Linearstaticfea Guide
Linearstaticfea Guide
Produced by Automated Analysis Corporation 2805 South Industrial Highway Ann Arbor, MI 48104
Copyright 1994 All Rights Reserved Automated Analysis Corporation Ann Arbor, MI 48104
ACKNOWLEDGEMENTS
The preparation of this finite element course required coordination with several groups. The MacNeal-Schwendler Corporation's Richard MacNeal and Bruce MacNeal allowed the use of materials and concepts from an earlier MSC/NASTRAN<B>video lecture series. Henry Fong and Mark Skidmore from MARC Analysis Research Corporation assisted with MARC lab problems. Dr. V. B. Venkayya from Wright-Patterson AFB helped with ASTROS concepts. Each of these three organizations has a University program to assist Universities in installing the commercial finite element program. My original attempt at finite element education using commerical codes was encouraged by Jeny Joseph (MacNeal-Schwendler), Louie Nagy (Ford Motor), Joe Eisley (University of Michigan) and Harry Schaeffer. It is relatively easier today to work with large codes at a university than it was 13 years ago. Doctoral students Michael Eldred, Jungsun Park and Michael Dungan have helped in the preparation of this material, particularly the laboratory sessions. Many unnamed students have contributed to this course by finding errors in the notes. They can feel good about being smarter . than the professor!
FOREWORD
The material in this course has been developed at The University of Michigan over the past 23 years, where I have taught Master's level finite element classes. The large class sizes dictated that I prepare the material in advance; it is difficult to teach extemporaneously to more than a hundred listeners, and errors in the material are difficult (and embarassing) to correct at the board. The course is based on virtual work and potential energy. Modeling concepts are emphasized because of their importance--the real problems in the field are to understand meshing, boundary conditions, reflective planes, multipoint constraints, convergence, thermal stress, etc. The course uses three commercial codes for laboratory sessions: MSC/NASTRAN~ MARC and ASTROS. These represent standard industry practice. MSC/NASTRAN is relatively strong for linear statics and dynamics , MARC is relatively strong for nonlinear problems and ASTROS is relatively stong for optimization. The course is intended to be even-handed with respect to all commercial codes, but in some areas of theory, it is necessary to choose a dominant method. Here, I often use the NASTRAN approach because of the importance of the various NASTRAN codes (COSMIC, MSC/NASTRAN, UAI and CSA). This occurs in partitioning of equations prior to solution, multipoint constraints and rigid body elements. The course is about 80/0 code-neutral, however. This is the first multimedia course presented with annotation of figures. The "LivePen" method was developed by Glen R. Anderson of Palo Alto, CA.. It allows the presentation of 30 hours of lectures on 5 CD-ROM discs, which is a compression and quality of material not previously possible. The material can be viewed on Macintos6M computers and Microsoft Windows computers. There are CD-ROM playback units that can handle 5 or 6 discs at once, and allow fast access of all information in the course. There is a window in which written text is scrolled, and this may help nonnative English speakers. Keyword look-up and figure selection is allowed for all940 figures in the study guide. A second series of 17 lectures on structural dynamics will be released in 1995. A third series on numerical acoustics will follow. William J. Anderson Ann Arbor, Michigan November 21, 1994
TABLE OF CONTENTS
CHAPTER
TITLE
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30.
Nomenclature Introduction Rod (Pin-Ended Truss) Element Constant-Strain Triangle Lab Problem 1: Membrane Case Study: 2-D Contact Problem Interpolation Virtual Work and Potential Energy Derivation of a Finite Element by Virtual Work Beam Element Case Study: Autohaul Trailer Lab Problem 2: Cantilever Beam Modification of Equations Linear, Static Equation Solvers Bandwidth and Wavefront Classical Plate Theory Plate Finite Elements Lab Problem 3: Plate Coordinate Systems Gauss Integration Isoparametric Elements Solid Elements Case Study: Hydraulic Valve Lab Problem 4: Gear Tooth Thermal Stress Axisymmetry Modeling Concepts Convergence Steady State Heat Conduction Steady State Heat Conduction (Continued) Dynamics and Nonlinear Problems
NOMENCLATURE
a
A
Value for constant strain in x direction, or dimension, or acceleration Area, or constant of integration Value for constant strain in y direction, or dimension Semi-bandwidth of a matrix, or constant of integration Strain matrix or damping matrix Scaling constant, used in shape function derivation, or constant of integration Flexibility matrix Strain-displacement matrix, or diagonal matrix Strain energy or Young's modulus Equilibrium matrix Nodal forces on element Force component Nodal forces on system Node (grid) numbers for beam (left and right ends) Shear modulus Stress-strain law in element Thickness of sheet material Transformation matrix relating nodal coordinate and generalized coordinates Second moments of area of beam cross-section, opposing deflection in principal planes 1 and 2, respectively Torsional stiffness or Jacobian Stiffness component, or spring stiffness Element stiffness System stiffness Length Line load (force /length) Lower triangular matrix Lower unit triangular matrix line segment
b
B
[B)
c
[C) [D)
E
[1:]
{f)
Fi
{F}
GA,GB G
[G]
[H)
k
[k]
[K]
L
{L} [L]
[Ll]
[m]
[M]
{M}
System mass Moment resultants Newtons Shape function Pounds per square inch Pressure distribution Generalized coordinates Rotational degree of freedom Field displacement Nodal displacement for element Nodal displacement for system Strain energy Upper triangular matrix Upper, unit triangular matrix Coordinate transformation at a node Translational degree of freedom
[N]
psi p(x,y,z)
{q}
~
{u}
{ue}
{u}
u
[U]
[U 1]
[t]
[T]
{.Y}
Coordinate transformation for an element Traction on surface (force I area) Volume, or shear force in beam Nodal shear forces in beam Intermediate vector in Gauss elimination procedure Work Work potential Body force (force/volume)
v
{V} {X}
w
'HI
{X}
Greek symbols
'Y 8 A
E
{E}
K
S(x)
e
'A
A
v II P
{<P} [<P]
at
