-[""v For ex|j.r i nnccd ' . i a r r . s of
j ;
iNTPC'mm ION
:;:; :;).
\ :.>'<> :.'_*-,
-intk;c ; o;j
| r; >i c.
is
a i ornv >r;
; i i na -
a':a i y I : \e anai'.-i . - ti - \
upon some
, . ! < "irai
I'IIPL- i a ^
r ! ' ! : r i he :a;-:_>-
i<--a! parpose
as CINTJA
:.; : h ; i-: app d e a l i ons . The ^(^ir!'.']] pLir'rosc heat .alvsis toois for conduction ode.s , si:r:h as <"INDA, p r o v i d e : . ' , ' . . ifu : 'a a . od<
the e n g i n e e r . translates
Ilowf-vni' , Inherent i ;! o
often
lore- learning
] arjre m a n - h o u r
i nvos linen Is
in da I a pre para t i on
drawbacks
are especially
uspr of
subsi anl.l al
amounts of
code o p e r a t i o n
the
t O br i d m probler:.
purpose
T h e C O Y O T E code
is c a p a b l e of heat
transH-m
linear or n o n - 1 i n o a r
conduction
problems COYOTE
cert a in 1y
d > < : . not have the g e n e r a l i t y ''>.;e that " ; ' ! . !J< red. - ' IIDII , an w-s has an Since
ol c o d e s
Lhat
is easily Hie
is based on
finite e ! e n c m
i nlerf'ai. o wi fh numerous 2
f i n i te element input
is q u i t e s t r a i g h t f o r w a r d . to the
I-inaily,
the basic
for COYOTE
: ! . v i den t i cal
format s for t he
i neomp-vsy i b]e
1 . j i j r i . u s " 1 ou' code . MARIAII . allowing users to greatly increase their analysis
1
.>::;})' ' i ly by ie-urning a siiip'.fi d ; i < ; i input convention. The- iisisic prog?"an organisation and many of the user oriented programming
f
eaiuios of the present code are derived from ' , he solid mechanics finite element developed by Becker and iJujihan., Among the programming features
. I t - :i.XGAP.
. v e i u p . i i in the TEX GAP code and incorporated in the present program are: (1) Command j ; i o d e input which allows the user to specify the sequence of operations required for each problem. i'2) Free field input that eliminates the need to recall field width and format speci f i t . i t ion for input variables . {: S } An isoparametric mesh generation scheme that allows complex boundary shapes to be modeled easily and accurately. ( \) A frontal solution technique foi processing elements that allows great flexibility in element choice. (5) An automated coarse to fine grid rezone procedure that allows specific areas of a solution field to b c examined in detail without an undue cost in solution time. In the following sections of this report, a brief description is given of the theory and computational methods used in the COYOTE program. Also, an input
guide for use of the program is provided along with several illustrative example prolj 1 ems.
F'V
<r
FOOT ] ()\S
"']( ; .-:
. I--TU n'
IM| aa I i on:-: f o r
I r^ir-:
.-n p i' 'I' I cms i s WP 11 dr., !! n = < < ' r piT'srnt formulation To s i ' i: '
!" -\
t h e j;];;p.'-
irraiod
t h e axi s\-n;:< I r: m i i o n s
in a s t r m t r h i :':.-!
n ' in aum Kauai ions Thr a p p r o p r i at o ma r l v " : " i < a .-.-.i I i -ri a I rctfi on f? i s ^ i r e n by ilcsrriplion of t he h e a t m r . i'!
p
:>('
k. . i p J J
J
Q = 0
.vhore o . un.snr,
che m a t e r i a l
density, heat
("
capacity, time, x.
i
k. . i h c .'.ndi.-,- c o o r d : r:
0 the volumetric
source,
the srvatia!
is USSMV.C-:
;
homogeneous and either isotropic (i.e., k.. = k) or ortho t rop i c (i.e.. . where k. . is written in terns of the principle material direct ions'i,
r
:n
ature: the heat source Q may also depend on time, spatial location, ard
a! tin- .
The boundary of the region d. is defined by 7 = 7. + 7 , where ':. ;u. ;:( parts of the boundary for which the temperature and heat flux arc sn<
' . ' h e rel evant boundary eonditions for Equation (1 ) may then be exp res icd !
10
i i
K j !r:) i c *r =
+ q
*"
, . ' ere T. is an applied boundary temperature, q. is the applied heat fl ux vector, : i? . th*' unit outward normal to I he boundary qT , c q tir.Ti, and q the heat flux due to radiation. the heat flux due to convec-
^c q
( T
" V
( 4 >
= h (T - T )
r r
(S)
uh<-rc h
f-
and h
; i n < i T , are equilibrium temperatures for which no convection or radiation occurs. Thn radiation coefficient is given by,
2
= 0(T
+ T^)(T + T )
r
(6)
Note that
the boundary conditions given in Equations (2) and (3) may again be functions of time, spatial location, and temperature. Equations (1) through (6) provide a complete description of the boundary value problem for the temperature, T. When considering the transient heat
conduction problem, a suitable set of initial conditions describing the initial spatial distribution of T is also required. An approximate solution to this
class of problems may be obtained by discretization of the continuum problem through use of a numerical procedure such as the finite element method. M_n ite Element Equations The spatial discretization of the above boundary value problem by use of finite elements may be approached by either of two methods. Historically, the
Tarsi and most popular approach consists of rewriting the boundary value problem
11
.-!il
:m"
ini e
Elements
_,
><>-e
B u u' 'Ir ,
Nodal
Points.
FIGURE ] .
Assume t h a t w i t h i n e a c h f i n i t e j p p j ' o x i r . a t e d by ,
eJcrifnt,
.he
tettine :;i i
T(x.,t)
Ev
] r. ma t r i x n o t a t i on ,
Tfxj.t) = * ( ) e ( t )
X i
is an N dimensiona
Equation (7) into the partial differential equation, Equation (1), and boundary conditions, Equation (3), yields a set of residual equations, due to the approx imate nature of Equation (7). The Galerkin method guarantees the orthogonality
of the residual vectors to the space spanned by the interpolation functions. This orthogonality is expressed by the inner product,
(*,R) = J
*R d I J=0
(8)
enclosed by the element. Carrying out. the above operations explicitly for Equations ( 1 ) , (3), and (7) yields the equation,
/ Ja
*<pC * - - J M k . . ^ - 9 ) - Q > do -| p- at s x ^ u 8 y *J
T X j
/ J p
|q.n. + k. . - ^ -\ J i x 1 1 3 x
en.+q
+ q I dr = 0
(9)
Equation (9) may be rewritten using Green's theorem (basically an integration by parts on the second order derivative term) to give the equation, f 36 / e 3$
1
3*
J/ jj
e
pcp.*
-^ / at dn + Jo
= / * Q
d n
- f
*K i
% q)
r
d r
do
13
Several terms in Equation (10) contain material properties which are gen erally assumed to be functions of time, spatial location, and/or temperature. To accommodate such arbitrary variations within each element, it is convenient to approximate these material functions through a set of interpolation functions. Thus, let pC and k.. be approximated by
nique ma;' be used to allow the volumetric heat source within an element to have an arbitrary functional dependence. Thus, let Q = n Q where Q is a vector of nodal point volumetric sources.
T
(11) and (12) into Equation (10) produces the following equation
* n Q dft-/
*(q.n. + q^ + q ) dr
(1
Once the form of the interpolation functions, * and n are specified for an element, the integrals in Equation (13) may be evaluated. Such an evaluation
leads to a matrix equation for each element of the following form, MB + KB = F + F , (14)
14
M = f
n^pC *
'fl ~ ~ p ~ ~ e T
,
e
F
n
.^ - - 3x.
*TI Q
T
>. . , - dfi
dfi
F = -f
*(q + q
n
+ q ) dr .
r
The previous discussion was directed toward the derivation of the equations for a single element. The finite element model for the entire region 5 is ob
tained through assembly of the element matrices by imposing appropriate interelement continuity requirements on the dependent variable. Such an assembly 9 (the so-called "direct stiffness" method ) yields a matrix equation of the form given in Equation (14). Discrete Boundary Conditions As noted previously, boundary conditions for heat conduction problems may lie of several types. straightforward. The discrete form of a specified temperature condition is
for that nodal point is replaced by a constraint condition enforcing the boundary value. The specification of boundary conditions in terms of various heat fluxes requires slightly more computation. In Equation (14), the boundary fluxes to
an element appear in the vector F as an integral taken along the element boun dary (only element boundaries coinciding with T need be considered as contribu tions from interior boundaries are cancelled by adjoining elements). In order
to understand the procedure for computation of these boundary integrals, refer ence must be made to Figure 2 which shows a typical finite element boundary. Considering first the case of an applied normal heat flux to the element, the contribution to F is expressed by,
15
-f
*n
d r
(15)
where the * functions are restricted to the boundary. the boundary is s, then r
dF
2
+
2i
= [(Si
&
J ds = A ds ,
Or)
ds
(1C)
S= +1
S=-l
FIGURE 2.
16
" "/ * c = S
q dF
(17)
where the definition in Equation (4) was employed. element boundary is given by, T(s) = 4> 8 where again $ is a boundary or edge function.
T
coefficient can be variable, it is convenient to represent it by, hc (s) = * -h ~c Substituting these relations into Equation (17) and using the definition for dV produces 1 F / ~c = -J_
1 T
?c = ~ ?hc
where, C = / 1 T T 4 F, f ' -he = J_ ~ h c~ c
x
'
< >
1 8
* h J J A ds
c
ds
Note that in Equation (18), the term CB contains unknown nodal point temperatures and will thus be moved to the left hand side of the matrix equation in Equation (14). A computation similar to the one above may be carried out for the radia tive flux boundary condition to yield,
17
(19) where, R = f * h e A ds
T T
Again, the R8 term will be moved to the left hand side of Equation (14) while F, is retained on the right hand side. Equation (19) is complicated by the
fact that both B and F. -hr are functions of 8 since the radiative heat transfer coefficient, h (Equation (6)), is an explicit function of temperature.
To summarize the modifications to the basic matrix equation, Equation (1), due to the application of flux type boundary conditions, Equations (16), (18), and (19) may be substituted into Equation (14) to yield, F F + ~r F + -Q F~ ~n + -c
MB + KB = F_ + F - C6 + F. - R8 + F. ~~ ~~ -Q ~n -~hc ~~ ~hr
(20)
Rearranging Equation (20) allows the final form of the discrete equation to be written as, MB + K*8 = F* with , K* = K + C + R F* = F + F + F. + F. ~Q ~n ~hc ~hr
n
(21)
18
The discussion of the solution procedures for the matrix equation, Equation (21), naturally divides itself into two sections; transient problems and steady state problems "ransient Analysis A large body of literature ' ' is available on possible time integration Both implicit and explicit In order
to efficiently apply typical explicit integration schemes to the heat conduction equation, the capacity matrix M defined in Equation (21) is replaced with an "equivalent" matrix M_ which has non-zero coefficients only on the diagonal. This lumping procedure, which expedites the inversion of M as required in ex plicit schemes, has been widely studied for use with lower order finite elements (e.g., bilinear quadrilaterals and triangles). Unfortunately, no generally
valid procedure is available for constructing diagonal capacity matrices for higher order elements such as the biquadratic quadrilateral. Mode superposition
schemes can be very accurate and computationally efficient for linear problems. However, modal analysis requires at least some of the eigenvalues and eigen vectors of Equation (21) to be computed which complicates code structure. Also,
for nonlinear analyses, the modal method loses a great deal of its computational efficiency. The problems listed above were circumvented in the COYOTE program by im plementing a family of implicit time integration schemes. Implicit methods in
general have the advantage of increased numerical stability which allows the use of a wide range of permissible time steps. This increased stability is
19
Two basic integration procedures were selected for use in COYOTE, the gen eralized Crank-N'icolson procedure, which is in fact a family of methods, and a modified Crank-Nicolson algorithm. The generalized Crank-.N'icolson scheme when 12
f i (e)
M
AK*(e )lB
n + 1
= J- M(e)6
+ AF*
n + 1
+ (1 - A)F*
+ (1 - A)9
0 S
* < . , 1
In writing Equation (22), the general functional dependencies of each matrix and vector have been indicated. The superscript n indicates the time level,
At is the time step, and A is a parameter that determines where in the time interval t
n
to t +At = t
With specific a) A = 1
produces a forward Euler or fully implicit method, b) A = 1/2 produces the standard Crank-Nicolson method, c) A = 0 produces the backward Euler or explicit 9 (also known as Donea's
When the M and K* matrices depend on temperature and A > 0, then Equation (22) produces a nonlinear set of algebraic equations that must be iteratively solved at each time step. Various predictor-corrector or extrapolation methods
could be used to reduce the work involved in solving Equation (22) at a given step. In the present program, a quasi-linearization scheme is used that avoids Setting 9 =6 allows
all the matrices to be evaluated explicitly and the solution advanced to 8 in a single step. This assumption is quite reasonable for many conduction
problems with thermal shock and some radiation problems being notable exceptions. The second integration procedure used in COYOTE is a central difference method that is related to the standard Crank-Nicolson scheme. The derivation
20
12 13 '
M ( 6 ) + K*(9 )la
= F*(e ) + ~
M(6 )6
(23)
where ,
= (e
n + 1
+ e )/2
The procedure in Equation (23), termed a modified Crank-Nicolson method, is identical to a standard Crank-Nicolson method (A = 1/2 in Equation (22)) ex cept that the solution is not extended to the end of the time interval. This
lack of extrapolation accounts for the improved behavior of the solution with respet_ to temporal oscillations when compared to standard Crank-Nicolson solu tions. For nonlinear problems in which M and K* are functions of temperature,
the algorithm in Equation (23) again produces a set of nonlinear algebraic equations. which 9 Iteration within a time step is avoided by quasi-linearization in
is set equal to 6 .
Steady State Analysis For problems that are independent of time, the basic matrix equation reduces to, K*(9)-6 = F*(8) . (24)
Considering first the case where K* and F* are not functions of 8, then Equation (24) reduces to a linear matrix equation which may be solved directly. When Equation (24) retains its nonlinear form, an iterative technique is required. The present code uses a simple Picard iteration (successive
K*(e )'9
n J 1
= F*(9 )
(25) is equivalent to solving a time dependent problem using, for example, Equation (23) with At*".
21
PROGRAM DESCRIPTION
The numerical procedure described in the previous section has been imple mented in a FORTRAN coded program called COYOTE. In the following sections, a
general description of the code is given along with a discussion of some of the computational procedures. Organization The organization of COYOTE reflects the steps taken in setting up, solving, and evaluating a finite element analysis of a conduction problem. self-contained with its own mesh generator and plotting packages. The code is COYOTE is
written in an OVERLAY form with each overlay handling a specific task in the analysis process. Transmission of data between overlays is through low speed
disc files and extended core storage (ECS) as shown in Figure 3. As indicated by Figure 3, the main overlay acts as a calling routine to the remaining primary overlays. The following list provides a brief description
of the functions of each primary overlay. (1) OVERLAY (1, 0) SETUP reads material property data, element and
boundary condition data, creates mesh, organizes data for equation formulation. (2) OVERLAY (2, 0 ) , OVERLAY (10, 0) FORMKF reads material property,
element and boundary condition data from SETUP, forms coefficient matrices for each element and applies boundary ;onditions, creates unique "nicknames" for each degree of freedom. the axisymmetric version. (3) OVERLAY (3, 0) ZIPP reads element coefficient matrices, assembles Overlay (10, 0) is
global coefficient matrices using "nicknames" to establish element connectivity, solves either transient or steady state matrix equation.
22
Input
/TV
COYOTE (0,0)
_/^\
Restart
( 7
v
y^\
output
SETUP (1.0)
FORMKF (2,0)
2IPP (3,0)
REZONE (4,0)
HEATFLX (5,0)
PLOTZ (6,0)
POINT (7,0)
PREFRN1
Plot Data
<tc^>-
Jl
Overlay
9 W
7 W
Disc File
Special Points
FIGURE 3 .
S c h e m a t i c D i a g r a m of COYOTF.
( 1 )
O V E R L Y (4, 0) nrZONT- -- refines mesh within a specified ^gian , organizers data for resolution of refined grid.
(5)
flux quantities for selected elements. (6) OVERLAY (6, 0) PLOTZ temperature histories. (7) OVERLAY (7, 0) POINT points within the mesh. COYOTE was originally designed for use on Control Data Corporation (CDC) series 6000 and 7000 computers. The code has also been made operational on the In the Cray version of COYOTE, the determines the location of special output plots grid point.-;, elements, isotherms, and
OVERLAY organization is replaced by a subroutine structure and data previously stored in ECS is relocated to main memory. The CDC version of COYOTE is pre
sently sized to accommodate grids with up to 500 elements: the Cray version allows up to 1000 elements. Increases in the element capacity of both versions
may be made through changes in several dimension statements in the code. Mesh Generation The generation of grid points in COYOTE is achieved through an isoparametric mapping technique developed by Womack codes, ' ' ' 14 and used in several previous
distinct operation in COYOTE and independent of the specification of the element connectivity. This separation of operations allows the user to generate more
nodes than may actually be used in the problem and to experiment with nodal point placement before element connectivity is established. These options are
especially useful when gridding large and/or geometrically complicated problems. For purposes of grid construction, the domain of interest is considered to be made up of ^arts or regions which are determined by the user. Within each
part, an isoparametric mapping is used to approximate the region boundary. Specification of a number of x, y (or r, z) coordinates on the boundary of each
region determines the limits of the region and the type of interpolation used
24
Complex,
curved
boundary includes
shapes
c*n
be
easily or
this s c h e m e w h i c h shape .
linear, quadratic,
! i . i b i <: i i , loj-puiat i on
'he b o u n d a r y a region
The r n e ; d i points within < : l ~ i) ides along each b o u n d a r y ing sysi cj! is used v . j hin
L
are g e n e r a t e d Within
once
the
number
is s p e c i f i e d .
an I, J
number
to identify
points.
point
spacing
the g e n e r a t i o n
of m e s h may
points.
A user s u b r o u t i n e by she
to g e n e r a t e m e s h
points
. ! - : .:'dinf.r to a
formula s u p p l i e d
user.
approximation func-
the e l e m e n t
geometry.
The
form of
the s h a p e types of
lions
depend
the element
being
four b a s i c
be J ow.
A.
Quadrilateral The
Elements elements used in C O Y O T E 4 and 5. the are -.he s t r a i g h t curved-sided, field given
sided,
four n o d e
shown
in F i g u r e
shown
in F i g u r e using
is a p p r o x i m a t e d
the b i l i n e a r
t)
( 2 0 )
- t)
(1 + s)(l + t) (1 - s)(l + t) whore the ordering of the functions corresponds to the ordering of the nodes in
Figure 4.
The temperature in the eight node element is represented by the biquadratic functions which can be expressed as,
25
FIGl'H! ' 1 .
Quadrilateral Element
FIGURE 5.
26
M = i
The shape functions in Equations (26) and (27) arc expressed in terms of the normalized or natural coordinates for the element, s and 1, which vary from -1 to +1 as shown in the figures. The relationship between ihe x, y (or r, z)
coordinates a.nd the natural coordinates is obtained through the isoparametric mapping concept discussed by Ergatoudis, et a!. That is, the coordinate
x = N x
N y
(28)
(29), x and y are vectors of coordinates for points on the element boundary (generally nodal point coordinates). Using the relations in Equations (26) through (29) and some algebraic manipulation allows the matrix coefficients defined in Equation (14) to be written as integrals of rational functions of the s, t coordinates. ation of these integrals requires the use of numerical quadrature. The evalu The COYOTE
program employs a 3x3 Gaussian integration for the evaluation of the matrix coefficients for a quadrilateral element. H. Triangular Elements The companion elements to the previously described quadrilaterals are the straight-sided, three node (TRI3/3) triangle shown in Figure 6 and the
27
0,
FIGURE 6.
Triangular Element
FIGURE 7.
28
cur vud-rfj tied, s i x n o d e ( T R I G / 3 and TUIfi/6) interpolation fund ions for these
t r i a n g l e .shown i n F i g u r e 7. by,
The
two t - l e n i e n t s a r e g i v e n
"
l 2
(30)
':i
. ' or
the
three n o d e e l e m e n t
and
l.(2I.. I. (21.
3 3
1) 1)
(31)
functions are expressed in terms of the natural coordinates for a triangle g (see, e.g., Zienkiewicz ) where (32)
The coordinate transformations needed to implement the triangular elements directly parallel those for the quadrilateral elements and need not be repeated. Note that when performing the numerical evaluation of the integrals
in Equation (14) for a triangular element, the COYOTE program uses a seven point quadrature formula developed by Hammer-, ot al . Boundary and Initial Conditions Boundary conditions arc accepted hy COYOTE on an element basis. Specifi
cation of temperature is possible on both a nodal point and element side basis with the latter type being restricted to having a uniform value along the side. Specified boundary heat fluxes are applied by element side; the flux is assumed
29
boundary condition for the heal conduction problem and is obcai ned l > y default in COYOTE, i.e., no boundary --nndi Lion is specified for an adiabatic surface. Convective and radiative \,,,;i\):\.,ry the omissi vi ty , t, conditions are also specified by clement, side, are assumed uni
convective fluxes) boundaries may vary arbitrarily in time through the use of user-supplied time functions. A volumetric heat source may be specified for any
material as either a constant or as a user specified function of time, spatial location, and/or temperature. Initial conditions for the solution of transient problems or nonlinear steady state problems mny bo input, by two methods. Through the standard data
input, the temperature field may be initialized at a different uniform tempera ture for each material. Arbitrary initial conditions may be input through a
user supplied tape file written in a specified format. Equation Solution The solution algorithms used to solve the matrix equations obtained from the finite element discretization were presented in a previous chapter. The
actual processing of elements during the solution phase is accomplished by use 18 of the frontal solution technique developed by Irons. The frontal method i s a Gaussian elimination procedure that operates with a single element at a time. As each succeeding element is processed, the coefficient matrices for the ele ment are added to the global coefficient matrix. Whenever a particular nodal
point unknown is found to be complete (i.e., all the elements contributing to a node have been processed), that particular equation is condensed from the system. This process allows the minimum number of equations to be kept in core The condensed equations are stored in
Extended Core Storage (ECS) until they are needed in the back-substitution phase. The logic for this elimi nation scheme is setup by an abstract elimina
tion procedure called the pre-front which occurs just prior to the actual solution. 30 The pre-front coding is located in a secondary overlay.
in heat conduc
tion analysis though it has found some previous use in solid and fluid mechanics. For regions in the domain of interest that cannot be finely gridded for reasons of economy or computational limits, the rezone technique provides an efficient method of increasing the locrl detail of a solution field. Basically, the re-
zone procedure consists of refining a part of the original coarse grid, applying boundary conditions to this region based on the coarse grid solution and solving the resulting boundary value problem. This process has been automated in the present code and allows local grid refinement to be made within regions containing quadrilateral elements. The
rezone region is specified by establishing a quadrilateral shaped boundary around the region c interest with all of the quadrilateral elements lying en After the region
to be rezoned is specified by the user, appropriate temperature boundary condi tions are automatically interpolated from the coarse grid solution and applied along interior boundaries of the newly refined grid. The rezone data is organ
ized sucf that subsequent calls to the element formulation and solution overlays alio"- computation of the temperature field within the rezone region. The pre
sent version of COYOTE only allows rezoning to be carried out for steady state problems. The decision whether to rezone a region or use an initially refined grid is very problem dependent and no clear guidelines for making this decision are currently available. Until a wider experience with the technique has been ob
tained, individual user experience and experiment remain the best guides. Heat Flux Calculations The calculation of heat flux values for a heat conduction problem is pro vided in COYOTE as a user option. directly from the definition, The calculation procedure employed follows
i - ij fir
< >
31
33
y = N y
or
y = M y
(34)
where the shape function vectors in liquation ( ; M ) are those described in the element library section. Direct substitution of Equation (34) into Equation Since
(33) shows that derivatives of the interpolation functions are required. the interpolation functions are in terms of the natural coordinates for an element, the following relations are needed fusing the N functions as an example),
a*
11
F-,o 12
-^
(35)
3*
22 where, 3N 11 3t
3.N
Jt
3l~
3N
F
'21 [J]
W
F
22
3l~
12
F 21
Qsing Equation (35), the definition of the heat flux components becomes, 3*
q
- x*( ii
it
34'
+ F
12
ITJT
3s
22 3
TJT 1-sW
(36)
32
l 2 TT
?)T7T
-k fr
.'ore* t !ia t Cor a ir i a nu'ul n r el einon L , the
22
?)TTT
' i ! h the expressions in liquation ( 3G } , the heat flux components may be calculated at any point, s . t \ within an element once the element geometry Ca1eu1 ati on of fluxes for axisymCOYOTE provides evaluation of the
: ' ! * ? t ; ' i c . problems follow a similar procedure. heat 11'.IN at points on the clement boundary.
In addition to the heal flux components, COYOTE also calculates the heat flux norma) to the element boundary. to an element boundary is, n = nx e x + nyey and the heat flux vector is, (38) Referring to Figure 8. the outward normal
(37)
+y qy n
l
(39)
where the flux components are given by Equation (36). components of the normal are,
n --*
y
(40)
(If)
V 3s J
33
Fff'ii ! ! - .
'.fii;M !r,n
f o r Heal
F l u x C'ompu I ai. i on
1 si
i'!(! y in K q u a l i o n
the
nori'in I in K t p i a i i o n ! run
t h e normal
i ho dof i n i L ion
s the t o t a l
l>y t a k i n g t h e h e a t
to the boundary
;wr;i .
' I ho COYOTi; prou rair. m n i n i n s iniciiiin of a varici;. . , ' n]ois Plots of nodal n^idii points, at
its own
plotting
packu^'
that
allows
the e o n -
in heal, transfer
analysis. analysis
are generated
of the user.
histories,
flux h i s t o r i e s , and
temperature
p r o f i l e s are a v a i l a b l e
:*!
INPUT GUIDE
The structure of an input data deck for the COYOTE computer code directly reflects the steps required to formulate and solve a finite element problem. Through the use of a series of command and data cards, the program is directed to such functions as grid generation, element construction, solution of the matrix equations and calculation of auxiliary data. The actual sequence of
commands to the program is quite flexible, though there are some obvious limits to the order in which operations can be specified. In the following sections,
the command and data cards required by COYOTE are described in roughly the order in which they would normally appear in an input deck. problems follow this section as an added aid to input format. In the following section, the conventions listed below are used in the description of input cards: (a) (b) Upper case words imply an alphanumeric input value, e.g., FORMKF. Lower case words imply a user specified value for the indicated variable, e.g., xmax. (c) All numerical values are input in a free field format with successive variables separated by commas. characters under this format. All input data is limited to ten WARNING: The CRAY version of COYOTE Several example
will read data with ten characters/word, however, only the first eight characters will be used. (d) [] indicate optional parameters which may be omitted by using suc cessive commas in the variable list. If the omitted parameter is not
followed by any required parameters, no additional commas need be specified. (e) < > indicates the default value for an optional parameter.
35
(f) The * character may be used to continue a variable list onto subse quent data cards. When the * character is used, it should replace the last comma in the data string. (g) The $ character may be used to end a data card allowing the remaining space on the jard to be used for comments. (h) The contents of each input card are indicated by underlining. (i) All quantities associated with a coordinate direction are expressed in terms of the planar x-y coordinate system. The corresponding
quantities for axisymmetric problems are obtainable from the associa tion of the radial coordinate, r, with x and the axial coordinate, z, with y. The descriptions of the command cards are presented in the following order: (a) Header Card (b) SETUP Command Card (c) FORMKF Command Card (d) OUTPUT Command Card (e) ZIPP Command Card (f) HEATFLUX Command Card (g) HEZONE Command Card (h) PLOT Command Card (i) RESTART Command Card (j) Program Termination Card Following the individual descriptions of the command cards are sections dis cussing input deck structure, user subroutines, initial conditions, error messages, and computer requirements for COYOTE.
36
Header Card
The header card must be the first card in a deck for any particular problem. If two or more problems are run in sequence, the header card for each new problem follows the END, PROBLEM card of the previous problem. A S symbol must
appear in Column 1; the remaining 79 columns are available for a problem title. The header card is of the following form: $ PROBLEM TITLE
37
The first task in formulating a finite element analysis of a problem in volves the specification of the material properties and the definition of element mesh and boundary conditions for the problem geometry. These functions
are accomplished through the SEfUP command and its three sets of associated data cards. The SETUP command card has the following form: SETUP, [iprint], [maxi], [order], [grid plot]
where, iprint <2>: determines the amount of printout produced during the setup oneration. Outnut increases with the value of iprint and
ranges from no printout for iprint = 1 to full printout for iprint = 4. maxi<18>: is the maximum number of I rows to be generated in the grid. Maxi need only be specified if there are more than 18 1 rows or more than 110 J rows. The limit on the maximum I's and
J's is I*J 4000 for the CDC version, and I*J < . 5000 for the Cray version. order<>: determines the numbering of the elements. For the default
value (i.e., order left blank), the elements are numbered by increasing I, J values (e.g., (1,1), (2,1), (3,1) ... (1,2), (2,2) . . . ) . For order = PRESCRIBED, the elements
are numbered according to their order in the input list. The element ordering should be chosen such that the front width of the problem is minimized.
38
If a
plot of the grid points is to be made in a subsequent call to the plot routine, then grid plot = PLOT; if no plot of the grid points will be made, the grid plot parameter is omitted. SETUP Data Cards Following the SETUP command card, three sets of data cards are required for material property specification, grid point generation, and element and boundary condition specification. Each of the data sets is terminated by an END card.
The last END card (i.e., the third), ends the setup portion of the program and readies the program for the next command card. Material Data Cards: The material data cards are of the following form: [Material Name], number, p, C , k.., k, 6,
END
where, Material Name: reference, number: is the internal reference number for the material. COYOTE will is an optional alphanumeric material name for user
ll' 22 * ll'
o n e n t s
(see below). B<0>: is the angle in degrees between the principle material axis and
39
material properties on temperature, time, and/or spatial location. If all material properties (p, C , k..) are constant, this parameter is omitted or set to CONSTANT. If one os more oi the properties are
variable, this parameter is set equal to VARIABLE. Q<0>'. prescribes the volumetric heat source for the material. For no
volumetric heating, this parameter is omitted: for constant volumetric heating, the parameter is set to the heating value. If the heat
source varies with time, spatial location, or temperature, this para meter is set equal to VARIABLE. initial temperature <0>: for the material. The material models allowed in COYOTE include homogeneous materials with either an isotropic or orthotropic conductivity tensor. For isotropic materials
2
(k. . = k ) , the conductivity is specified by setting k,. = k; the parameters k and S are omitted. For orthotropic materials ( k = S^.k
= k . ) , the conduc
tivity tensor is determined by specifying components of the tensor with respect to the principle material axes. Referring to Figure 9, the 1 and 2 axes indi
cate the principle material axes while 6 specifies the orientation of the mater ial with respect to the spatial reference frame. COYOTE requires k,. and k to
be specified for an orthotropic material; ( 5 must be specified only if the mater ial axes are not aligned with the coordinate axes. Note that B is measured from
the positive x(r) axis and is positive in a clockwise direction. The variation of material properties and volumetric heating with time, space and/or temperature is indicated by setting the parameters "temperature dependence" and "Q" equal to VARIABLE, respectively. The actual variation of these quanti If variable material
properties have been indicated, COYOTE requires SUBROUTINE CAPACIT and SUBROUTINE CONDUCT to be supplied by the user; a variable heat source requires SUBROUTINE SOURCE. Further details on these subroutines are given in a later section of
this chapter.