Dundurs variable in contact problems, or coefficient of thermal expansion (linear) Shear strain component Variational parameter Increment Direct strain component Strain vector (contains both direct and shear components) Dundurs' variable in contact problems Angle of twist along beam Rotation at node (in structural elements) Eigenvalue Shifted eigenvalue Poisson's ratio Potential energy Mass density Eigenvector Displacement function matrix Direct stress component Stress vector (contains both direct and shear components) Shear stress component Curvature Frequency of vibration, rad/ sec Natural frequency of vibration, rad/ sec
Subscripts
0 a e ext I
II III
Initial (reference) configuration Analysis set of degrees of freedom Element External State One: unloaded and undeformed State Two: loaded and in equilibrium State Three: virtually displaced from State Two
int
r
s v
-1
(i)
Inverse (of a matrix) ith iteration cycle Transpose (of a matrix) Wire Pliers
Overstrikes
T
w
Modified quantity
Velocity Acceleration
Additional Definitions "Elasticity" elements: Those elements derived from classical one, two and three-dimensional elasticity in which no rotations at a point are defmed. These include truss, axisymmetric, plane stress, plane strain and solid elements. "Structural" elements: Those elements derived from beam, plate and shell theory, in which rotational degrees of freedom are defined at nodes. These include all beam, plate and shell elements. "Detailed" nodal numbering: The scheme of nodal numbering where every degree of freedom is numbered and given a separate subscript. Computer programs report bandwidth in this notation. "Compact" nodal numbering: The scheme of nodal numbering where all degrees of freedom at a node are represented by a single symbol. This leads to a simpler notation for discussing assembly concepts, which depend primarily on nodal connectivity. This is a shorthand, nonphysical notation
I I I I I I I I I
I
I I I I I I
LECTURE 1
INTRODUCTION
COURSE OVERVIEW PERSONAL INTRODUCTION DEFINITION OF FINITE ELEMENT HISTORICAL REVIEW OF THE FINITE ELEMENT METHOD EMERGENCE OF F. E. A. AS STANDARD PRACTICE EXAMPLES OF PROJECTS COMMERCIAL PROGRAMS (PARTIAL LIS'}'
USA STRUCTURAL FEA PROGRAMS EUROPEAN STRUCTURAL FEA PROGRAMS STRUCTURALBOUNDARYELEMENTPROGRAMS FLUID FEA PROGRAMS
REFERENCES
Turner, M. J., Clough, R. W., Martin, H. C. and Topp, L. J., "Stiffness and Deflection Analysis of Complex Stn.Ictures," Journal of the Aeronautical Sciences, Vol. 23, No. 9, September, 1956, pp. 805-823. Williamson, F., "A Historical Note on the Finite Element Method, " Int. Journal for Numerical Methods in Engineering, 1980, pp. 930-934.
Hibbitt, Karlsson and Sorensen, Inc., 1080 Main St., Pawtucket, RI 02860 (401) 727-4200 FAX: (401) 727-4208 ADINA R & D, Inc., 71 Elton Ave., Watertown MA 02172 (617) 926-5199 FAX: (617) 926-0238
150 Beta Drive, Pittsburgh, PA 15238-2932 (412) 967-2700 FAX: (412) 967-2781
Swanson Analysis Systems, Inc, P. 0. Box 65, Houston PA 15342-0065 (800) 937-3321 FAX: (412) 746-9494 United States Air Force, Wright-Patterson AFB OH
1.1
COSMOS/M
Structural Research and Analysis Corporation, 1661 Lincoln Blvd., Suite 200, Santa Monica CA, (213) 452-2158 FAX: (213) 399-6421 CSA Corp., Agoura Hills, CA, (818) 707-6060 CASA/GIFrS, Inc., 2761 North Country Club, Tucson, AZ 85716 USA, (602) 795-3884 FAX: (602) 795-3886 Structural Dynamics Research Corp., 2000 Eastman Dr., Milford, OH. (513) 576-2003 Celestial Software, 2150 Shattuck Ave, Suite 1200, Berkeley, CA 94704 (510) 843-0977 FAX: (510) 848-9849 Livermore Software Technology Corp., 2876 Waverley Way, Livermore CA 94550 (415) 449-2116 FAX: (415) 373-0588 MARC Analysis Research Corporation, 260 Sheridan Ave., Suite 309, Palo Alto 94306 USA (415) 329-6800 FAX: 415-323-5892 Rasna Corp., 2590 N. First Street, Suite 200, San Jose, CA 95131 (408) 922-6833 FAX (408) 922-7256 The MacNeal-Schwendler Corporation, 815 Colorado Blvd . Los Angeles, CA 90041, (800)-336-4858 FAX: (213)-259-3838 Structural Analysis, Inc., Austin, TX (800) 388-8134 Engineering Mechanics Research Corporation, P. 0. Box 696, Troy, MI 48099 (313) 689-0077 FAX: (313) 689-7479 PDA Engineering,2975 Red Hill Ave, Costa Mesa CA (714) 540-8900 STARDYNE.TITAN Corp., 20941 Devonshire St., Chatsworth CA 91311 (818) 773-8174 FAX: (818) 773-8915 Universal Analytics, Inc., 3625 Del Amo Blvd. Suite 370, Torrance CA 90503 310-214-2922 FAX: 310-214-3420
CSA/NASTRAN GIITS I-DEAS IMAGES LS/DYNA MARC MECHANICA MSC/NASTRAN mTAB/SAP386 NISA
FEGS Ltd, Oakington, Cambridge CB4 5AF England 0223 237111 FAX: 0223 234192 FEA, Forge House, 66 High Street, Kingston Upon Thames, Surrey KT1 1HN, England, 081-541-1999 FAX: 081-549 9399 PSI/ESI, 20, rue Saarinen Silic 270, 94578 Rungis, Cedex, Paris, France. 33 (1) 49 78 28 00 INTES GmbH, Nobelstrasse 15, D-7000, Stuttgart 80, Germany, (49) 711 68783-0 FAX: (49) 711 68783-10 Belgium VERITAS SESAM SYSTEMS A. S., P. 0. Box 300, N-1322, Hovik, Norway, +47 2 47 89 96 Salvia Engineering AB, Ostra Ringiagen 4, S-722 14 Vasteras, Sweden +46-21 144050 FAX: +46-21 188890 FRAMASOFT, 10 rue Juliette Recamier, BP 3083 69398 LYON CEDEX 03, France 72 74 89 60 FAX: 72 74 89 99
SOLVIA SYSTUS
COMET/BEA
FLOTRAN
PH LEX
RAMPANT
I COURSE
GOALS:
OV~RVIEWI
To understand F. E. Theory To model structural systems To use commercial FEA codes Lectures, laboratories, case studies Engineers, physical scientists, mathematicians Introductory
1.4
SUBJECT:
~~~~-~:
EFFECT
I PERSONAL INTRODUCTION
William J. Anderson
pe~rees_:
full:time jo~~;
Prof. 1/Lt
1.5
1965-present 1962-1965
G)
~_y.mmer_ jops;
Ford, Chrysler, GM, Caterpillar Business Interests: President and Chairman of the Board of Automated Analysis Corporation, Ann Arbor Hobbies:
: xx x
I A hypothetical subdivision of a structure or system, ! ! possessing a simple shape which can be analyzed. ;:
~ ~~ ---~""""'~~"'~~--...v~'N>.V."""''~!'... .? -"._.y.AJ~V~.A~.-.~ANV~~'-'""~".WW
.....
rFiNrfEELEMENr: -
l
..Il
The finite element method requires: a) development of individual elements, often with concepts from classical mechanics b) assembly of elements into structure or system c) solution of the assembly using modern numerical analysis and computing d) recovery of field variables (stress, strain) within the interior of the elements
1.6
o)
Line element (truss, beam, pipe, electrical resistor)
~-u~
fl
~~~
C>
~~~--,.--------....)
f2
Solid element
1943
1956
C ...,_____.F--~-.J:!,.~. ,.~~.