40
FIGURE 9.
COYOTE does not contain dimensional constants and, therefore, the units for the material properties are free to be chosen by the user. a table of consistent units is given in Appendix A. Grid Data Cards: Following the material specification, the grid points for the finite ele ment mesh are generated. In contrast to many finite element codes, the COYOTE For convenience,
code separates the generation of nodal points and the generation cf element? into distinct operations. The calculation of nodal point locations is accom
plished by use of an isoparametric mapping scheme that considers quadrilateral parts or regions of the problem domain separately. For each part, a series of
coordinates are specified which determine the shape of the region boundary. The node points within each part are identified by an I, J numbering system; the I, J axes must form a right handed coordinate system. The location of
points in a region is controlled by user specification of the number of points along a boundary side and a local gradient parameter. These ideas are more
41
clearly fixed through a description of the data cards required to generate points in a part. For each part or region in the mesh, three data cards of the following form are required: imin, jmin, imax, jmax, |gl1, [g2] , [g31, [g41, [POLAR], [xol, lyol
[yi2l
END where, imin, jmin, imax, jmax: are the I, J limits for the region being gener The difference between the maximum and
minimum values determines the number of grid points generated in a particular direction. gl<l>, g2<l>, g3<l>, g4<l>: specify the gradients for the node spacing The gradients are defined by,
and are illustrated in Figure 10a. equal node spacing along a side.
than unity may be used to bias the spacing in either direction. POLAR, xo<0>, yo<0>: specify the use of polar coordinates for describing With POLAR speci
fied, the x's and y's on the second and third data cards are inter preted as polar radii and angles referred to the local origin xo, yo. The polar angle is referenced to the positive x axis; positive angles are measured in a counterclockwise direction.
42
7n
i m i n , jmax
xlO.ylO
'
Y 9
"2.V2
(c)
AHC Nomenclature
FIGURE 10.
43
y i,
xl x2 xl2' -. ' ' "* ',o! define the coordinates of the four corner and optional
T i 0
ya ,
...
y 1*5.
only, the four corner coordinates (i.e., xl to x4 and yl to y4) need be specified. If any of the region sides are curved, then the appro If one
side node is defined, a quadratic interpolation is used to define the boundary; specification of two side nodes allows a cubic interpola tion. Side nodes should be located near the midside and third points
for quadratic and cubic interpolation, respectively. There are no limits on the number of parts which may be used to define a grid. The only restriction on the total number of grid points is that max
(I*J) 4000 for Lhe CDC version of COYOTE and (I*J) <. 5000 for the Cray version. Node points may be generated for a triangular shaped region by allowing two adjacent corners of the quadrilateral part to coincide. However, when tri
angular meshes are created in this manner, the location of interior node points is generally unpredictable. The user is advised to verify the quality of such
a mesh through use of a node point plot. In generating mesh points for complex geometries, it is often convenient to be able to position an individual node point or a line of nodes. These
situations are provided for in COYOTE by use of the following data cards: POINT, i, j , xl, yl, [POLAR 1, [xol, [yo] where, i, j: xl, yl: is the I, J name for the point, are the coordinates for the point. specify the use of polar coordinates as described
44
ARC, imin, jmin, imax, jmax, fgll , [POLAR], [xol , [yol xl, x2, fx31, [x4] yl, y2, fy31, fy41 where, imin, jmin, imax, jmax: Figure 10c. are the I, J limits for the arc as shown in
;
Since a one-dimer.~ .onal array of points is being gener The difference between the
maximum and minimum values determines the number of grid points generated along the arc. gl<1.0>: specifies the gradient for the node spacing along the arc.
The gradient is defined by, , _ node spacing at ijmin _ Aijmin ** node spacing at ijmax Aijmax and is illustrated in Figure 10c. POLAR, xo<0>, yo<0>: previously. ' y2, ' y3, ,' y4 " ': define the coordinates for the ends of the arc and the yl, optional intermediate points. If the arc is a straight line, only The generation of a
v] vp
v"3
v4 '
curved arc requires the specification of one or two intermediate points as shown in Figure 10c. There are no limits on the number of POINT and ARC data cards that may be used in generating a mesh. Both types of cards may appear at any point within the
grid point data section of the SETUP command. In certain circumstances, it is convenient to specify nodal point locations according to a given formula or a specific pattern that is not available in the basic mesh generator. These situations are provided for in COYOTE through use This subroutine is accessed by the following
45
EXTDEF
ThJs data card causes the user supplied subroutine EXTDEF to be called by the mesh generator. Within this subroutine, the user may create an arbitrary array The EXTDEF data card may be used in conjunction
with the standard nodal point generation schemes or may be used to create an entire mesh. The data card may aopear at any appropriate point within the grid Details on the form of SUBROUTINE
point data section of the SETUP command. EXTDEF are given in a later section. Element and Boundary Condition Tata Cards:
Following the generation of the grid points, the program is ready to accept element and boundary condition data. Since the mesh points for the problem
geometry are generated independently of the elements, the selection of nodes from which to construct a given element is vory flexible. The actual construc
tion process for an element consists of identifying an appropriate group of previously generated mesh points that will serve as the corner and midside nodes of the element. This concept is apparent from the form of the element data card
element type, mat, il, j 1, fi2], [ j2] , ... [in], f . j n l where, element type: is an alphanumeric name for the type of element. The
element types used in COYOTE are described below. mat: is the material number for the element. This number should be set
to correspond to the material number used on the material property card, il, jl, [i2], [,j2] , ...: in the element. is the list of I , J values for the node points
around the element starting with any corner as shown in Figure 11.
some situations, the list of I, J values may be significantly conden sed. When only the first node I, J values are specified for a quadri
lateral element, the following values for the remaining nodes are assumed, 46
4 Node Element (QUAD4/4): i2 = 13 = 11 + 1, 14 = il j2 = jl, J3 = j4 = jl + 1 8 Node Element (QUAD8/4, QUAD8/8): i4 = i8 = il, 15 = i7 = 1 1 + 1 , i2 = i3 = i6 = il + 2 j2 = j5 = jl, j6 = j8 = jl + 1, j3 = j4 = j7 = jl + 2 When I, J values are specified for only the corner nodes of an eight node quad rilateral, the midside I, J values are computed as the average of the corner values. All of the appropriate I, J values for a triangular element must be
specified explicitly.
47
The specification of different types of elements is provided by the "element type" parameter on the previously described data card. element names for this parameter include the following: (a) QUAD4/4: a four node (isoparametric) quadrilateral with arbitrarily The permissible
oriented straight sides. (b) QUAD8/4: An eight node (subparametric) quadrilateral with arbitrarily
oriented straight sides. (c) QUAD8/8: sides. (d) TRI3/3: a three node (isoparametric) triangle with arbitrarily an eight node (isoparametric) quadrilateral with curved
oriented straight sides. (e) TRI6/3: a six node (subparametric) triangle with arbitrarily oriented
straight sides. (f) TRI6/6: a six node (isoparametric) triangle with curved sides.
Note that in the generation of the eight and six node subparametric elements, the physica] coordinates (x, y) of the midside nodes need not lie precisely on the element side as these coordinates are not used in the element formulation. During the construction of the element mesh, two points about the 1, J identification scheme should be noted. Each element is identified internally
by the I, J values for the first node named on the element data card, i.e., il, jl. Since any corner node may be named first, two or more elements may share This situation must be avoided if any
of the elements sharing names are to have boundary conditions imposed on them. A simple reordering of the nodes in the appropriate elements remedies this situation. Secondly, during the generation of grid points, there is no require When
elements are constructed along boundaries of such adjacent parts of the grid, it is imperative that common nodes between elements have the same I, J values. Element connectivity is generated from " ^ I, J values for each node and, therefore, common nodes with different I, J labels will not be properly connect ed. The generation of element meshes using node points with non-continuous I, J
The boundary conditions for the problem are specified by element and may appear at any point in the present data section after the element to which they apply has been defined. Boundary conditions are specified to have either a The
uniform value along an element side or a particular value at a node. boundary condition data card has the following form: BC, b.c. type, il, jl, side/node, value/set no./time curve no. where, b.c. type:
The types of boundary conditions used in COYOTE are described below. il, jl: is the I, J identification of the element (i.e., the first 1, J
named on the element data card) to which the boundary condition applies. side/node: identifies the side or node of the element to which the The numbering of nodes and sides
begins with the identifying node (i.e., the first node named on the element data card) and proceeds counterclockwise as shown in Figure 11. value/set no./time curve no.: is the numerical value of the applied
boundary condition, the number of the boundary condition SET in the case of convective or radiative boundary conditions or the number of the time history curve for time dependent boundary conditions. For
specified temperature or heat flux boundary conditions, this parameter is set to the numerical value of the boundary condition (e.g., 100.0 C 2 or 50 watts/m ). The specification of a set or time curve number is explained below. The permissible types of boundary conditions as specified by the paramete."b.c. type" include the following: (a) (b) T: specifies the temperature at a node. specifies the temperature to have a constant value along an
TSIDE:
element side.
49
(c)
TVARY:
function of time. (d) QSIDE: specifies a constant heat flux (energy/unit area/unit time)
along an element side. (e) (f) QVARY: QCONV: specifies a time dependent heat flux along an element side. specifies a constant convective heat transfer process along
an element side. (g) QRAD: specifies a radiative/convective heat transfer process along
All of the above boundary conditions can be employed with any of the previously described elements. Note that in the present version of COYOTE, only one QRAD.
QCONV, QVARY, and TVARY boundary condition can be specified for any single element; other boundary condition types are not restricted. The boundary condition options, TVARY and QVARY, permit temperatures or heat fluxes along an element boundary to vary with time. The time histories of
the boundary condition are input through a set of user supplied subroutines (SUBROUTINE CURVEn). The association of a particular time history with the
appropriate boundary condition is accomplished by use of the "time curve no." which appears on the boundary condition data card and in the user supplied sub routine name. input. section. The use of convective and radiative boundary conditions requires the appro priate heat transfer coefficients (h , h ) and reference temperatures (T to be specified. These parameters are input through use of a SET data card T ) COYOTE will allow a total of six different time histories to be
where,
50
b.c. type:
for which data is being specified, i.e., either QCONV or QRAD. set no.: is the number of the particular data set. The code allows up The set no. is
used to identify the appropriate boundary condition set on the BC card. h, T: are the parameters required for specifying a convective or radia For convection, h = heat transfer coeffi For fourth power radiation, h = emissivity For generalized radiation/
The generalized radiation/convection condition mentioned above has been provided to allow an arbitrary variation of h with temperature or time in the flux expression, q = h(T, t)(T - T )
r
To incorporate such a boundary condition, the parameter "b.c. type" must be set to QRAD, "h" must be set to VARIABLE, "T" is set to the appropriate reference temperature, and the user supplied subroutine, HTCOEF, which describes h(T, t) must be appended to the COYOTE program. in a later section of this chapter. The SET data cards may appear anywhere in the present data section; SET cards for both types of boundary condition may appear in the same problem. Looping Feature: In order to permit easy specification of elements and boundary conditions which appear in regular patterns in the mesh, a looping feature is incorporated in COYOTE. This feature allows the definition of ILOOP's and JLOOP's (which This subroutine is described in detail
are similar to FORTRAN DO loops) for incrementing data in the I and J directions. He. ting of the loops may be in any order but no more than one loop in a given direction may be used at one time. Note that within each loop, all the I (or J)
51
values are given the same increment. form: I LOOP, nioass , inc
or
JEND
specifies the number of passes to be made through the loop, specifies the increment to be added to the I or J values found The "inc" parameter may be negative.
The looping commands may appear at any point within the element, boundary condition data section of the SETUP command. illustrated in the example problems. The use of the looping feature is
52
With the completion of the SETUP command, the next task in the analysis sequence is the formulation of the matrix equations for the individual elements in the mesh. This function is carried out by the command card, FOHMKF, [geometry 1 where, geometry: is an alphanumeric name to indicate the type of coordinate If geometry is omitted, the two-dimensional planar
system desired.
53
Prior to solution of the heat conduction problem, it is sometimes conven ient to limit the amount of printed output generated by COYOTE. It may also be
desirable to specify spatial locations within the mesh, other than nodal points, at which temperature output is required. through use of the OUTPUT command card. Both of these functions may be invoked The selective printing of the computed
temperature field is set up with the following command card: OUTPUT, delimiter type, nl, n2, n3, ... n50 where, delimiter type: is an alphanumeric name to indicate how the subsequent
element numbers are to be interpreted in limiting the printed output. If delimiter type = SINGLE, individual element numbers are assumed to be listed in the following data. If delimiter type = STRING, the
following data is interpreted as pairs of element numbers with each pair defining a sequence of elements, nl, n2, ... n50: is a list of element numbers indicating which elements A
are to have temperature data printed during the solution process. maximum of fifty individual element numbers or twenty-five element pairs may be specified on a single command card.
There is no restriction on the number of OUTPUT command cards of this form that can be used in a COYOTE input deck; OUTPUT cards with different delimiter types may be mixed in the same input deck. The OUTPUT command card(s) must preceed If no OUTPUT command is used in a
COYOTE input deck, the entire temperature field is printed at each output interval.
54
The specification of special temperature output points is achieved with the following form of the command card: OUTPUT, POINTS, x
r
y^
* , y . ... x ,
2 g 2 5
y^
tions for the special points. A maximum of twenty-five points may be specified on a single command card. COYOTE allows a maximum of fifty special points to be specified during any par ticular analysis. The OUTPUT command card in this form must proceed the ZIPP command card; OUTPUT command cards of both types may occur in the same input deck. The temperatures at the special output points are printed at each normal
output interval. The temperatures at these points are also stored for possible later plotting as temperature histories.
55
The assembly and solution of the global system of matrix equations is carried out by the ZIPP command card. is of the following form: ZIPP, STEADY, [no, iterl, [iprint], [toll, [initial conditions! END where, no. iter<l or 5>: specifies the number of iterations to be used in solving the steady state problem. The default values of 1 and 5 For steady state problems, this command
correspond to iteration limits for linear and nonlinear problems, respectively. iprint <no. iter -1>: specifies how often the solution field is printed. tol <0.0>: specifies a convergence tolerance that may be used to When the largest difference in temperature
between successive iterations falls below the specified tolerance, the iteration process is terminated. All nodes in the mesh are
checked for adherence to this criteria. initial conditions < > : specifies the source of the initial estimate for the temperature field. If this parameter is omitted, the initial
estimate for the temperature is set from data on the material property cards. If this parameter is set to TAPE, the initial temperature
field may be input through a tape file as explained below. The solution of steady state problems requires the specification of only one ZIPP command; for compatibility with the transient solution option this command is terminated by an END card.
56
The solution of transient problems is carried out by a command card of the following form: ZIPP, integration type, [X] , [iprint] , t At, . [tfjnal '
1
l n i t j L a l
END where, integration type: is an alphanumeric name to indicate what type of When integration type = TRANS1,
the modified Crank-Nicolson procedure (Equation (23)) is selected. For integration type = TRANS2, the generalized Crank-Nicolson family of integration procedures (Equation (22)) are selected. X<. 5>' specifies the weighting parameter for the generalized CrankFor X = 1.0, a forward Euler method is
produced, X - .5 produces a Crank-Nicolson method, and X = 0 produces a backward Euler method. The X parameter may have any value in the
range 0 <. X <. 1.0. This parameter is omitted for the TRANS1 integra tor, iprint <10>: specifies how often the solution field is printed.
U : s
tinitial' tfinal'
Pe"fy
t h e
i n l t i a l
, t
f i n a l
) and
the time step (At) for the time integration procedure. no. steps: indicates the number of time steps to be taken in the tran
sient analysis. initial conditions: the problem. specifies the source of the initial conditions for
parameter is set equal to TAPE, initial conditions may be input through a tape file as explained below.
57
lumping:
specifies if the capacitance matrix is to be diagonalized If this parameter is omitted, a consistent capacitance
(lumped).
matrix is used; when lumping is set to LUMP, the capacitance matrix is diagonalized. This option is only valid when using the three node
triangular element and/or the four node quadrilateral element. In performing the analysis of transient problems, one or more ZIPP cards may be
required depending on the need to change time step (At) or printout frequency (iprint) during the analysis. When using a sequence of ZIPP cards, all infor
mation pertinent to the continued analysis of the problem must be specified on cards following the first ZIPP card. The control of the integration interval is accomplished by either of two methodsspecification of a final time ( t of time steps to be taken (no. steps). ^
f l
the specified number of time steps is equaled, the program checks for the pre sence of another ZIPP command and the definition of a new integration interval. This process continues until an END command is encountered, thus ending the integration process. When the initial condition parameter is set to TAPE, COYOTE expects the initial temperature field to be supplied from a disc (or tape) file, called TAPE 19. Details on the format for this file are given in a later section of
this chapter. The choice of an appropriate time step for a transient problem is often a critical point in the analysis for reasons of computational economy and accuracy. A procedure for estimating a meaningful integration time step, based on problem boundary conditions and the element mesh, is outlined in Appendix B.
58
To aid in interpreting and using the computed temperature solution, COYOTE allows the computation of several heat flux quantities on an element basis. computation of heat fluxes is initiated by the following command card, HEATFLUX, time step no., location where, time step no.: is the number of the time step for which heat flux com For steady state problems, this parameter is The
every element at every time step. location: specifies where the heat flux calculations are to be made.
For location = FULL, heat flux calculations are made for every element in the mesh. A second option allows fluxes to be calculated in up to This latter option is speci
fied by listing the required element numbers after the "time step no." parameter. This parameter is omitted if time step no. = ALLTIMES.
Within a given element, heat flux values are calculated on the element boundary midway between nodes. Calculations are made for the x and y (or r and z) com
ponents of the flux vector, the heat flux normal to the element boundary, and the heat flux integrated over each side of the element (i.e., total energy transferred across the element boundary). When using the HEATFLUX command in
conjunction with a transient analysis, note that the time steps are numbered continuously from 1 to n beginning with the initial conditions (i.e., solution at time At is step number 2 ) . Heat flux calculations can thus be made at any
particular time step by appropriately setting the "time step no." parameter.
50
Note that the use of the 'time step no. = ALLTIMES" option does not result in any heat flux data being printed by COYOTE. This option should be used only to
create the data file for the subsequent plotting of flux histories (see HISTORY option under the PLOT command). There is no limit to the number of HEATFLUX
60
The automatic coarse to fine mesh rezoning of a coarse grid solution is initiated through the command card, REZONE, [xmin, ymin, xmax, ymaxl , ni, n j , fiprintl, [imin, jmin, imax, jmaxl where, xmin, ymin, xmax, ymax: to be rezoned. ni, n j : specify the x, y coordinate limits on the region
specify the number of fine grid elements to be placed in each ni and nj must be less
coarse grid element in the I and J directions, than 10. iprint<3>: cedure,
specifies the amount of output produced by the rezone pro iprint may vary from 1 (limited output) to 4 (full output). specify the limits of the region to be rezoned Only elements entirely within the
in terms of I, J coordinates.
region bounded by these limits are rezoned. The region to be rezoned may be specified in terms of either the physical coor dinates of the mesh (imax, jmax, etc.) or the I, J coordinates of the mesh (imin, jmin, etc.). Following a REZONE command, the program should be directed The REZONE command only
provides mesh refinement and interpolation of new boundary conditions and not automatic re-solution. The present version of COYOTE only allows time indepen
61
The COYOTE program contains a plotting package to facilitate the interpre tation of data obtained from a solution and to aid in setting up element meshes. There are six basic types of plots which are available in COYOTE and are ob tained with the following corm.iand card, PLOT, plot type, xmin, ymin, xmax, ymax,
fimin^ .jmin, imax, .jmax] , [xscale], [yscale], [number 1 , [special ptsl where, plot type: desired. is an alphanumeric name which specifies the type of plot The permissible parameter values are catalogued below. specify the range of coordinates for the area
to be plotted in a line drawing or define the range of the ordinate and abscissa for a graph. For line drawings (e.g., element plots,
outline plots., contour plots) the xmin, ... ymax parameters define a rectangular window; only elements and their associated data that fall entirely within the window will be plotted. histories or profiles), the For graphs (e.g., time
maximum and minimum valuer for the ordinate (ymin, ymax) and abscissa (xmin, xmax) of the plot. If these parameters are omitted, the axes
for the graph are set by COYOTE. imin, jmin, imax, jmax: is an optional specification of the limits of If the I, J limits of the region are speci
fied, the xmin, ... ymax parameters are used to set the border for the plot. These parameters are omitted for graph plots.
62
xscale<1.0>, yscale<1.0>:
y coordinates of the plot. The default values produce a correctly proportioned plot of the largest possible size consistent with the plotting device. The use of a scale parameter produces a non-
proportional plot. The use of these parameters is illustrated in the example problems. number<>: These parameters are omitted for graph plots.
specifies if the element numbers are to be displayed on the If number is omitted, element numbers are not plotted;
element plot.
if number = NUMBER, the element numbers are plotted at the element centroid. special pts <>: specifies if the location of the special output points
(see OUTPUT command) are to be displayed on the element plot. If special pts is omitted, special point locations are not plotted; if special pts = SPECIAL, the special point numbers are plotted at the appropriate location on the element plot. The parameter on the PLOT command card, denoted "plot type" may be set to any of the following values as required. (a) POINTS: generates a plot of the grid points generated by the SETUP command. The "grid plot" parameter on the SETUP command card must be
set to PLOT to obtain this type of plot. (b) ELEMENTS: (c) CONTOUR: (d) OUTLINE: generates a plot of the element mesh. generates contour plots of the temperature. generates an outline plot of the problem domain with material
boundaries indicated. (e) HISTORY: flux. (f) PROFILE: generates plots of temperature versus position with time as generates time history plots of the temperature or heat
a parameter. Following the command cards for the CONTOUR, HISTORY, and PROFILE plot options, a series of data cards are required.
63
Contour Data Cards: For the CONTOUR plot option, the required data cards are of the following form: contour type, time step no., no. contours, [cl, c2, ... c20]
END where, contour type: contoured. time step no.: quired. is an alphanumeric name indicating the variable to be This parameter must have the value ISOTHERMS. is the number of the time step for which a plot is re
For steady state problems, this parameter may be omitted. specifies the number of contours to be plotted. A maximum
no. contours:
of twenty contour lines is allowed on each plot. cl, c2, ... c20: are optional values that specify the value of the con If these parameters are omitted, the plotted con
tour to be plotted.
tours are evenly spaced over the interval between the maximum and minimum values of the variable. Note that when the contour values are left unspecified, the maximum and minimum values used to compute the contour levels are those for the entire mesh. This
procedure may produce an unsatisfactory (or blank) plot in the event only a small portion of the mesh is plotted and the number of contours specified is small. plotted. This situation may be avoided by specifying the contour values to he When using the contour option with a transient analysis, note that
the time steps are numbered continuously from 1 to n beginning with the initial conditions (i.e., solution at time = At is step number 2 ) . A contour plot can thus be made at any particular time step by appropriately setting the "time step no." parameter. Any number and/or type of contour data cards may follow a CONTOUR command card; the sequence is terminated by an END can ..
1
64
series of contour plots for a transient analysis, a looping feature is available. The looping command has the following form, PLOTLOOP, no. plots, plot inc contour type, time step no., no. contours, [cl, c2, ... c20] PLOTEND
END where, no. plots: plot inc: specifies the number of plots to be generated, indicates the frequency at which a plot is generated, i.e.,
every "plot inc" time steps a contour plot is produced beginning with the "time step no." indicated on the contour data card. There is no limit to the number of PLOTLOOP data sets that may follow a CONTOUR command card; the sequence is terminated by an END card. PLOTLOOP, only a single contour data card may be defined. Within any g-/en PLOTLOOP's and indi
vidual contour data card may be mixed under a single CONTOUR command. History Data Cards: For the HISTORY plot option, the required data cards are of the following form, location, no. elem no., node no., elem no., node no.,
END
where, location: is an alphanumeric nane indicating the variable to be plotted For location TLOCATION, nodal point 65
as a function of time.
temperatures are plotted versus time; location = QLOCATION implies that heat fluxes (normal component) are to be plotted; location = TSPECIAL allows temperatures at the special output points (see OUTPUT command) to be plotted. no. points: specifies the number of histories to be plotted. A maximum
of ten time histories per plot is allowed. time step 1, time step 2: indicate the time step numbers at which the The maximum differ
ence allowed between time step 2 and time step 1 is 400, i.e., only 400 time steps may be represented on a given plot. elem no., node no.: are pairs of numbers indicating the element and node A maximum of ten such
number for which a time history is required. pairs per data card is allowed.
case, the numbers of the special points to be plotted are listed in sequence. A maximum of ten such numbers per data card is allowed.
There is no restriction on the number or type of LOCATION data cards that may follow a HISTORY command card; the sequence is terminated by an END card. numbering of the element nodal points is shown in Figure 11. Note that the The
numbering of fluxes in an element differs from the node numbering convention; fluxes are numbered sequentially (counterclockwise) around the element beginning with the point nearest the identifying node for the element. Also, the plotting
of flux histories assumes that the HEATFLUX, ALLTIMES command was executed prior to the command for the history plot. Profile Data Cards: For the PROFILE plot option, the required data cards are of the following form:
66
TIMEPLANE, no. planes, tl, t2, ... tlO location, delimiter type, no. pairs, elem no., side/node no., elem no., side/node no
END where, no. planes: plotted. specifies the number of profiles (at different times) to be A maximum of ten profiles per plot is allowed, specify the time step numbers at which the profile is For steady state problems, this data is omitted,
is an alphanumeric name indicating the variable to be plotted Location should be set to TLOCATION.
is an alphanumeric name indicating how the profile If delimiter type = SIDES, the profile For delimiter type =
geometry is to be specified.
NODES, the profile is given in terms of individual nodal points, no. pairs: specifies the number of data pairs (elem no., side/node no.) A maximum of 20 such data
needed to describe the profile geometry. pairs is permitted, elem no., side/node no.:
geometry in terms of an element number and a side or node number. There is no restriction on the number of TIMEPLANE/LOCATION data cards that may follow a PROFILE plot command. However, TIMEPLANE and LOCATION data cards must The data sequence for the PROFILE
occur in pairs with the TIMEPLANE card first. option is terminated by an END card.
With the delimiter type set to SIDES, the temperature profile is plotted along element sides with the spatial distance being measured along the element boundary. When delimiter type is set to NODES, the temperature profile is In
plotted o,' constructing the straight line path between successive nodes.
67
both options, the input sequence of elements determines the profile path.
There
is no requirement that the profile have a continuous path, i.e., some elements along a path may be omitted. dual element are permitted. Multiple side/node specifications for an indivi When the SIDE option is used, the profile path is
directed along element sides with the direction of increasing path length deter mined by increasing corner node numbers. Thus, for an eight-node quadrilateral
with side 1 specified, the path proceeds from nodes 1 to 5 to 2; with side 3 specified, the path is directed from nodes 3 to 7 to 4. If the path direction
68
The COYOTE program allows a computed solution and its associated problem data to be conveniently saved for further post-processi'.g through the use of a restart command. In order to save a previously computet) solution, the following
command card may be used, RESTART, SAVE In order to restart a previously saved solution, the following command card is used, RESTART, RESET, [nsteps] where, nsteps <0>: is a parameter to indicate at what time step number the If
solution file is to be positioned at during the restart process. nsteps is omitted, the solution file is rewound.
restarting of the solution process with the nsteps + 1 time step. This parameter has no meaning for steady state problems. The command to save a solution may occur at any point after the ZIPP command sequence, that is, after a solution has been obtained. In order to restart a
solution, the RESET command should immediately follow the header card. If the solution is to be continued in time following a restart, the FORMKF command must be executed prior to the ZIPP command as the restart procedure does not save the matrix equations for the problem. Also, the initial condition The solution
file should be re-positioned at the appropriate time step through use of the nsteps parameter.
69
The RESTART commands when executed from the data deck, direct the program to collect (or distribute) pertinent element and solution information onto (or from) two files, TAPE13 and TAPE19. To complete the restart process, these
files must be saved (or attached) by the appropriate system control cards. Typically, these files would be saved on a magnetic tape or catalogued on a permanent file. section. The restart procedure is illustrated in the example problems
70
If two or more
problems are to be run in sequence, then the appropriate termination for any particular problem is, END, fiprintl However, if the program is to be terminated with no further computational oper ations, then the following command card is used, STOP, fiprintl
where, iprint: is an optional alphanumeric parameter that allows the printing If iprint is omitted,
71
As noted previously, the order of the commands to COYOTE is dependent on the needs of the user. However, some limitations on the command sequence are
obvious as some operations are necessary prerequisites to other computations. The following comments provide some guidelines to specific sequencing situations. (a) The POINTS plot option must follow the SETUP command sequence as the file used to store the grid point coordinates is rewritten in later operations. Typically, a grid point plot is used in setting up a grid
and is not generated during a complete solution sequence. (b) The ELEMENTS and OUTLINE plot options can be located anywhere after the SETUP command sequence. (c) The remaining plot commands can occur at any point after the values to be plotted have been computed. (d) The OUTPUT command must occur after the SETUP command and must precede the ZIPP command, (e) The P.ESTART, SAVE command can only be executed after a solution has been obtained. No provisions are made in the restart process to
retain the element coefficient matrices and thus the SAVE option has little meaning prior to obtaining a solution. (f) If a solution is to be saved, the RESTART, SAVE command should gener ally be the last command (except for STOP) in a data deck. The restart
process uses several tape files that are also employed by other pro gram operations. The execution of commands following a RESTART, SAVE
command could result in a conflict in file usage. (g) If a transient solution has been saved, it may be continued in time by use of the RESTART, RESET command. The RESTART command should be
followed by the FORMKF command and the ZIPP commands with the TAPE parameter indicated. 72
There are a number of situations which require the user to supply FORTRAN coded subroutines to COYOTE; the definition of special nodal point locations, the use of variable material properties, the use of a variable volumetric heat source, the specification of a general radiation/convection boundary condition and the use of time dependent temperature or flux boundary conditions. When the parameter EXTDEF is encountered during the generation of nodal points, COYOTE accesses SUBROUTINE EXTDEF which is supplied by the user. This
subroutine allows nodal point locations to be defined in any appropriate manner by the user. This subroutine must have the following form: SUBROUTINE EXTDEF (X, Y, MAXI) DIMENSION X(MAXI, 1 ) , Y(MAXI, 1)
RETURN END where the variables in the subroutine parameter list are X, Y: two-dimensional arrays containing the x, y (or r, z) coordinate The indices for each two-dimensional
array correspond to the I, J name for the nodal point (i.e., X(I, J ) , Y(I, J) for point I, J ) . MAXI: an integer specifying the largest I value that can be defined in This parameter corresponds to the maxi parameter specified
the mesh.