G ...,___ _ _ _ _ _ _
M
... P
0 y_2 y_l
Yo
X
iiJIIa'
yl y2 y3
y4
'
!Y
Finite element mesh
1956 1960
I
1970
I
1990
I
Electrical Engineering
r-1.10
--t>
IEXAMPLES OF PROJECTS
PROGRAMS
ASTROS CSA/NASTRAN MSC /NASTRAN NASTRAN UAI/NASTRAN ANSYS ABAQUS
MARC
US Air Force, WP-AFB CSA Corp. MacNeal-Schwendler US Govemment, COSMIC Universal Analytics
Swanson Analysis Systems Hibbitt, Karlsson, Sorensen MARC Analysis Res. Corp. @
1.12
Algor, Inc. Structural Res. & Analysis Structural Dyn. Res. Corp. PDA Engineering U. of ARIZONA, H. Kamel ADINA R & D, Inc. Celestial Software RASNACORP. Structural Analysis, Inc. Engineering Mech. Res. Corp. Univ. Calif., Ed Wilson STARDYNE MIT
NISA
SAP IV, SAP6 STARDYNE
STRUDL
EUROPEANFEAPROG~S
ASAS ASKA BERSAFE CASTOR DIANA FAM LUSAS PAFEC PAM-CRASH PE~S Atkins, England IKOSS,Gennany Nuclear Electric, England CETIM, France DIANA Analysis, Netherlands FEGS Ltd., England FEA, England PAFEC Ltd., England PSI/ESI, France INTES,Gennany
1.13
RADIOSS
France
1.15
I I I I I I I I I I I I I
I I
LECTURE2
GENERAL TWO-NODED ELEMENT OVERVIEW OF TWO-NODED ELEMENTS ROD ELEMENT ASSEMBLY PROCESS OTHER SIMPLE LINE ELEMENTS PROBLEM SESSION
1. 2. 3. 4. 5. SOLUTION OF ROD ASSEMBLY STIFFNESS AND ASSEMBLY OF SPRINGS ASSEMBLY OF RODS AND TRIANGLES VARYING-AREA ROD ELEMENT DISPLACEMENT AND SHAPE FUNCTIONS
2.1
~;
~
~ ~~{
/----------~ ;=~;
.;
Forces and displacements at nodes are related. Examples: rod (pin-ended truss), torsional member, spring, beam, pipe, column, gap.
2.2
'
- - -extension - --
--
-~
-torsion ---- -
-~
The NASTRANROD element combines these properties. Civil engineers call the extensional element the pinned-ended truss.
A general relation between nodes can violate equilibrium. (Examples: ELAS l and ELAS2 in NASTRAN.)
o-= fl
ul
Even though the stiffness is symmetric, the element can unbalance the structure. We need physical laws--equilibrium or energy conservation--to "tame" the relation. G)
2.3
IROD ELEMENTI
(~o----------~--------~o>
~X
Features: two nodes constant area A straight axis, loads aligned on axis element coordinate system, x Assume: linear elastic material
O"x = Ex
Many engineers have trouble understanding displacement and strain. The displacements are infinitesimal (although enlarged, here). All geometry ideas are referred to the undeformed body.
~........ii}~
u(x)
~f.--.....----u-2-::.: : :"
fu
0
l u(x)
I
L
2.4
o--;:
fl
ul
!2=- fl
/).f
L
:::::: u2- ul
= E(u2 ~ul)
Relate nodal forces to intemal stresses through free-body diagrams:
Rearrange:
In matrix form:
{f} = [k]{u}
EA[ 1-1] L -l 1
Comments
o
0
Stiffness terms kij are conductors of force. The electrical analogy is conductance. Each kij is the force at d. o. f. "i" due to a unit displacement at only d. o. f. ']". Theoreticians love stiffness. Experimentalists hate stiffness; they love flexibility cij (the displacement at a d. o. f. "i" due to a unit force at only d. o. f. ']"). @
2.6
!ASSEMBLY PROCESS
Use a topological approach, with truss elements. Distinguish element nodal displacements u ei from assembly nodal displacements u i.
~ Uel ~---- Ue2 ~----'~ U e3 ~-----"" Ue4
-r..'
/....
2
Ue,_1
.;
K--, ..
where:
{\. J -
r - E1A1
Ll
~-- ~ .... -
~:, - E2A2 L
2
~ .. _,
s...;~.,j
+ ft .I ~J: )
- t .t- -~
.r~:
2.7
t..
...6...
~-- _ ;.:
--~
a..~---
The stiffness matrices add "term-by-term." { )( + 17} = [_p= + (} ] {Z} The rod elements are assembled:
> ,-~ :
! ":'
, ,_ . :
: ,,. :.
~k., q , ..::..
~ '
..:
!<"
A
, ....
L.
u~
e2
;_, '
,:,
"
sum of / stiffnesses
independent d.o.f.
Surviving d.o.f. are the assembly coordinates. We need to clarify what the forces are. @
F.~r~~ ~q~ilil:Jrat~g~H . Cl-t H Cl- ~99.~
:r
L forces on the pin
= 0
f2+!3=:T
C,::
2.8
{F} = [K]{u}
Comments Do this process 2 or 3 .times to develop intuition. Thereafter, use assembled notation only. Nodes are actually "welds," not pinned joints. The rod element is an "elasticity" element.
Heat conduction:
Electrical resistor:
2.9
IPROBLEM SESSION)
ProbleDl 1. Solution of Rod Assen1bly Find the displacements in the rod assembly:
~ ul ~~-~==~,. . u2 ~----=~=::,. . . u3
1
II
CD
4
';
7to0oN
4
E 1A 1 1L 1= 1.0X10 N/mm
E 2 A 2 1L 2 = 2.0X10 N/mm
3 -2 -2 2
_1
0 ]{ ~ . } {
u2 =
u3
J~}
1000
-1 104 [ 3 -2 -2 2
o]{u~} u; =
F1 } 0 1000
The first equation is partitioned out and set aside. This will allow solution for the reaction, later. -. , 10 4
!-~1
=-
U2
2]{u2} _{1000 0 }
u3 -
~u2 = 0.15 mm
If desired, one can retum to the neglected equation to get the force of constraint:
F1 =-10 4 u 2
= -1000 N
~--------:::;;:-:--0
~~---:::;;,,,. 0 .1 0 ~-~--~~::;;,,. .
1000N .. (o
(o
2.11
Q)
A linear, elastic coil spring is to be treated as a single finite element with 4 d.o.f. The spring is 200 mm long and 40 mm in outside diameter. It is made of steel wire of l mm diameter.
u1
u
2
-~--~-~:::;,.._.
t:~-'
-~::;-...;,-x.......
Two experiments have been carried out to provide stiffness data. The left end of the spring is clamped.
u = 1 rad
4
Assume the spring remains linear. A) Using elastic symmetry, geometric symmetry and equilibrium concepts, construct the stiffness matrix for this spring. B) If two such identical spring elements are joined in series, find the stiffness matrix for the assembled system. @
2.12
k1.3
-100
k23
0 1 0
-50
!3 !4
Hence: k 13 = -100 k 23 =-50 The second experiment gives: -50 -100 kl4 0 -50 k24 0 - -100
0 1
!3 !4
kt4 =-50
k24 =-100
From equilibrium: !3 =-ft The first experiment yields: -100 -50 0 -50 -100 0
k33
k 43
-100 -50
1()0
50
Hence: k 33 = 100 k 43 =50 The second experiment yields: -100 -50 0 -50 -50 -100 0 -100 100 k34 0 50 50 k44 1 lfJO Hence: k44 =100
2.13
ij
=k
ji:
-50
-10()
Due to geometric symmetry, the forces are identical in the following two experiments:
. . . .\::~ ~:
::::~:::..:
::-~::;~~:::;.......... ..-.;:::;::...........
~:::::-:::::::
Exp. 1':
-100 -50 -1 -50 -100 0 k2.1. -100 -50 100 50 0 -50 -100 50 100 0
k11
Hence:
kll
= 100
k21
=50
@ @
-.~::!:-> -:;:::=~~;::::.
-::~:::::::::Y..
. .:::~~;:;:::.-- :::::::
Exp. 2':
100 k12 -100 -50 0 50 k ~~ -50 -100 -1 L..:..., -100 -50 100 50 0 -50 -100 50 100 0
k21
=50
k22 =
100
0 0 100 50 -100 -50 0 0 50 100 -50 -100 200 100 -100 -50 -100 -50 200 -50 -100 -50 -100 100 100 50 0 0 -100 -50 0 50 100 0 -50 -100
;,
'-./
'".N'rtNY'a~~~"'it/'..'NV''t
Solution
............
a a a
a a+ - =,. a
;.;~
a a a+ li
, ~
0 0
8
0
...~~
..:.
0 0
0 0 0
0
-~ ..l y
(j
:::
0 0
0 0 f3 + l) f3 f3 {3 f3 + '/ {3
;1
f3
2.15
f3
f3
Find the exact stiffness matrix for a two-noded, varying area element:
!2
fl
~o--------<0 ~
A 0 (1 + Bx/ L)
!2 - = -(J = - - E
EA 0 (l + Bx/ L)
f = J0
!2
EA 0(l + Bi/ L)
dx + u1
X
= !2
Lln(l+Bil L) +u 1 EA 0 B 0
~
f2 L =--ln(l+Bx/L) +u 1 EA 0 B
2.16
h = (u2 -ul)
By equilibrium:
II=-
EA 0
ln(l +B)
B
h = (ul-u2)T ln(l+ B)
EAo
Problem. 5. Displace111ent & Shape Functions The displacement function for a two-noded line element is: u (x ) = q + q x
1
2
x/L]{~J
Relate the constants q 1 and q 2 to the nodal displacements u 1 and u 2 . Solution Evaluate the field displacement at the nodes:
'itX-.V..;)i WJI. )o())::O:,O
u(0)=u 1 =q 1 +q 2 0
u(L)= u 2 = q 1 + q 2 L
2.17
q 1 =u 1
LECTURES
DERIVATION
A. B. C. EQUILIBRIUM STRAIN-NODAL DISPlACEMENT RElATION STIFFNESS
3.1
IoVERVIEW I
The aircraft industry of the 1950's needed new methods for stress and vibration analysis. Aeroelasticity lagged. Membrane theory was (and is) used for wings. The birth of practical FEA: Turner, Clough, Martin and Topp, "Stiffness and Deflection Analysis of Complex Structures," Journal of the Aeronautical Sciences, Sept. 1956. Rectangle and triangle elements were presented.
3.2
IEIASTICI'IY .REVIEW]
A. 3-D
stressstrain strain-
displ.
<== [G]
<==
{Ux}
Uy
Uz
[D]
equilibrium
equations
B. 2-D
d:l. {Ux}
[D]
Uy
strain-
equiJibriurn
equations
t : ................
~-x
h~O
h~oo
The mappings needed for plane stress are the stress-strain (constitutive) law:
{ax} cr
r
y xy
vE
- vE
l!v2
I-v2
1-v2 E 1-v2
\....
0
v
~ { ~~} G Y.xy
,)
Ex
a
0
ey
Yxy
}=
(}X
a {Ux}
Uy
dx
[D]
nodal forces
stress
strain
{!} = [k]{u}
3.4
A. EQUILIBRIUM
Develop a mapping from intemal stresses to nodal forces:
This is risky because the element is statically indeterminate. Consider a sheet of metal with a constant cry stress. We will superpose cases, later.
+
/
+
/
/ /
'
- ---
- --
'
3.5
C8
'
Consider f4 , say:
14=
cryc.x{-x 3 )h
2
-
cryvr{-x 1)h
2
!4
fs
!2 !3
16
\..
XI3h
[~]
v
{~;}
.J
The entire matrix can be constructed this way. Finding the terms due to shear is difficult. One must deal with both horizontal and vertical forces in the f. b. d. (See Problem 5.)
3.6
B. DISPLACEMENT RELATIONS
Tumer, et. al., assumed constant strain:
{ a } [D] b =
c
{ux}
Uy
ul
=
[?]
u2
strain
{c
dydx There are three equations in two unknowns! This requires the concept of "compatibility." @
~} = ~ 1 {:x} a a
y
ax
=ax+ f(y)
arbitrary function
-ey
J
=by+ g(x)
3.7
- auy
arbitrary function
Can we choose f(y) and g(x) so as to have a unique displacement field? This is the compatibility question; in general you cannot! The third equation requires:
c=-x+-
au auy ay ax
_ df(y) dg(x) c- dy + dx
This can only be true if all three terms are constant: , say. c = A + c-A
==> ==>
f (y) = Ay + B
g(x)=(c-A)x+C
dg(x) _ A dx -cux
Uy
= ax + f(y) = ax + Ay + B
=by+ g(x) = by+ (c-A) X+ C
a
{::}=[~
\.....