73
When the "temperature dependence" parameter on the material data card is set equal to VARIABLE for any material in the problem, COYOTE expects SUBROUTINE CONDUCT and SUBROUTINE CAPACIT to be supplied by the user. problems, only SUBROUTINE CONDUCT need be supplied. For steady state
the user to conveniently specify the conductivity and heat capacity (actually p.C ) as arbitrary functions of the temperature. In the event only one of the
two indicated properties is to vary with temperature, the remaining property must still be set through use of the aDpropriate subroutine, i.e., COYOTE ex pects both subroutines to be present. Each subroutine is used to evaluate the
appropriate material property for all materials labeled as temperature dependent on the material data cards. The required subroutines must have the following forms: SUBROUTINE CAPACIT (RHOCP, T, X, Y, NNODES, MAT, NELEM, TIME) DIMENSION RHOCP (1), T(1), X(l), Y(l)
FORTRAN coding to evaluate p-Cp for each material with temperature dependent capacitance
RETURN END and, SUBROUTINE CONDUCT (CONDI, C0ND2, T, X, Y, NNODES, MAT, NELEM, TIME) DIMENSION CONDl(l), C0ND2(1), T(1), X ( l ) , Y(l)
FORTRAN coding to evaluate k n and k22 for each material with temperature dependent conductivity
RETURN END
74 ri
where the variables in the subroutine parameter lists are: RHOCP (NNODES): an array containing the values of p-C . Arrays containing the values of the
F o r
principle conductivities k,, and koo' C0ND2 (NNODES) = COND 1 (NNODES). T (NNODES): nodes. X (NNODES), Y (NNODES):
s o 1 ; r o
Pi
materials, set
r, z) coordinates for the nodes. NNODES: an integer parameter specifying the number of nodes at which
the material property is to be evaluated. MAT: an integer specifying the material number as set on the material
data card. NELEM: TIME: an integer specifying the element number. the current value of the time.
When the volumetric heating parameter, Q, on the material data card is set equal to VARIABLE, COYOTE expects SUBROUTINE SOURCE to be supplied by the user. This subroutine must have the following form: SUBROUTINE SOURCE (OVALUE, T, X, Y, NNODES, MAT, NELEM, TIME) DIMENSION QVALUE(l), T(l), X(l), Y(l)
FORTRAN coding to evaluate the volumetric heat source for each material with a variable heat source
RETURN END where the variables in the subroutine parameter list are: QVALUE (NNODES): heating rate. an array containing the values of the volumetric
75
T (NNODES): nodes.
X (NNODES), Y (NNODES):
r, z) coordinates for the nodes. NNODES: an integer parameter specifying the number of nodes at which
the volumetric heating is to be evaluated. MAT: an integer specifying the material number as set on the material
data card. NELEM: TIME: an integer specifying the element number. the current value of the time.
The use of a generalized radiation/convection boundary condition requires that the variatioi: of the heat transfer coefficient with temperature and/or time be specified. When the "h" parameter on the SET data card is specified
as VARIABLE, the COYOTE program expects SUBROUTINE HTCOEF to be supplied by the user. The required subroutine has the following form: SUBROUTINE HTCOEF (HT, TSURF, TREF, XSURF, YSURF, TIME, ISET, NELEM)
FORTRAN coding to evaluate the heat transfer coefficient for each set number
RETURN END where the variables in the subroutine parameter list are: HT: the heat transfer coefficient. the local surface temperature of the material. the reference temperature of the environment as specified on the
TSURF: TREF:
SET data card. XSURF, YSURF: TIME: the x, y (or r, z) coordinates on the surface.
76
ISET:
an integer specifying the "set no." as set on the BC and SET data
Note that if more than one generalized radiation/convection boundary condition occurs in a problem, SUBROUTINE HTCOEF is used for the evaluation of all heat transfer coefficients that vary with time and/or temperature. The set no.
(ISET) parameter is used to distinguish the different boundary conditions. When the boundary condition options TVARY and/or OVARY are employed, COYOTE expects the appropriate time history subroutines to be supplied by the user. The correspondence between a particular boundary condition and its variation with time is established through the "time curve no." which appears on the BC data card and in the name of the subroutine providing the time history. required subroutines have the following form: The
RETURN END where the variables are: n: an integer specifying the "time curve no." as set on the BC data card (1 ; n i 6). NELEM: TSURF: TIME: an integer specifying the element number. the local surface temperature of the material. the current value of the time. the value of the boundary temperature or heat flux (as required)
BCVALUE:
77
The subroutines described above, when required by COYOTE, should follow the main overlay in the loading sequence. The control cards necessary to
78
Initial Conditions
The analysis of a transient conduction problem requires the specification of a set of initial conditions for the problem. For cases where the initial
temperature field may be assumed uniform (or at least constant over each material), the initial conditions may be set through a parameter on the material data card. For problems where the initial temperature field is more complex,
COYOTE accepts the initial conditions from an external storage device (disc or magnetic tape file) denoted TAPE19. In order to be compatible with COYOTE, the initial temperature field must be written to unit 19 with the following unformatted FORTRAN write statement, V7RITE(19) TIME, TMAX, TMIN, NUMEL, ((T(J,I), I = 1,8), J = 1, NUMEL) where, TIME: is the initial time. are the maximum and minimum temperatures in the field.
requires that NUMEL 500, the Cray version requires NUMEL <. 1000). T(J,I): is the array containing the initial temperatures.
Note that the I index has an upper limit of 8 which corresponds to the number of nodes in a quadrilateral element. If triangular elements are present in the
mesh, the I index must still run to 8; the last two entries in the T array are ignored by COYOTE when a triangular element is encountered. Ordering of the
nodal point temperatures for each element must be as shown in Figure 11. Initial estimates for the temperature are also necessary for steady state problems when radiation boundary conditions are used.
79
When the file containing the initial conditions is attached to the job, the file name must be TAPE19. It should be noted that the output format for
COYOTE is the same as the above; the output file for COYOTE is also TAPE19. Thus, solution fields from COYOTE may he used directly as initial conditions for subsequent problems. chapter. The use of these features are demonstrated in the next
80
Error Messages
COYOTE has been supplied with a number of error checks and tests for bad or inconsistent input data, overflow of storage, etc. When an error is encoun
tered, a message is printed indicating in what subroutine the error occurred, the type of error, and any pertinent data associated with the error; the program is terminated with a STOP 1 if the error is fatal. The error messages imbedded
in COYOTE are listed below according to overlay along with brief explanations of the error. (0,0) OVERLAY DRIVER-UNRECOGNIZED COMMAND: a command instruction was expected but was
not found or was misspelled. FFLD-END OF DATA: an end of file was encountered on the input file, check
termination command. FFLDSB-INPUT VARIABLE TOO LONG: characters was encountered. RESTART-UNRECOGNIZED COMMAND: a restart command was used with an incom an input variable with more than ten
plete or misspelled parameter list. PRINTER-UNRECOGNIZED COMMAND: an output command was used with sn incom
plete or misspelled parameter list. (1,0) OVERLAY ELDATA-UNRECOGNIZED COMMAND: erroneous element, boundary condition, or
looping specification, check spelling. ELDATA-LOOP PREVIOUSLY DEFINED: error in looping specification, check for
third loop within two existing loops. ELDATA-MAXIMUM NUMBER OF ELEMENTS EXCEEDED: used or re-dimension code. reduce number of elements
81
incomplete or misspelled parameter list. ELDATA-BC APPLIED TO AN UNDEFINED ELEMENT: a boundary condition was applied
to an element that has not yet been defined. ELDATA-EXCESSIVE BOUNDARY CONDITIONS ON ELEMENT: too many boundary condi
tions (i.e., >6) have been applied to an element. ELDATA-UNRECOGNIZED BOUNDARY CONDITION: error, check spelling. ELDATA-BC ON IMPROPER SIDE OP ELEMENT: a boundary condition was specified boundary condition type is in
for an improper side of an element (1 <. side ^ 4 ) . NMESH-IJ MAX OR MIN EXCEEDS SPECIFIED VALUE: during generation of the grid
points, a part was found with an imax, jmax, imin, or jmin that ex ceeded the specified maxi on the SETUP command. MATREAD-TOO MANY MATERIALS SPECIFIED: fied. (2,0) and (10,0) OVERLAY FORMKFP-BAD ELEMENT JACOBIAN: a negative element area was found, check more than ten materials were speci
element coordinates and connectivity. TRI-ZERO JACOBIAN: a triangular element with a zero area was found, check
element coordinates. QUAD-ZERO JACOBIAN: a quadrilateral element with a zero area was found,
check element coordinates. (3,0) OVERLAY ZIPP-UNRECOGNJZED COMMAND: a solution command was used with an incomplete
or misspelled parameter list. ZIPP-MAXIMUM STORAGE EXCEEDED: maximum active storage exceeded, reduce
problem size or re-dimension code. ZIPP-COMMAND CARD MISSING: another solution command was expected but not
82
ZIPP-ZERO PIVOT: zero on the diagonal of the equation being eliminated, equation system is singular or element connectivity is in error. BCTIME-TIME CURVE NUMBER TOO LARGE: a time curve number greater than six
was encountered, check data cards and time history subroutine names. PREFRNT-INSUFFICIENT STORAGE: insufficient storage for element connecti
vity, reduce problem size or re-dimension code. PREFRNT-INSUFFICIENT STORAGE FOR FRONT WIDTH: dimension code. (4,0) OVERLAY REZ0NE-TO0 MANY SUBDIVISIONS OF COARSE ELEMENT: more than ten fine grid elements per coarse grid element have been specified. REZ0NE-T00 MANY COARSE GRID ELEMENTS IN REZONE: more than one hundred coarse grid elements have been found in the rezone region. REZONE-EMPTY REZONE REGION: rezone region. (6,0) OVERLAY PLOTZ-UNRECOGNIZED COMMAND: a plot command was used with an incomplete or no coarse grid elements were found in the reduce problem size or re-
spelling on contour type, looping command, and termination. TIMEPLT-UNRECOGNIZED COMMAND: spelling and termination. SPLOT-UNRECOGNIZED COMMAND: error in profile plot specification, check error in history plot specification, check
83
COYOTE is a fairly large code (approximately 6000 source statements) that operates most effectively on a large computer system. The code was originally
developed on a CDC 6600 computer with an extended core storage (ECS) capability. The present version of COYOTE is designed primarily for execution on the CDC CYBER 76 computer system and the Cray SI computer. The following discussion of
computer request parameters and control cards is specifically directed toward the use of COYOTE on the Sandia National Laboratories computer system. The central processor time needed to run COYOTE is directly dependent on the number of elements in the mesh and to a lesser extent on the front width of the problem. To roughly estimate the CPU time required for a typical job, the
following formula may be used for the CDC version of COYOTE: CPU = (Number elements) x (0.02) + (Number elements) x (Number time steps or iterations) x (0.007) .
Note that the constants in the above formula are accurate for problems of moder ate size but could increase by a factor of two for very large analyses. Also,
the above estimate of CPU time does not include the time for post-processing the solution, i.e., computing heat fluxes, plotting, etc., nor does it include the time charged for unoverlapped I/O operations. In terms of CPU time, the
Cray version of COYOTE executes approximately twice as fast as the CDC version. The amount of Large Core Memory (LCM) needed for COYOTE (CDC) is also directly related to the number of elements in the mesh and the problem front width. Unfortunately, the LCM required cannot be estimated a priori. The
experience of the user is, thus, the best guide for setting this parameter. After completion of a solution, COYOTE reports the total LCM required for the
84
job allowing the user to adjust the requested value for future runs. request is needed for COYOTE (Cray).
No LCM
The COYOTE program is maintained in an UPDATE form in the Sandia National Laboratories permanent file system and may be accessed by executing the follow ing control cards: PFGET, FLNAME, COYOTE, AU = DKGARTL. and, FILE, FLNAME, RT = S. PFGET, FLNAME, COYOTEC, AU = DKGARTL. (CDC Version)
(Cray Version)
Before executing, the file attached with the above control cards must be run through the UPDATE processor and then compiled. In the following sections,
control card decks are listed for using COYOTE in its standard modes of operation. Run with Plotting: The control card sequence shown in Figure 12a allows the user to run COYOTE (CDC) and plot results using the RSCORS software package. The graphical
output is post-processed and returned through a Versatec plotter at the appro priate Remote Job Entry Terminal (RJET). Other options for processing and dis To access the Cray
version of COYOTE, the control stream in Figure 12b should be executed. Run with Restart: Figure 13a shows a control card stream to run COYOTE (CDC) and save the solution for later restarting. In this case, the solution and element data were The
saved on magnetic tapes; permanent file storage could also have been used.
listing in Figure 13a is also suitable for restarting a job if the STAGE commands are modified to cause files TAPE13 and TAPE19 to be PRE staged. The control
cards needed to run COYOTE (Cray) with restarting are shown in Figure 13b.
85
Run with Initial Conditions: The control stream shown in Figure 13a are also suitable for running a transient analysis with initial conditions provided from an external source. If the initial conditions are written on TAPE19 (in the format discussed previously), then the listing in Figure 13a is immediately applicable if the request for TAPE13 is deleted. The control stream for the COYOTE (Cray) version
of this procedure is shown in Figure 13b. Run with User Subroutines: The control cards required tc execute COYOTE (CDC) when user supplied sub routines are to be included are shown in Figure 14a; the equivalent COYOTE (Cray) control stream is shown in Figure 14b. In Figures 12b, 13b, and 14b, the execution of COYOTE (Cray) is carried out through use of a set of procedure files (SLTLIB 21 ). The use of these procedures
simplifies the access to COYOTE on the Cray though in fact, the basic Cray Job Control Language can also be used to run the program.
86
COYOTE, TXXXX, ECYYY. USER PFGET, FLNAME, COYOTE, AU=DKGARTL. UPDATE, P=FLNAME, N, F. FTN, I=COMPILE, L=0. * PFGET, RSCOR76, AU=SLTHOHP. * PFGET, RSCDI76, AU=SLTHOMP. * LIBRARY, RSCOR76, RSCDI76. RFL, L=YYY. LGO, PL=777777B. * PFGET, P0P76, AU=SLTHOMP. * POP76, TAPE10, rOPOUT, IIC1. * COMQ, POPOUT, HC1. 7-8-9 7-8-9 COYOTE DATA 6-7-8-9
NAME
BOX NO.
NOTES: (1) For large jobs producing large quantities of output, consideration should be given to using a F1CHE, OUTPUT command to divert the printed output to microfiche. (2) Graphical output may also be generated via the SCORS software system by appropriate substitution for the lines marked by * (see Reference 21). The above control statements assume plots will be generated on
a Versatec hardeopy plotter; other options are available (see Reference 20)
FIGURE 12a
87
NAME
BOX NO.
CUPDATE, P=COYOTEC, UN=DKGARTL, F. CFT, I=COMPILE, L=0. * RSCORLB. * ASSIGN, DN=POPIN, A=FT10. * LDR, LIB=RSCORS. * POP, POPIN, POPOUT, HC1. * XCOMO . PriPnilT. HC1 7-8-9 7-8-9 COYOTE DATA 6-7-8-9 CS=R7 .
NOTES: (1) For large jobs producing large quantities of output, consideration should be given to using a XFICHE, $OUT command to divert the printed output to microfiche.
FIGURE 12b
88
COYOTE, TXXXX, ECYYY. USER PFGET, FLNAME, COYOTE, AU=DKGARTL. UPDATE, P=FLNAME, N, F. FTN, I=COMPILE, L=0. LABEL, TAPE13, W, L=$$. LABEL, TAPE19, W, L=$$. STAGE, TAPE13, VSN^AAAAAA, HY, POST. STAGE, TAPE19, VSN=BBBBBB, HY, POST. RFL, L=YYY. LGO, PL=777777B. 7-8-9 7-8-9 COYOTE DATA 6-7-8-9
NAME
BOX NO.
NOTES: (1) The above listing uses (skeleton labeled) magnetic tapes to store the element data and solution fields. The permanent file system could also be used to save the data by replacing the STAGE and LABEL commands with the appropriate PFSAVE commands. Note that PFSAVE commands should follow the LGO card. (2) The above listing is suitable for restarting a COYOTE run from tape. When restarting, the POST parameter on both STAGE commands should be changed to PRE. When restarting from permanent file storage, the PFSAVE commands should be replaced by an appropriate set of PFGET commands. (3) The inclusion of plotting in the above run is accomplished through the addition of the appropriate commands from the control stream shown in Figure 12a (i.e., those with an * ) . FIGURE 13a 89
NAME
30X NO.