0 0 y X
y 1 0 b -x 0 1 c
-......r
./
[c/J]
3.8
b
c
[?]
A B
<==
u1 u2
u6
nodal displ.
The remaining mapping is found by a trick. Evaluate the field displacement at the nodes:
ux(x 1,y 1)
=u1 =ax 1
. . .
+ Ay 1+ B
by 1 +cx 1-Ax 1
. . .
+C
+ c {,;;;\
\!;
0
y1
u3
u4 u5
x2
0
y2
0
x3
u6
0
y3
0
'-
1 0 a x1 -x 1 0 1 b 0 c y2 1 0 x2 -x 2 0 1 A y3 1 0 B 0 x3 -x 3 0 1 c _,y ............,_. 0
y1
[H]
The relation must be inverted:
-1
{q}
{q}=[H] {u}
One can invert [H] in sY!_Ilbol form using Mathematica, Maple, etc.
3.9
C. STIFFNESS
a
b
c
[Hfl
=
A B
ul u2 u3 u4
us
u6
(O,y3)
=x 2 - x 1.
@
[k]=
y3 x2l vx 2
X21
@
VX2 Xzl x2
2
--
--
y3 Xzl vx 2
X21
1 --
vx
Xzl
-v
xlx2 Y3X2l
vx 1 X21
0 0
0 0
-v xl Y3
Eh 2(1- v 2 )
_ _h
X21 1 -X21 vx
Y3X2l _vx 2
X21
x2 Y3
Y3
X21 vx 1 x21
xlx2 Y3X21
x2
Y3X2l
0
-v
0
x2 Y3
0
v
0
Y3
0
X21
XI
Y3
(continued)
x2
2
x2
X21
Y3X2l x2
X21
--
x2
Y3
0 0
0
Y3
X21
-Y3 X21
X21
-1
xlx2
+Gh 2
Y3 X21
- x2
X21
- XI -
x2 Xt
xl
y3
1
X21
xl -
-Y3 -1
0
Y3
X21
X21
0 0
0
x2
Y3
0
Xt
Y3
0
1
0
Y3
0
3.11
IASSEMBLYI
a a a a
ex a ex
a
aacta aaact
e t a aa
K21
K22
K15 K25
F3 - . F4
Fs K51
K52
Kss
: li!llllillil!:!lil,i.lillllll:
a
a
a
a+JJ
f-J:
t:{
{'t
.{ .......
0 Ff
!/
i
[K] =
0 0
a
/3 +
'1/
~
l
0 0
'V
i
I
.",~
0 a+ JF-...J:
."',<'
,,.
i3 + ,r,
.. .
a+ /3
s
t!
i3 + ""/
~
a+ /3: +
. . . . I'
~
>'
3.12
(COMMENTSI
Think of element stiffness as conductors of force. Think of interelement boundaries as welds (in addition to nodes as welds!). Compact notation simplifies assembly. Use of constant strain elements in 2-D and 3-D causes error when stress reversal across an element is required (as in a single layer of elements through a beam thickness).
IPROBLEM SESSIONI@
ProbleiD 1. Assembly of triangles
Consider a two-dimensional assembly made of three constant-strain triangles. Using the symbols ~ [J 0 , show the form of the stiffness matrix.
3.13
Solution The element~ connects nodes 1, 2, and 4. The element o connects nodes 2, 3, and 5. The element 0 connects nodes 4, 5, and 6.
= {F}
~- x
t
3.14
Solution Complete the nodal numbering. The displacements at the wall are constrained. The forces are given on the outboard nodes.
u4 u6
0 0 0 0
FI F2 F3
us
ul
u6 u7
Ug
F4 0 0 0 -1000
/
Proble111 3. Deflection of a triangle Consider an equilateral triangle made of steel of thickness 10 mm. The sheet is clamped at the bottom edge. A vertical load of 1000 N acts at the top. Use a single constant strain element to estimate the deflections at the upper vertex.
E = 207,000 MPa
V=
lOOON
0.3
3.15
_t __ _
173.2 mm
u 5 =0
Model the right triangle with one triangular element. Find the deflection under the load.
= 7 95
.
X
104 MPa
0
u3 u4
F1 F2
0 -1000
Fs F6
3.17
Y3
Eh x2I 2 2(1-v) ~
X21
u3}-{ 0 } u -1000
4
2( -1000) x 21
2( -1000)(100)
r xyh(x3 - x 1)
.....
!
:
r xy h(x2 - x3 )
.
~i ~
""-"
('(')! ~!
,--...,_
N ~
~'
I
('(')
...s::
~ '-"'
3.18
Propose that the concentrated forces match the total distributed load on the opposite face:
~
~ .
~
rxyh(x 3-x1)
I
(x 3 - x 1)/2
--.---r
'
(y3-Yl)/2
"-" :
~!
~'
Forces balance in the x andy directions. How about moment equilibrium? L M z ? 0 (-:. . .=;"'
rxyh(x 3- x 1 )[(y 3 - y 1) I 2]- rxyh(y 3- y 1) [(x 3 - x 1 ) /2] == 0
~andy .. J 1J
= y.1 y.: J
0
x32
0
[~]=-
h y3l 2 0
Y12
0
xl3
0
X21
The [H] matrix found earlier can be inverted and the stiffness found:
[k] = [!E] [ G] [D] [ l/J] [H]
3.19
-1
I I I I I I I I I I I I
I I I I I I
LECTURE4
LAB PROBLEM 1:
MEMBRANE
PROBLEM STATEMENT
PHYSICAL MODELING
DATA SET
RESULTS
4.1
IPROBLEM STATEMENTI
A sheet of aluminum measures 1000 mm x 1000 Inm x 1 mm (thick). It has a circular hole of 300 Inm diameter which is centered. An extemal stress of 100 MPa acts on the two vertical sides. Find the most severe stresses .
.. ... ....
....
.....
...
...
....
0
4.2
.... lOOMPa
.... .... ..,..
...
...
_____ ___ __
..
IPHYSICAL , M ODELINGI
The sheet and loading lie in a plane, and will be considered a two-dimensional, plane-stress problem. Buckling will not be considered. The material will be assumed linear elastic, withE= 68,950 MPa and v = 0.3. Strains will be assumed to be infinitesimal, so that linear elastic theory applies. Stresses at the top and bottom of the hole should be high in tension, those on the horizontal axis high in +3 compression. If the plate were infinitely large, the stress concentration factors would be +3, -1.
0-1
Exploit symmetry by modeling only 1 I 4 of 0 the structure. Vertical and horizontal reflective planes of symmetry are cut, and the first quadrant solved. The reflective planes are modeled by greased surfaces--no shear forces.
zero forces zero force.--------- zero zero displacement ~1~
I
t
zero displa9ement
force
.........~ ... - 100 MPa
~~
~-
- 10
..
.,~\
.\
'~~~
1 -:
33333 N
o----o------o
8333 N
The loads are apportioned by a method called "equivalent nodal loads" (studied later). This distribution causes the same amount of work to be done as the original distributed loads. @
4.4
assign output2='p1.f12',unit=12
~rhe
ID ANDERSON,MEMBRANE TIME 5 ~"'-~. .._ , , i?i~-~~- <~l""'Y;~npr:<' ...-h(-,~('f.:)~-~ <..-.., .t<:., <-t.UJt,_,z_ U. SOL 10 1 CEND
t~.- '- . ~..ti ~-"'
-->~
cu~e:
TITLE=PLANE STRESS. PLATE WITH HOLE. ECHO=BOTH DISPLACEMENT=ALL STRESS=ALL ELFORCE=ALL SPCFORCES=ALL SUBCASE 1 LOAD= 50 SUBTITLE= EDGE LOADING OF 100 MPA
GRID, 1,,0.0,150. 0. ,13456 GRID, 2, 1 0.0 1 325. 1 0., 1 13456 GRID/ 3 11 0.0 1 500. 1 0. 11 13456 GRID 1 4, 57.4 1 138.6 1 0.0 GRID 1 5, 1 250.0,500.0 1 0.0 GRID 6,,106.1,106.1 0.0 GRID, 7,,303.0,303.0 0.0 GRID, 8,,500.0,500.0 0.0 GRID, 9 ,138.6, 57.4 0.0 GRID,10 ,500.0,250.0 0.0 GRID/11 1 1 150.0,0. 1 0. 11 23456 GRID,l2 1 ,325.0,0. 1 0. 1 1 23456 GRID,13,,500.0,0.,0. 23456 CQUAD8 1 1,29 1 1 1 6 1 8 1 3 1 4 1 7 1 +A1001 +A1001 5,2 CQUAD8,2 1 29 1 6,11 1 13 1 8 1 9 1 12 1 +A1002 +A1002,10,7
1 1
1 1
1
11
PSHELL,29,13 1.0 MAT1,13,68950., 0.3 FORCE 50 8 0 1 . 0 8 3 3 3 . 0 . 0 . FORCE,50,l0 1 0 1 1.0 1 33333. 1 0. 1 0. FORCE 1 50 1 13 1 0 1 l.0 1 8333. 1 0. 1 0. END DATA
1 1
I I I I I I I
message:
***USER INFORMATION MESSAGE 5293 FOR DATA BLOCK KLL LOAD EPSILON EXTERNAL WORK SEQ. NO.
- 1.7 45 43 13 E- 15 2.2748490E+04
4.6
V E C T 0 R D I S P L A C E M E N T POINT T -ci:l..ST~:l(:t<::::t::rrl : : :_:=t"I.t: T2 1 ID ,.,\. -3.126035-01 .0 1 '<\... ..., .~4. 319586-01 2 .0 ._:._=~ ~: .-: :_:_:~: . _ ~-~ !: _ _ ; \.. .. .. 3 0 3.055194E-01 -2.977902E-01 4 2.896056E-01 -2.985195E-01 5 5.517983E-01 -2.186938E-01 6 7. 5.421658E-01 -1.359472E-01 6.710781E-01 -7.273364E-02 8 7.343332-01 -1.161699E-01 9 3.664567-02 9.142293E-01 10 7.836932E-01 .0 11 12 9 .@ 13 1 . l_J 1 , _) 1 ~ t 0 J 0 dtSplac~enterz.t
\ \_ /)
......... ;...:t
~-
.... __
---
= ..
. . .:_ _. _
.4;~;;?~~?; ~.highestx
STRESSES
IN
QUADRILATERAL
ELEMENTS
(QUADS)
VON MISES
1.346208E+02 2.589606E+02 1.635533E+02 1.105713E+02 4.486930E+01 9.315877E+01 1.655113E+02 9.975479E+01 9.728526E+01 1.300757E+02
3.292986E+01
2 CEN/8.5 6 11 13 8 .5
5.639891E+01 5.461864E+01
1.229330E+02 -1.326844E+01
4.7
T2 .0 .0 .0
3.562568E+03 2.218100E+03 -5. 780667E+03
The horizontal forces should equal the live load. The vertical forces should add to zero.
IMARC DATAl
The MARC solution will use element #26, which is an 8-noded quadrilateral, plane-stress element.
~
4.8
COMMENT THE FOLLOWING CARD SETS ARE FROM LOAD ID 8333.000 0.0 0.0 0.0 0.0
8
67
0.0 0.0 0.0
0.0 0.0
0.0 0.0
0.0 0.0
0.0 0.0
0.0 2 11 12 0.0 13
1 1 2 3 CONNECTIVITY 2 1 26 1 2 26 6 COORDINATES 6 13 1 0.0 2 0.0 3 0.0 4 57.40000 5 250.0000 6 106.1000 7 303.0000 8 500.0000 9 138.6000
6 11
3
8
13
7 12
5 10
2 7
4.9
13 68950.00 1 TO GEOMETRY
1
0.300000 2
4.10
o t a 1
llx
d i
s p 1 a c e m e n t
lly
10
11
12 13
tht~
nodes:
!xy
a l i
z e d
O'y
s t
r e s s e s 3 . 3 6-82 11.484 14.690 -26.165 9.7889 -31.378 -9.0459 2.5393 -23.052 -9.5745 -16.316 -16.881 -16.048
1 2 3
4
5 6 7
8 9
10 11 12 13
118.67 75.915 228.67 90.166 154.03 117.52 115.70 62.853 107.72 18.944 66.343 100.69
44.138 9.6650 8.9647 18.644 4.8227 -25.046 -17.296 -4.9284 -56.917 23.204 13.291 51.291
4.11
IASTROS DATAl
ASTROS data are compatible with COSMIC NASTRAN. Eight QUAD4 elements are used. Stresses are given at element centroids.