CUPDATE, P=COYOTEC, UN=DKGARTL, F. CFT, I=COMPILE, L=0. LDR. XTAPEOT, FT13. L=($$), VSN=AAAAAA. XTAPEOT, FT19, I , = ($$), VSN=BBBBBB. 7-8-9 7-8-9 COYOTE DATA
NOTES: (1) The above listing uses (.skeleton labeled) magnetic tapes to s .ore the element data and solution fields. The permanent file system
could also be used to save the data by replacing the XTAPEOT commands with the appropriate XPFSAVB'. commands. (2) The above listing Is suitable for restarting a COYOTE run from tape. When restarting, the XTAPEOT commands are replaced by
XTAPEIN commands; the XTAPEIN commands precede the LDR command. When restarting from permanent file storage, the XPFSAVE commands are replaced by XPFGET commands. (3) The inclusion of plotting in the above run is accomplished through the addition of the appropriate commands from the control stream shown in Figure 12b (i.e., those with an * ) .
FIGURE 13b
90
COYOTE, TXXXX, ECYYY. USER PFGET, FLNAME, COYOTE, AU=DKGARTL. UPDATE, P=FLNAME, N, F. FTN, I=COMPILE, L=0. RFL, L=YYY. LGO, PL=777777B. 7-8-9 * IDENT, USERSUB * DELETE, USER.2, USER.38 FORTRAN CODED USER SUBROUTINES 7-8-9 COYOTE DATA 6-7-8-9
NAME
BOX NO.
NOTES: (1) The inclusion of plotting in the above run is accomplished through the addition of the appropriate commands from the control stream shown in Figure 12a (i.e., those with an * ) .
FIGURE 14a
91
NAME
BOX NO.
('UPDATE. P=COYOTEC. UX=DKGART1.. F. CFT. I=C'OMPII.F.. I . = 0. !.DR. 7-8-'.) I DENT. I'SKRSUU DELETE, USER. 2. USER.38 FORTRAN' CODED USER SUBROUTINES 7-8-D COYOTE DATA 6-7-8-9
NOTES: (1) The inclusion of plotting in the above run is accomplished through the addition of the appropriate commands from the control stream shown in Figure 12b (i.e., those with an * ) .
FIGURE 14b
92
V'XAMPl 1 PVMW M 1
\ s e r i e s (if sis proh lems h;ivi- IK-I n in- ! ,.':>' in ' ) ii^ i!)r u s e of of the p r e v i o u s l y described ennr and
-e-.iior
! > t : . - e .-t ; . , .
; ;
i he C O Y O T E c o d e .
Though
the e x a m p l e s
I t a: ; , : . -
H o w e v e r . a can-ful
I o sueoess ! , ;i ! \
p o i n t s about
the
relationship
between
15 is to be u r i dded u s i n c in Figure
a \\
listing shown
location of
the nodes by
remained
the n o d e s , as g e n e r a t e d
the P I O T , P O I N T S
ei>nnand.
Figure | 7 . The Note that input deck for Case A resulted in i he- element e l e m e n t s . only m e s h shown t he fi rst in F i g u r e ] ,s .
;i
in generali ng t ho e i g h t - n o d e
I. i tor t he
values
dt^-k
H. produe--d
in F i g u r e
In this c a s e , all
the r e r m r 1
. 1 values were
for e l e m e n t s
in the left-hand s i d e of
i be m i d - s i d e
g r i d s A and H i l l u s t r a t e
g r o u p of nodal p o i n t s p r o d u c e d by
of an e l e m e n t .
93
? . : , 4
. 0
; . 0,4 .0 1 . ' , 9
- . 9
10.0,
^
Part 1
1",
Pai t
0.0,0.0 1=1,J=l
5.0,0.0 13,1
The input deck for Case C i 1 lusiral e . s the use of a discontinuous I. . ! num bering scheme between parts cf the grid. Observe that in the second pari oi the
e n d generation (Figure 16c), the I numbering begins with 23 instral of 1 : i as in Cases A and fi. The nodes t.iat are common to the two parts (i.e.. share th.same physical location, denoted by X's in Figure 17) thus ba\c two ! eg i t I ma ! 1. J names (e.g.. (!3.!i = (23.1)). !n the construction of elements that us,-
these doubly named nodes, care must be taken in consistently use only . T . t ol i ; , , names. This is shown in input deck I " where the 1, J names from grid n a n 1 havi
Th.e final case (D) demonstrates the appropriate formats for the element cards when using four-node elements. In the first part of the grid general urn
(Figure 16a,, only the first I, , J for the element was specified with the program assuming default values for the remaining corner l, J quantities; the I and , t
94
DIRECT SEXAM'Lr
LIST
tf
ir.oili
3AT
P10bL
S T . : P , J ,:=i
,PRJ:JCPI3;:',=L3T
i s-: r JP : : -iti'o s: JJ
* JIID DIA-*T 1
:. ,o...,.
13,1 ,19.5 5 . f 1 C. 1 0 . , 5 . CC*,2>, 2 11 NO 12 I L C O P . S . 2 ! 3 JL0CP,4,2 I ' Q'JADbM, 1 , 1 , 1 15 J L N D 16 I-:NO > 7 ILC0P,3,2 15 J L D O P , 2 , 2 19 Q U A D S ' * , 1 , 1 3 , 1 20 J L N D 21 IEND 22 :ND
31 f l T . ^ T f f
L1Sr
D4TA
24 PLOT,:LM-:MTS,O. 25 I N D . N T a i H T
,o.,u.,.
S M3 OF 3 E T J P t * s 3 t L * 2 I " 7 S L 1 E * T PLOT
S3U>CE Lir
'!csh A 26 27 29 2? 30 31 32 33 3% 33 36 37 33 39 43 1 2 43 SEKAM'LE P*0BLH ONHSH STUP,2 1 9 , P S : ? C ; l B : 3 , ? L 0 T MATRIL,1,1.0,1.C,1.3 END !ilil!i9 3,.5.,5. ,3. Q.,3.,4.,4. 13,1.1?,5 a.1J.10.,3. a.,o.,2. ,2. END ILOO,3,4 JLOOP,4,2 QUAOS/4,1,1.1,5,1,5,3.1,3 JCNO ITND IL00P,5,2 JLOOP,,2
I t
S >IU
iL1Nr
OATA
Mi-sli I)
FIOL'RE
16.
95
51 52 53 5* 55
tEHAMPLE PROBLEM O N E ' - H ^ S H iENEIRATION C SETUP.2,29,PRESCRIBED,PLOT S SETJP :3N1MO S l f l J E M C : MATERIAL,1.1.0.1.0.1.0 * 1 A T i I L 3ATA ENG 1.1.13,5 * OATA-PART 1
3 R I : )
56 57 53 59 60
52
S 'AU 2
61 END
* EL:IENT DATA
63 64 65 66 67 63 6? 70 71 72 73
7*
JLOOP.4.2 9UA08/4.1.1.1 JEND IENO QUA0f'4l,13,l,2 5,l,25.3,13.32 4,l, 25.2.24.5.13,2 QUAD8/4.1,13,3.2 5.3.25,5,13.5.2*.3. 25.42,i.13.4 ILOOP.2.2 JLOOP.2.2 QUA06/4.1.25,1 JEND IENO
FND t -:NJ OF
a
SETUP
SEQUENCE
75 P L O T , J O I N T S . 0 . . 0 - . 1 0 . . 4 . 76 P L O T , : L E M E ^ T S , a . . 0 . . 1 3 . 4 .
77 ENO,N}INT
(c) Mesh C
81 END 32 1 , 1 . 1 3 . 9
53 0 . . 5 . . 5 . . 3 84 G . , 0 . , 4 . . 4 .
35 36 87 89 9) 91 92 93 94 95 96 97 93 99
100
ioi
3
13,1,19,5 5..10..1U..5. 0..0..2..2. ILOOP.12,1 JLOOP. 3 , 1 QUAD4/4,1,1,1 JEND IENO ILOOP,3,2 JLOOP,4,1 QUAD4^4,1,13.1.15.1,15,2,13,2 JEND IENO END
PLOT,?OINTS,?.,0..15.,4.
LOT,:LEWE>IT3,O. , o . , i a . , 4 .
98 END
t N03AL
t
3INT
PLOT
PLOT
IL-'IENT
1C2
STOP.mPRINT
Mesh D
96
1,9
i i i i i i i i i i i
13,9
i
Part 2
1,1
19,1
sliow.s how any suitable nodal point combination can be used to define an element. In this case, all four corner node I, J values are specified with the I and J loops being i ncrer-r-ntfid by two and one, respectively. deck U is shown in Figure 18c The mesh generated using
(a)
Mesh
A a n d C.
Cb)
Mesh B
(c)
Mesh D
FIGURE 18.
98
Heat Conduction . i n a Steel Bar The second example consists of a steel bar (square cross section) with a circular hole that is subjected to prescribed temperature boundary conditions. Til' two-dimensional idealization of the problem is shown in Figure 19. Due to
symmetry considerations, only one-eighth of the cross section needed to be con sidered in the finite element model. A COYOTE data deck used to solve the
steady state version of this problem is shown in Figure 20a, As indicated by the input listing, the grid points for this problem were generated in a single part; the curved part of the boundary was produced by use of a iiuartratic interpolation. The finite element mesh is shown in Figure 21.
The input listing in Figure 20a also indicates that a steady state solution was saved by a RESTART command to allow additional processing at a later time. control cards needed to complete D I P restart procedure are described in a previous section. y The
Steel
T = 100
FIGURE 19.
99
DIRECT L I S T 1 2 3 4
OF INPUT 3ATA
ON3U:TI0N IN A STEEL BAR
t t
5
6 7 3 9 13 11 12 13 l 15 IS 17 1} 11 20 21 22 23 24 25 26 27 ZA 29 30 31
1,1,25,21
3 . . 4 . 3 , , 3 5 3 , 0 . , , > 1^1 0 . , 0 . , 3 . 6 * 5 , 3 . 5 , , , 3 . 539 END ILOOP,12,2 JLOOP,10,2 QUADS/?,1,1,1 JEND IENO ILOOP.12,2 8 C T S 1 D E , 1 , 1 , 1 , ICC. 3 C , T S I D E , 1 , 1 9 , 3 , 30 C. IEND JL00P,t!),2 ^C,QSIDE,1,1,4,0 JEND END FORMKF ZIPP.STEADT END PLOT,ELEMENTS,0. , 0 . , 4 . , 4 . PLOT, ELEMENTS, 9 . , 0 . , 4 , , 4 . , , , , , , PLOT,CONTOUR,0. . 0 . , 4 . , 4 . ISOTHERMS,1,10 END ESTAU,5AVE STOP
J SSI 3 DATA
* ELEMENT
DATA
S3JUOAR*
CONDITION DATA
$ EN3 OF SETJP SE3UENCE S - 0 * 1 MATRIX EQUATIONS I STEADY STATE SOLUTION * ,i\JMZR t 'LOT RESULTS ELEMENT PLOTS
S SAVE
SOLUUON
(ai
Sleudy S t a t e
Solution
ti 1 2 3 4 5 5
DIRECT
LIST
OF INPUT DATA
SEXAM'LE P*ORLEt TWO--HEAT CONDUCTION I N 5 T E E . BA* SESTA1T,*ESET * RECALL SOLUTION HEATF-JX,l,l,U,21,3l,41,5l t ; 3 1 U T E HEAT FLUXES HEATF.UX.1,61,71 , 8 1 , 9 1 , 1 0 1 , 111 H E A T F L J X , I , 1 0 , 2 0 i 3 0 , 4 0 , 5 0 , 6 3 , 7 0 , ) 0 , 9 0 , 1 0 0 , I t 3,123 STOP.NOPRINT
<h)
Restart
FIGL'HK 20.
Input L i s t i n g E x a m p l e
ProtUom 2
100
LINE 1 2 3 4 5 6 7 3
3
DIRECT LIST 3F INPUT DATA *EXAN>L PROBLEM T y O - - H E A T SETUP.2,25PRESCFIBED STEEL,I , 7 . 6 8 9 , . 4 6 , . 2 2 END 1.1,25,21 0 . , 4 . J , . 3 5 3 , 0 . > 19 1 0 . . C , 3 . 6 4 6 , 3 . 5 , ,,3.538 END
tL00Pti2,z
CON3U:TI0N
ELEHENT
DATA
ID 11 12 13 14 15 16 17 13 19 20 21 22 23 24 25 26 27 23 29 30 31 32 33 34 35 36
JLOOP,10,2 QUADS/B,1,1,1 JEND IEND ILOCP,12,2 tBOUVOAFtr C O N D I T I 3 N DATA BC,TSIDE,1,1,1,3C0. 3CTSIDE,1,19,3 300. IENO JLOOP.10,2 BCQSIDE.1,1,4,0. JEND END S EN3 OF SETUP SESUENCE FORNKF S F O i MATRIX EOUATIONS ZIPP.RANSl, , 2 , 0 . , 4 . ,.4,10,TAPE S T R * Y S I E M r SOLUTION ZIPP,TRANS1,,2<4.0,10.,.5,1J ZIPP.TRANS1.,210.,15.,1.0,5 END PLOT,:DNTOUR,0. , 3 . , 4 .,4. S L0T RESULTS PLOTL3DP,Z3,2 I S O T H I R M S . l , 9 , I 2 C . , 1 4 0 . , 1 6 : . , 1 3 0, , 2 0 3 . , 2 2 J . 2 4 0 . , 2 6 3 . , 2 3 J . PLOTEMO ENO PLOT,HISTORY * *I3TORr PLOTS TLOCATI ON,4,1,30,2,4,4,4,6,4,8,4 TLOCATI ON,4,1,30.112,3,114,3,116.3,118,3 END STOP,<OPRINT
(c)
Transient Solution
FIGURE 20.
Input ListingExample
Problem 2
101
The input listing in Figure 20b indicates the use of the restart capability. [i: this case, heat flux calculations were performed for a number of elements in r . h e previously computed solution. A second problem was analyzed to demonstrate the setting of initial condi tions from an external tape file and a transient solution procedure. In Figure
20c, an input listing is provided for a transient analysis of the steel bar geometry. For this case, the outside temperature of the bar was raised from The ZIPP
command card indicates that the initial temperature field was to be read from a file called TAPE19. Through the job control cards, the steady state solution The steady
state solution thus became the initial condition for a related transient analy sis. Figure 22 shows several of the contour plots of the isotherms obtained
FIGURE 21.
102
t = 0.0
103
Conduct, i on
To d e m o n s t r a t e
c o n d i t i o n s , a o n o - d i m o n s i una! is encased
was
I ay or of
1 ay or of material
hav i ng" n
T h e on tor s u r f a c e of convection.
by natural
The g e o m e t r y ,
properties and
heating; history
the heat
inar and
I . urbul on I natural
convect i on shown
to t h"
1 he input
the m a t e r i a l
data card
for M A T
1 i n d i c a t e s a v a r i a b l e heat
boundary
cone i v i on - F T < * a rrl i ndi ca to.s a v a r i a b l e heat analvsi'- o[ this problem was carried cards. using The
A transient
;
out using
intounTtion
nicrv;i]s
temperature
several plot
radial
is shown
the e l e m e n t m e s h
reference.
1 0 1
q = 0 HAT 1 - MAT 2
0)
g
3 0 CO
4J
/
r
\ \
\
1
i
MAT 3 Q(t).
q = hAT
1
J
.
16 Time
\ \ . \
20 24
.