3
8
=
:=
Yt
.46
~
9
X
u - - - - - - e : : l: o :: - - - o
11
12
13
4.12
BEGIN BULK GRDSET,,,,,,,3456 GRID,1,,0.,150.,0.,,13456 GRID,2,,0.,325.,0.,,13456 GRID,3,,0.,500.,0.,,13456 GRID,4,, 57.4,138.6,0. GRID,5,,250.0,500.0,0. GRID,6,,106.1,106.1,0. GRID,7,,303.0,303.0,0. GRID,8,,500.0,500.0,0. GRID,9, ,138.6, 57.4,0. GRID,10,,500.0,250.0,0. GRID,l1,,150.0,0.,0., ,23456 GRID,12,,325.0,0.,0.,,23456 GRID,13,,500.0,0.,0.,,23456 GRID,14,,191.2,325.0,0. GRID,15,,325.0,191.2,0.
CQUAD4,1,23, 1, 4,14, 2 CQUAD4,2,23, 2,14, 5, 3 CQUAD4,3,23, 4, 6, 7,14 CQUAD4,4,23,14, 7, 8, 5 CQUAD4,5,23, 6, 9,15, 7 CQUAD4,6,23, 7,15,10, 8 CQUAD4,7,23, 9,11,12,15 CQUAD4,8,23,15,12,13,10 PSHELL,23,33,1.0 MAT1,33,68950.,,0.3 FORCE,67, 8,0,1.0,12500.,0.,0. FORCE,67,10,0,1.0,25000.,0.,0. FORCE,67,13,0,1.0,12500.,0.,0. END DATA
4.13
D I S P L A C E M E N T
V E C T 0 R
POINT
ID. T1 1 O.OOOOOE+OO
2 3 4 5 6 7 8
9 10 11 12 13 14 15
l1ighest x. translation
RFORCE
LOADTYPE(8)
....
-1.98096E+04 0.00000E+00 1 -2.32470E+04 0.00000E+00 2 -6.94344E+03 O.OOOOOE+OO 3 0.00000E+00 4.37528E+03 11 0.00000E+00 -1.88103E+02 12 0.00000E+00 -4.18718E+03 13 6 ENTRIES SELECTED
4.14
IRESULTSI
mesh is shown, using
4.15
/
/
/ / /
/ /
/
/ / /
----~==<
1.13 mm
MSC/NASTRAN
NLt\RC
3,563 4,168
4.16
t.
119 : ;:
136
:. 1o" .. !
88
MSC/NASTRAN
MA~RC
112
::::
::::
::::
19 ~-69
66
101
The highest predicted tensile stress is 2 7 4 MPa. The highest compressive stress is 94 MPa.
COMMERCIAL SOLUTION
,.
finer mesh, with MSC /NASTRAN. This provides a more accurate solution to judge the simple models.
4.17
,..,
00 00
~
00
~
u
t= c
~
ro
111.
145.
120
Cl.i 8
94.0
.
~
><
68.4
42.8 17.2
-8.45
378.
-8.45
418
5?. 8
43.0
max
X
dis pl.
mm
1.1318 1.0862 1.0973
1.1631
Commercial
(MSC/NAS, 1280)
I I I I I I I I I
I I I I I I I I
LECTURES
5.1
IPHYSICAL PROBLEMI
The linear theory of elasticity admits singular solutions (infinite stress) at reentrant comers:
This also can happen when bodies contact, and at abrupt change in materials:
5.2
The stress singularity is characterized by a radial stress in the local region with:
K
G)
O"r=1-p r
-+------r
= 0:
If pis less than 1.0, the stress is singular. If p lies between 1.0 and 0.0, the stress is singular, but the displacement is "integrable." If p is negative, the problem is hopeless (has infinite displacement!)
5.3
Williams, Dundurs, Bogy, Comninou, Lee, et. al. have studied this problem. Dundurs showed that the elastic constants are described by two variables, a and ~ Only a interests us:
a = ( G 1 I G 2 )( K"1 + 1)- ( K"z + 1) (G2/G1)(K1 +l)+(K2 +1)
where:
90
80
en en
K 1
70
=3-4v1 =3-4v2
~ ~
(1)
60 ~50
30
K2
OO"'' 40
No Power Singularity
20
10 0
o
o
o e
Developed by Suh, Anderson, McDonald and Park. (Ref. 1, 2, 3) Independently developed by others in England and USA at slightly later time Also called "biological growth" method Basic idea: a live body will grow in a way to reduce stress (e.g. bone) The cross-section of a member will be scaled up to reduce stress (optimality criterion)
5.4
Consider a 2-D body lying in the xy plane. Locate the principal stress directions at a typical point. A small rectangle of material is isolated, aligned with principal coordinates.
1112
I
I~
illl
=-
.. 1
Jill
Ill~
=I
(jl I (}1
11112
(jl
I (}Io
1-1
The goal of the geometric redesign is to achieve a fully-stressed state in a design region nf. Tractions are prescribed on the rt portion of the boundary; displacements are prescribed on the r u boundary .
.Z
1o .... y
IFEASTUDYI
Consider three contact problems:
0 0
contact of elastic wedge and elastic plane contact of elastic wedge and rigid plane contact of bonded elastic wedge and plane
In each case, we will start with a singular baseline design, and progress to an optimum redesign that has constant stress along the wedge surface. Load will be held constant, and the shape of the wedge is optimized. The wedge/half-space will be modeled by plane strain.
5.6
The baseline design for the elastic wedge and half-space with identical properties yields a singularity: No friction.
"??3.
~--<
71.5.
@
An enlargement of the
lC.?
5.7
Optimal redesign of elastic wedge and elastic half-space with equal properties. No friction.
283 .
271.
284.
1v..:...,
lSO.
13~~-
0 ')
118.
5.8
G4S.
l"'!"f
0.:
~
E .;l?. 587.
~
~ --
...........
b
cr.;
~
~ --
55G .
~:~2~~ .
::/:;
1)
~
t85.
4G!.\.
'~ ~:! ~~1 .
~
~
cc
~
~
.. .:..,)
c.~
4 ~
0-n
~
403.
3?~3,
~~ -t2.
!"""'t
~ i-<
312,
281.
2~1.
crj
::2
220.
An enlargement
5.9
325.
0.:
~
~
32:3. : 320.
4~
l-<
~
300.
29:3.
295.
"i-<
Cj
~
23'3.
2:90.
@
5.10
.....
,.... .
Baseline design for the bonded wedge and half plane. Identical materials. Singular solution.
4.2G
5.11
Enlarg contact . : .: . optimum :. wedge/ plane. Id~ - ~~~-~- ~~m_:: :.::. :.: : : : : : : :: : : : : : : materials-m: -.: : HH:{/{'H::{::-:n-::::::r:::?::::_':::-::;:;-::;:;:;:;:.:-.-.,:,:::::;:;:
:m:-;::: :-: : : :::
281.