28
12
~T
Section for Finite Element Model P r = 5.0 r = 6.0 1.5 MAT 1 MAT 2 MAT 3
S
3
2 274 x 1 0 ~
r = 50.0
h - .29(f)-
25
-q = 0
h = .19(AT)
.33
for
10
> Gr
Gr = p g B L A T / u
FIGURE 23.
LIN:
1 2 3 5 6 7 3 9 13 11 12 13 K IS lo 17 13 19 21 21 22 23 2* 25 26 27 23 29 "' X 32 33 3 35 36 37 38 39
DIRECT LIST IF
INPUT DATA
S E X A I ' L E PROBLEM THREE--ONE-DIHE1I S I ONAL C 0 N J J C r i 3 N I N A CYLINDER SETUP.2,61 * SETUP C3HMAND SEQUENCE M A T 1 , 1 , 2 . 2 7 E - 3 , . 2 , . 0 1 , , , . VA*[BLE,20. SHATERIAL DATA < 1 A T 2 , 2 , 1 . 0 3 1 E - 3 t . l 2 2 , . 0 : * i t i t i 2 5. HAT3,3,2.l62E-5,.2 44,.0,,.20. END 1,1,11,3 * S ^ I J 3ATA-PART 1 0.,5.,5.,D. Q.,0.,1.,1. 11,1,13,3 J 'Ail 2 5.,6.,6.,5. 0.,C.,1. ,1. 13,1,51,3 I A*T 3 6.,50.,50. ,6. a.,0.,1.,1. END ILOOP,5,2 * ELEMENT DATA QUA08/,1,X IEND QUAOS/4,2,11,1 ILOOP,24,2 9UAD8/4,3,13,1 IENO SETQiUD,l .VARIABLE,20. * lajHOlRt CONDITION DATA 3CQRAJ, 5 3 , 1 , 2 , 1 END * END SETUP SEQUENCE FORMKF.AXISITH t FO MATRIX EQUATIONS ZIPP,nANSl,,20 ,2.0,.10,23 t TRANSIENT 33LUTTON ZIPP.rRANSl,,5,2.0,11.0,.25,36 Zr P P . T R A N S , , 5 , 1 1 . 0 , 1 3 . 0 , . 1 , 2 0 ZIPP,HANS1,,5.13.,30.,.2S,S8 END P L O T , E L E M E N T S , 3 . , 0 . , 5 0 . , I . , , , , , 1 0 . A . O J ELEMENT PLOT HEATFLUX.ALLTHES S : a l U T E HEAT FLUXES PLOTHIST01Y -USTORr PLOTS TLOCATI0N,*,l,157,1,8,5,6,6,6,30i6 9l.OCATION,2,l,150,5.3,63 END STOP
3
(a)
Input listing
FIGURE 24.
106
SUBROUTINE! 30URC: (OYA-U:: .T i X i Y i M M a O E S . H A T i N E L E M . T I M E ) 01 HENS ION O V A L U E d ) iT 1 > , X 11 > f f 1 I c C C SUBROUTINE TO EVALUATE T i V0LU1E 1EATINS
I F ( T I H E . G T . 2 4 . 60 TO 51 I F U I M E . G T . 1 3 . ) GO 0 0 I F C T I M E . G T . 1 2 . > S3 TO 50 I F ( T I M E . S T . 1 1 . 1 GO TO 2D I F I T I M E . 6 T . 1 . ) GO TO 1]
r
C DO 5 I =ltNNOOES <mLUE<I> = .123TINE CONTINUE RETURN CONTINUE DO 15 I = l i N N O D E S OVALUEd=.l23 COMTINUE RETURN CONTINUE DO 25 I = 1 . N N 0 D E S avALJE < i ) = . i 2 : . : 6 c ( T I K E - U . i CONTINUE RETURN CONTINUE DO 35 I = l . N N O J E S QVALUE(I)=.16: CONTINUE RETURN CONTINUE DO 45 T=l,NNOOES QVALUE(I> = . 1 8 r - . 1 3 ' ? < r I H E - 1 8 . ) / & . CONTINUE RETURN CONTINUE DO 55 I = l i N N O O E S OI/ALUE ( I ( = 1 . CJNTINUE RETURN END
5 10
15 20
25 30
35 40
*5 55
55
(b)
Source Subroutine
FIGURK 24.
107
SJ8RDUTINE C C
C
HTCDE-
3UHR0UTIK:
FREE
TO r n . u ^ E
;
( HT , X S J f>^ ,TREF , <SURF , YS J l F ,11 M" , 1 S^ T ,>JE " 1 ( " THE HEAT n N S = : R C O E F = r C I ENT FOR
A;S
CONVLCTI:'-<
^ N<INE
A
L11.5)/TFI.H*I93.>
C C C C C C 10
EVALUATE SRASHCF NUMBER 6 R = l R H O R H C 9 B E T A * A L H L A L O E L T ) ' f AiUAHJ EVALUATE H IF<G.LT.1.0E9> 13i25 CONTINUE LAMINAR FLOW HT=.2 9*CDELT/AL>.25 IF{GR.LT.1.0E> HT-0. HT=HT*CONVERT RETURN 2C CONTINUE TURBULENT FLOW H T = . 1 9 C D E L T . J 3 3 333> rir=HTCONi*ERT RETURN END
(;)
Ileal
Transfer
Coefficient
Subroutine
FIGURK 2-1.
Input
Listing--Fxample
Probjc-m
lot)
..
* - "
/ ** / (
V \
\
-"Point 1
f-*
~s
J
J
J Poin t 3 -
Poin 2
\ \
\ \ \\ \\ \\
V
^ ^i
Po nt 4
""N
*i
**<ai
aa>*
**Maa^
1
T I ME FIGURE 25. Time History PlotExample Problem 3
109
Hvai Conduction in a Finned Radiator The finned tube radiator section shown in Figure 26a was chosen to illus trate the use of time dependent boundary conditions. Symmetry considerations
allowed the finite element model in Figure 2Gb to be constructed for this probien;. The inside temperature of the aluminum tube was maintained at 100 F: the
outside surfaces of the radiator were subjected to the heat flux history shown in Figure 26b. The input data deck for the analysis of this problem is shown in Figure 27a. The use of the QVARY boundary condition requires that a time history sub
routine (corresponding to the time history number) be appended to the COYOTE program. For the present case, the flux history shown in Figure 26b was sup As a result of
the transient analysis of this problem, the contour plots in Figure 28 were produced, indicating the thermal response of the radiator.
FIGURE 26a.
110
<J
( ! ) .0025
.1
.2
.3 Time
.4
.5
T = 100
FIGURE 26b.
iIN:
ENO
b 1,1.11,11.,,,.POLAR t 5 R I 3 OATA-PART 1 6 , 75,.75,I.0,1.0 ,..75,,1.0 7 9O.,60. , 6 0 . , 9 0 . 75.75. 3 l i t t , 2 1 , ' t l t t t i t POLAR * U 2 9 .75,.75,1. J,1.3,.7e,,1.0 10 6 0 . , 3 3 , , 3 0 . , S 0 . , 4 5 . , , 4 5 . 11 2 1 . 1 , 3 1 , l l , , , , , P O L A a * Mir 3 12 . 7 5 , . 7 5 , 1 . 0 , 1 . J , . 7 5 , , 1 . : 13 3 0 . , C . , ! ! . , 3 0 . , 1 5 . , , 1 5 . 14 2 1 , 1 1 , 3 1 , 4 1 . 5 , , . 5 t PA)T 4 15 . 9 6 6 , 1 . 0 , 3 . 0 , 3 . 0 , . 9 6 5 ? 16 . 5 , 0 . , G . , . 7 5 , . 2 5 8 8 17 END 18 J L 0 0 P , 5 , 2 * ELEMENT DATA 19 I L 0 0 P , 1 5 , 2 20 aUADBf*,1,1,1 21 IENO 2 2 JENO 23 J L O O P . 1 5 , 2 24 I L O O P , 5 , 2 25 0 U A D 8 / B , 1 , 2 1 , 1 1 26 TEND 27 JEND 28 I L 0 0 P , 1 5 , 2 * 33UMDART CONDITION DATA 23 B C T S I O E , 1 , 1 , 1 , I C O . 30 IEND 31 I L O O P . 1 0 , 2 32 d C , Q V A R T , l , 9 , 3 , l 33 IEND 34 J L O O P , 1 5 , 2 35 B C f l V A R r , 2 1 , 1 1 , 4 , 1 36 JEND 37 ENO S END OF SETUP SEQUENCE 3B PLOT,ELEMENTS,-.CI,0.,3.3,1.01 S L3T ELEMENTS 3J LOT,ELEMENTS,-. 31,0. ,3. . ' . ,1.01, ,,,, ,,NUH3ER 4 0 PLOT,ELEMENTS,-.01 , 0 . , 1 . 0 , 1 . 0 1 41 FORMKF S F 0 * 1 MATRIX EdUATIONS 2 Z I P P , r R A N S l , , 5 , 0 . , . 5 , . 0 1 , 5 0 S T R I M S I E U SOLUTION 43 ENO 44 P L O T , ; 3 N T O U R , - . 0 1 , 0 . , 3 . : , 1 . 0 1 * :>LOT RESULTS 45 P L O T L O 0 P , 1 0 , 5 45 I S O T H E R M S , 6 , 1 2 47 PLDTEN3 43 END 45 PLOTHISTORT 53 TLOCATION,4,1,51 ,61,4,66 I 4,71,4,73,7 51 TLOCAr[0N,4,l,51,96,4,9i,7,146,4,14 8,7 52 END 53 STOP
(a)
Input Listing
FIGURE 27.
112
SUBR3UTINE C l H V E l
(b)
FIGURE 27.
Tl* . m i t
113
Heated Salt 131ock The penultimate problem has been selected to demonstrate the use of sub routines for I ernperature dependent properties and temperature dependent heat transfer coefficients. The physical problem consists of a right circular cylin The annu
der of bedded salt that is heated internally by a <-y I i ndri eal heater.
lar space between the heater and salt block and the cylindrical space above the heater is filled *vith crushed salt. The salt block is confined on the upper The problem geometry and finice The heater is assumed to be pro
and lower surfaces by circular aluminum plates. element mesh are shown in Figures 29a and 29b.
ducing heat at a constant volumetric rate; the cylinder is assumed to lose energy to the environment by natural convection through the top and vertical surfaces. The bottom surface is assumed insulated.
In the analysis of this problem, it was assumed that the thermal conducti vity of the bedded salt and crushed salt varied with temperature. and heater properties were assumed constant. The aluminum
coefficient as a function of temperature were also to be included in the analy sis. Figure 30a shows the input listing" for the steady state solution to this The material property subroutine and heat transfer coeffi
nonlinear problem.
cient subroutine that were appended to the COYOTE program for this problem are listed in Figure 30b. Note that SUBROUTINE CONDUCT is used to evaluate the
thermal conductivity for both bedded salt and crushed salt; SUBROUTINE HTC0F.F is used to evaluate the film coefficient for both top and vertical boundary conditions (based on ISET parameter). The results of this analysis are shown
114
. . q = hA
... ,..
Aluminum
1*1
,., ,.,
... ...
...
... ... ... ... ,., . ... ... , ... ,., ...
Itl
., ,
... ... ... ,..
tot
,
HI
...
Crushed Salt
l'
... !* ... . .
5 Salt q = * hAT
! ... ,..
II*
, ...
IH
'" ...
... ,., ...
,.. , , , . ,
...
... ...
... ...
-Heater (Q = Constant)
" 41
"
11 (f
10
to
"
It
IT
t*
10
f<
1*
'
"
"
b)
FIGURE 29.
Salt Block
1 SEXAH'LE
2
PnOBLE.1 F l V E - - H E A T r D
5A.T
BLOCK
* SLTJP :JHIN3 S:QUENC-:
SETUP,2,23,PRESCRIBED
, l
3 4 5 i 7 d 9 10 11 1? 13 14 15 IS 17 13 13
ALUMIVUK,1,2.71 , 2 . 3 5 E - 4 , 2 . a 4 2 , , , : 0 < $ T A N T , . 4 ) i t HATEUAL SALT,2 2.1S2.56E-4,.0 3 5i.,, l/A U*3LE.,40. HEATC. 3 3 . 1 9 , l . 2 8 E - 4 , . 4 5 0 , t ,C3 M3 TA NT t . U S 3 t 0 . C * U S M - S A L T . 4 t l . 5 7 . 2 . 5 S E - 4 , , 3 C 35 11 , V AR I A B L E , . * 0 . END 1,1,5,37 * 5 " I 3 OtTA-PART 1 C.,5.,5.,3. 0.,u.30.f33. 5,1,7,37 * AU 2 5.,6..6.,5. C.,C.30.,n. 7,1,23,37 S ?A*T 3 6.,50. ,50. ,6. Q.,G.t3Ci.tS0. END JLOOP,2,2 S E L ^ E N T 3ATA IL00P,11,2
1 l
OATA
20 21 22 23 24 25 26 27
23 29 33
31 32 33 3t 35 36 37 33 33 43 41 42 43 44 45 4S 47 48 4? 50 51 52
53 54
IENO QUADfl/t,4,5,11 ILOOP,3,2 QUADS'*,2,7,11 IEND JEND JLOOP,3,2 IL00P,J,2 DUADa/4,tfl,27 IENO ILOOPt3,2 QUADB't,2f7,27 IEND JEND IL00P,11,2 QUAD8/4,1,1,33 IEND I LOOP,13,2 dUAOB/4,1,1,35 IENO TRI6/3tl .21,35,23,35,23, 37 TRI6/3,li23,37f21,37,21,3^
SET,QUO.1,VARIABLE.2). 3ET,O^D,2,VARIAHLE.2CI. S 33JYOAR CONDITION DATA
(a)
Input Listinp
FIGURK :in.
116
J -!S3 OF S r r i i P
S >l)I >03EL
SEQUENCE
63
P L O T . 3 J T L I N E . 0 . , 3 . . S O . . - C,
67 END
63 PLOT,:JNTOURO. J..50. co. 63 I S O T H E R M S , , 1 5 70 END 71 STOP (a) input Listing s L:r *E:SJLTS
(con!)
SUBROUTINE CONDUCT t : 0 \ I D l . C D N D 2 . T , X , r , GNOSES . M S T . N E L i H . T I H E > DIMENSION C 0 N D 1 ( 1 ) C D N } 2 ( 1 ) , T f 1 > . < ( 1 ) , ( I ) C C C SUBROUTINE TO EVALUATE H E 3 H A L IF(MAT.Ea.2) IF(HT.E8.4) CONTINUE SALT SO TO 13 GO TO 53 CONDUCTIVITY
15 C C C
CONDUCTIVITY
00 20 I = 1 , N N 0 D E S TT=TCIJ AK=3.7C2l-.0a6871*TTK .252E- + TT*TT*<-.0E-7ITTTTTT CONOKI>=AK/m. CON02<II=COND1<I) 2!) CONTINUE RETURN SO CONTINUE C C C CRUSHED SALT CONDUCT! VI T1 DO 60 I = 1 , N N 0 0 E S TT=TI> AK = 3 . 7 0 2 C - . 0 3 3 8 7 1 T r ( . 2 5 2 E - * ) r r T T K - * ] E - 7 l * r T T T * T T AK=AA.10 CONDI(I)=AK'1Q0. C0ND2<I)=C0NDMI> 6C CONTINUE RETURN END
1
(b)
FIGURE 30.
117
3UUR0UTINE H I C = C C C C
C C C
>
l9.^5.
C C C
EVALUATE A I R PROPERTIES BETA=1./TFILM M U = < 7 . 3 < m E - 7 > C < T F ! L 1 1 . 5 > f C T - I L M * 1 }Bi )> RrtO=39.68/TFILM
C C C C C C C C C 10
EVALUATE GRASHOF NUH8EI GR = <fHORHOGBETAALLALQELT>'t AMU* AUDI CHECK BOUNDARY COHOUUH SD TO 3 0 CTLINOER SET NUMBER
IF<ISET.E0.2
EVALUATE H - - V E R T I C A L
IF(GR.LT.1.0E9I 10i20 CONTINUE LAMINAR FLOW HT=.29*DELT/AL*.25 IF(G).LT.l.OEt) iT=0. HT=HT*CONVERT RETURN 20 CONTINUE TURBULENT FLOU HT=.19(DELT.333333) HT=HT*CONVERT RETURN 3D CONTINUE EVALUATE H - - H O R I Z 0 N T A L IFCGR.LT.1.0E9) 3,50 CONTINUE LAHINAR FLOU Hr=.2 7C0ELT/AL>*.25 I F ( G R . L T . 1 . 0 E * > HT=0. HT=HTC0NVERT RETURN 50 CONTINUE TURBULENT FLOU HT = . 2 2 * 1 C E L T * ' . 3 3 3 3 3 3 I HT=HT*CONVERT RETURN END 40 (b) Heat Transfer Coefficient Subroutine Input ListingExample Problem 5 PLATE
C C C
FIGURE 30.
118
FIGURE 31.
119
fl _ ( >_a t_ C o n d u c t i o n
in a Slut i_ _ * I 1 J; i_r
1
The final example problem was cho.s^n Lo illu.slrat.r_ Lhe use. of four _ ; ' ! I'lmr-nis and the selective fuitput options in COYOTK. Tlie problem of JnuT-st
''insists of a copper bar of iv.-tangular cross-section that contains an : r,' e rn . " riangular slot. The bar is assi.nod insulated on three sides; energv is : r;\
r.iiterl in the bar through the fourth side by placing it in contact with > > ; ' . " bar of higher temperature, i i o t l i bars arc initially at uniform temperatures.
steel is at 30QC: and i.he copper at 100C. The finite element model used to solve this problem is shovn in Fin-ire " i along with the material properties and boundary conditions ?<>v the problem is given in Figure 33. The input lis: in Al
several points have been identified within the mesh as special output points. A transient solution to the problem is indicated with the generalized CrankNjcol.son time integration scheme fa = 0.667). In Figures 34-35 are shown some of the graphical results generated by analysis. ih
contact line is shown in Figure 34 ; a typical contour is shown in Figure 3o . Finally, Figure 36 shows time histories of some of the special output points noted in Figure 32.
120
St:el
T|=300C
V
Con act Line \
X
1
' J^
1 !
f
1 1
r
i Copper T, = 100 C - L ^ ^
_._ 1
I
Sipe] Propertii'!
3
Copl'i'i0 8 . SI
I>ro[io;-ti i-few'''
P = 7.fi c / c t n
K
C = 0 . 4 5 Ws/ K k = 0 . 5 4 '.V/cmK
KIGURI-. ; t 2 .
fxompjc
I'M b l e m
f i - l i c a r . Con.
$ SRI3
OATA-PARTl
9 1.9.14,15
10 11 12 13 1* 15 IS 1? 16 1? 20 21 22 23 Z* 25 26 27 28 23 30 31 32 33 3*
35 36 37 33 39
S EN3 SETUP SEQUENCE * - 0 * 1 MATRIX E 8 U A T I 0 N S OUTPUT, POI N T S , 1 . = , . 3 75 , 2 . 5 , . 375 , -5 , . B, 1 . 5 , . 9 , 2 . 5 . . 9 OUTPUT, POI N T S , . 5 , . 2 , 1 . 5 , . 2 , 2 . 5 , . 2 PLOT,OUTLINE,0. , 0. , 3 . , 1 . 2 5 $ L3T H03EL PLOT, ELEMENTS, 3 . , 0 . , 3 . , 1 . 25 , , , , , , ,N UMBER ZIPP.T3ANS2..6667,5.0. ,15.,.5,30 $ r f t l N S I E N T SOLUTION
c
.75..75.1. 25,1. 25 EMO JLOOP.3,1 ILOOP.15,1 QUAD*/*.2.1.1 IEND JENO JLOOP.2.1 ILOOP.13,1 QUAD*/*.2.6.* IEND JEND JLOOP.3.1 ILOOP.15.1 QUAD*/*.2.1.6 IENO JEND JLOOP.6,1 ILOOP.15.1 0,UAD4/4,1,1,9 IENO JENO END FORMK
$ ELE1ENT
OATA
40 END
* 1 PLOT,10FILE * iLDT RESULTS 42 TIMEPLANE, * , 6 . 1 1 , 1 6 , 2 1 43 T L O C A n O N , S I D E S , 1 5 , 9 6 , t , 9 7 , l , 9 B , l 9 l , 1 3 3 , l , 1 0 l . l , 1 0 2 , l 1 0 3,l,10*,l
44 45 46 47 43 49 50 51 52 53 5* 55 56
57 55 59 6C 61 62
1 0 5 , 1 , 1 3 6 , 1 , 1 3 7 , 1 , 1 0 8 , 1 ,10 3 , 1 , I I 0 , 1 TIMEPLANF,*,6,ll,16,21 TLOCATI ON, SI OES. 10 ,56,1,5 7, 1,53, 1,5 9, 1,63, 1.51,1, 62, 1,63, 1,64,1,65,1 TIMEPLANE,4,6,11,16,21 TLOCATI ON,NODES,13,31,4,32,4,33,I,34,4,35,4,36,4,46,4,5b,4,73,1, 69,1* 68.1,97,1,66,1 END PLOT.HISTOSY TLOCAriCN,3,1,31,6,1,56,1,151,* ENO PLOT,41 STORY TSPECIAL.2,1,31 ,3.6 END
PLOT,CONTOUR,0. , 0 . , 3 . , 1 . 25 PLOTLOOP.S.l ISOTHE3MS,2,9,12 0 . . 1 4 0 . , 1 6 0 . , 1 3 B . i 2 0 5 . , 2 2 1 . , 2 4 3 . , 2 6 3 . , 2 3 3 . PLOTENJ END STOP.N3PRINT
FIGURE 3 3 . 122
.
. 4 _ -
j~ " 1
;
I
t,
;
4 -
-;--,-i ;
i
:1 \
j
:
>
\ ! *
=
Hs-
.........
- -- f
i
!
3=!
PSsat.
1
1
1- J
1 " ^ -
. 1
.
I
DISTANCE
FliilHh
:'.-}.
'!'( :II;HM';I 1 u r n
IM'n 1 i 1 t K--L',>;arnpK
Problem
123
FIGUHh 35.
; i
j i
1
1
1
1
: !
1 1
' ;
\ ! ^
-x
!
i j r+
H
z:
/^
i 2 -
J
TIME j
FIGURE 36.
124
The following list provides examples of consistent units for quantities encountered in the use of the COYOTE program:
English foot (ft) second (s) Mass Force Energy Temperature Density Specific Heat lb m lb -ft/s m Btu Fahrenheit (F) Rankine (R) lb / f t n v Btu/lb -F ' m Btu/s Btu/ft -s Heat Transfer Coefficient Thermal Conductivity Volume Heat Source Btu/ft -s-F Btu/ft-s-F Btu/ft -s 4.755 x 1 0 ~ Btu S-ft -R
2 4 1 3 3 2 2 3 2
Metric centimeter (cm) second (s) gram (gm) , 2 gm-cm/s calorie (call Centrigrade (C) Kelvin (K) gm/cm cal/gm-C cal/s cal/cm -s 2 cal/cm -s-C cal/cm-s-C cal/cm -s 1.355 x 1 0 ~ cal
1 2
meter
im..
second (s) kilogram (kg) Newton (N) Joules (J) Kelvin (K) kg/m J/kg-K J/s (Watt) J/m -s
J/m -s-K J/m-s-K ,, 3 J/m -s 5.6697 x 10 J s-m >-K 4
125
The analysis of a transient conduction problem requires the selection of a suitable time step for the integration procedure. The selection of too large
a time step can result in a loss cf the transient temperature response and the generation of only a steady state solution. An inappropriately small time step
may produce nonphysical spatial oscillations in the early time temperature field due to the limited resolution ability (for temperature gradients) of the finite element mesh. Obviously, the judicious choice of a time step is important for The method for estimating a time
23 step as outlined below is due to Nickell and Levi and is based on an analy tic solution to tue one-dimensional heat conduction equation. In the following discussion, it is assumed that a finite element mesh has been constructed that will adequately model the thermal phenomena of interest (e.g., thermal shock problems will require a fine mesh near the boundary). a given spatial model, a local characteristic length, Ax, is chosen based on element size. Typically, this characteristic length is measured normal to a For
boundary on which a temperature or heat flux (source) disturbance occurs. Based on the characteristic length, the local heat transfer coefficient, h, and the local thermal conductivity, a local Biot number (Bi = hAx/k) can be com puted. For prescribed temperature, heat flux, or heat source boundary condi
tions, the Biot number is assumed large. In order to bound the thermal gradient that will occur during the first time step, the ratio of the temperature at a distance Ax (characteristic length) from the boundary to the boundary temperature is selected. defined by 9 = T ( A x ) /
T s u r f a c e
126
' . ' , ' : ' ' > i}-... f-s: i ir;n C-! vn1ues :=;;); li>'.;rii'r ! V . ; ! : ; 1 ; <' ;'Fe = -i-'w/^s") ..!'..-: : ! ' .
fer
local :nav be
temperature ratio, a in F i g u r e s 37
o
from the - h a r t s solution for : ' ; u i chart ; : i re based on an analytic 24 .. !I<J!I i i j i . i l a ,Tinv('(M ivr boundary j:.nr ' i i ! ' * : . i : f : and values fur condition. A v a l u e for the length and thermal local diffusivity, one-dimensional
the c h a r a c ; e r i s t i c computed.
the p r o c e d u r e o u t l i n e d
a b o v e , c o n s i d e r the
transient material
P'-eb' described
The pertinent
y< r t i es ( s t e e l ) w e r e given
length
Ax is b a s e d on the s i z e of e l e m e n t
number
(he b o u n d a r y Assuming
is s p e c i f i e d , the Biot n u m b e r is t a k e n
infinite. Thus.
6 = 1 0 ^ , then
F i g u r e 39 y i e l d s
a F o u r i e r n u m b e r of F o = 0 . 2 ,
(0.2K0.35)
(0.0622)
2 =
u..}^ s
s was used
in the
first
i n terval .
127
1.0
p~rT T l
-
["IT
1 MJ,
r-i
II_
/ /
"
10
/
_ -
/ /
1
id
-2
i i i
1 /
i
/
III
io
i 10
I I I _L
1 I 1
III 10
ll.U
1 1 I I L.-ULL
10
1.0
FOURIER
NUMBER, Fo
FIGURE 39.
< Bi <
i i i
-1
10
V
ft
-2 10
<i
io"
3
-4 10 10 10
10 FOURIER 10 NUMBER, Fo 10
10
10
FIGURE 38.
10
i
id
oy
r
b
10
- ^r
io
4
10
10
10
10
FOURIER
10
NUMBER, Fo
10
,12 10
FIGURE 37.
< Bi < 10
i i i !
1.0
_ -
10
REFERENCES
i. \i.
in I [roncing Analyzer for 3rd Generation Computers (CIKDA-3G) , " TN-AP-67-287, i' vy;,]r-i' Corporation, Space Division, Now Orleans, LA, October 1967. 1 ' K Gm-t 1 i ng , "MKRJ.IN--A Computer Program to Transfer Data Between Finite
! ifuont Meshes.' SAMiSl -0463, Sandia National Laboratories, Albuquerque, " . : ! . ..-pi ember 198 L. : ' . K. Gartliiig, "NACTIOS A Finite Element Computer Program for Incompres sible Flow Problems." SAND77-1333 and SAND77-1334, Sandia National Labora tories, Albuquerque, NM, April 1978. U. K. Gartling and C. E. Hickox, "MARIAHA Finite Element Computer Program for Incompressible Porous Flow Problems," SAND79-1622 and SAND79-1623, Sandia National Laboratories, Albuquerque, NM, August 1980. R. S. Dunham and E. 1 ! . Decker, "TEXGAPThe Texas Grain Analysis Program," TICOM Report 73-1, Texas Institute for Computational Mechanics, Austin, TX, August. 1973. I. I . Wilson and R. E. Nickel], "Application of the Finite Element Method to Ileal Conduction Analysis," Nuclear Engineering and Design, Vol. 4, 1966, pp. 276-286. G. C.'omini , 8. del Guidice, R. W. Lewis, and O. C. Zienkiewicz, "Finite Element Solulion of Non-Linear Heat Conduction Problems with Special Refer ence to Phase Change," Int. J. Num. Meth. Engng., Vol. 8, 1974, pp. 613-624. II. A. Finlayson, The Method of Weighted Residuals an' Variational Principles, Academic Press, New York, NY, 1972. (I. c. Zienkiewicz, The Finite Element. Method, McGraw-Hill, London, 1977.
131
REFERENCES (cont)
10.
W.-l,. Wood and R. K . Lewis. "A Coinnarison of Time Marching Schemes for Hie Transient Heal Conduction Equation," Int. .1. Num. Moth. Engng., Vo! . '.) . 1975, pp. 670-089.
11.
J. II. Argyris, 1 . . . E. Vaz , and K. J. William, "Higher Ordor Method., I ' o r Trapsient Diffusion Analysis," Comp. Meth. App] . Moch. Engr., Vol. 12. 1977, pp. 243-278.
12.
H. E. Nickel 1 and II. I). Hibbitt, "Thermal and .Mechanical Analysis oi' Weld-.i .SI.nict.uros," Nucl. Eng. Dos. , 32, 1975, pp. 110-120.
13.
H. E. Nickoll, "Applications of the Finite Element Method in Solid Mechan ics, Fluid Mechanics, and Heat Transfer," Developments in Mechanics, V. 11. S. Proc. l<lth Midwestern Mech. Conf., Norman, OK, 1975, pp. 599-626.
14.
S. J. Womack, "Shape Function Techniques for Generating Finite Element Grids," TT00.M Report 73-3, Texas Institute for Computational Mechanics. Austin, TX, 1973.
15.
D. K. Gartling, "Texas Fluid Analysis Program--User's Manual," TICOM Report 75-2, Texas Institute for Computational Mechanics, Austin, TX, 1975.
16.
I. Ergatoudis, B. M. Irons, and 0. C. Zienkievicz, "Curved Isoparametric, 'Quadrilateral', Elements for Finite Element Analysis," Int. J. Num. Meth. Engng., Vol. 4, 1968, pp. 31-42.
17.
P. C. Hammer, 0. P. .Marlowe, and A. H. Stroud. "Numerical Integration over Simplexes and Cones," Math. Tables Aids Comp., Vol. 10, 1956, pp. 130-137.
18.
B. M. Irons, "A Frontal Solution Program for Finite Element Analysis," Int. -1. Num. Mot I). Engng. , Vol. 2, 1970, pp. 5-32.
19.
S. L. Thompson, "HSCOItSA Revised SC0RS Plot System," SAND77-1957, Sandia National Laboratories. Albuquerque, KM, May 1978.
20.
S. I . . . Thompson, G. C. Padilla, and IV. A. Sexson, "RSC0RS Plotting at SNLA, Version 1.08," Internal Computer Document (unpublished), Sandia National Laboratories, Albuquerque, NM, January 1982.
132
REFERENCES (conI)
21.
R. W. Simons, "Remote Hard Copy Programming Reference Manual," Internal Computer Document (unpublished), Sandip National Laboratories, Albuquerque, Ml, September 1979.
22. 23.
R. E. Nickell, private communication, 1977. I. M. Levi, "User's Guio-? to the Transient Heat Conduction Finite Element Code HTCON," MM70-5424-17, Bell Telephone Laboratories Memorandum, tthippany. N.J, 1970.
24.
H. S. Cars law and J. C. Jaeger, Conduction of Heat in Solids, Oxford, NY. 1947.
133