238. 214.
-=cs
1.4~.
121~
5.12
ICONCLUSIONSI
The geometric strain method has been shown to be useful for redesign of contact problems, where the goal is to eliminate stress singularities. Cases presented were plane strain studies. Specific optimum shapes are found for each elastic wedge/contacting body problem. These shapes may be useful in certain design situations.
5.13
I I I
I I I I I I I
I I
LECTURE6
INTERPOLATION
OVERVIEW
A. DEFINITIONS B. EX~LESOFnnERPOLATION C. SOLIDS, SURFACE, CURVES AND POINTS D. TRICKS FOR CREATING SURFACES FROM CURVES
FUNDAMENTAL THEOREM OF ALGEBRA PASCAL'S TRIANGLE INTERPOLATION IN FEA DISPLACEMENT FUNCTIONS 1-D SHAPE FUNCTIONS
A. LINEAR INTERPOLATION B. QUADRATIC INTERPOLATION C. CALCULATION OF SHAPE FUNCTIONS
REFERENCE
Hamming, R. W., "Numerical Methods for Scientists and Engineers, " Second Edition, McGraw-Hill, Inc., 1973, pp. 332 and 329.
6.1
IOVERVIEWI
A. DEFINITIONS
Interpolation: The process of estimating a function in the interior of a region bounded by known values of the function.
f(x)
I
/
~
X
-- a
~~~
? 1f(b)
f(x) Extrapolation: The process of estimating ~ functional values exterior / jf(a) f(b)f to a region of known --+--li-------:----___,.__ X values. a b
------
6.2
Interpolation can use (Hamming): o function values only o values of derivatives o differences of function values e arbitrarily placed samples
Uses of interpolation: o fit experimental data o exact fit to polynomial o approximate fit to higher degree polynomial or transcendental function Interpolation and numerical integration are closely related:
B. EXAMPLES OF
INTERPOLATION
6.3
One equation One linear equation Two simultaneous eqns. Two linear simult. eqns. Three simultaneous eqns. Three linear simult. eqns.
surface (usually) plane (always) curve(s) (maybe) str. line (usually) point(s) (maybe) point (usually)
(~)
6.4
2-D (PLANAR)
SPAC~
y
\ \~'
~
=0
--"' \
= Cf 1 (x,y)
This creates a surface \ x passing through the curve in the xy plane. Given several equations of lines in the xy plane, set z = Cf 1 (x,y)f 2 (x,y) This surface passes through both curves.
6.5
--~
...-~~
'-,
~
FUNDAMENTAL
THEOREM OF ALGEBRA
A polynomial p(x) has exactly n roots (zeroes), counting repeated and complex roots.
p ( x ) = qo + qlx+ q2x + q3x +. +qnx
2 3
n
\
~.:.
\ j
..f
I
X
\ .J
~=
'::\~~
:c~..
,/
-+----c~
,,::::.::.:.':: .)
\ .....,.._,..<",,/
IPASCAL'S TRIANGLE I
In two dimensions, one can group the polynomial terms in a triangular pattem:
1
X
x2
.xy
y2
x3 x4
x5 x4y x3y
2 xy
x3y2 x2y3
6.6
xz 2
X Z
yz 2 y2z xy2 y3
x3
x2y
INTERPOLATION INFEA
The "displacement method" in FEA either: assumes the strain field {} (Tumer), or assumes displacement functions [], or assumes shape functions [N]:
{!}
[G]
{o}
=
[D]
[ ~]
[H]
=
-1
{e}
{u} = {q}
{ue}
~ '==rNlA
nodal forces stresses strains
[B]=~ general. displ. coord.
6.7
nodal displ.
A. LINEAR INTERPOLATION
Linear interpolation is used for trigonometric and logarithm tables.
0
Find sin 24
Extrapolation: o 10 20 3o 4o so /(24) =/(20) + 0.4 [/(30)- /(20)] Interpolation: f(24)=0.6f(20)+0.4f(30) The coefficients (in red) can be interpreted in terms of length ratios:
f(x)
6.9
f(x) = : ;~ :
1 '---y---J
1V1(x)j(x 1) + l\f2(.x)j(x2)
@)
N(x)
f(x)
6.10
B. QUADRATIC INTERPOLATION
To "fit" a parabola, one requires 3 points.
f(x) = N 1(x)f(x 1)
+ N 2(x)f(x 2 ) + N 3 f(x 3)
f(x)
N(x)
~(X ). ~ N 1 (x)1 , .. 1
x 11
x?.
/ x3
6.11
The shape function must be unity at the home node and zero at other nodes.
N1 (0)=1 N1(L/2) = 0
N 1(L)=O
which can be solved for the constants a, b, c. This approach applies to elements with many d. o. f., but involves large sets of equations. @
PRODUCT FORM
One can exploit the zeroes of the function by using the product ("factored") form of the polynomial. For the same rod element:
N 1(x) = C(x- L/2 )(x- L)
The zeroes are "built-in," but we must find the amplitude scale C:
N 1(0) = 1 = C(-LI 2)(-L)
1
:
~~--.,,_<~'<-.
.......
....,.1'.;.-.~ .
~- e=~
.o),o. -
C=21L
L
12-DSHAPE FUNCTIONS)
The constant-strain triangle is straight-sided, and can be solved by a product form which causes zero functional val~es along lines.
ft(x,y)=y=O
b f2(x,y)=y-b+ax=O
co,br-3 Y,o
0
ty
'f..~
Q.:t-~
f 3 (x,y)=x=O
(0,0)
1
Q .;: -
y =0
2 (a,O)
b =C(y-b+ax)
We must scale this function to pass through unity over 1 node 1: N1(0,0)=1 = C(O-b+ ~0)
C=-1
b b N 1(x,y)=-J; (y-b+ax)
IN1 (x,y)=l-~-t I
Likewise:
N 2 ( x, y) = Cx N 2 (a,0)=1=Ca
C=l a
X
6.13
Finally,
N 3 (x, y) = Cy
1
N3 (0,b)=l=Cb
C=l
b
The eight-noded quadrilateral is another example: We will develop shape functions as we need Q them. We will typically use the product form.@
IPROBLEM SESSIONI
ProbleDl 1. Six-noded triangle
Propose a shape function N 4 (x,y) for the plane strain triangle shown. The midside nodes are located exactly at the midpoints.
y
6.14
Shape functions for u(x,y) are the same as for v(x,y). Use shape functions repetitively, i.e., N 4 (x,y) serves both u and v. Consider the displacement function:
2 N4(x, y) = ql + q2x + q3y + q4x + qsxy + q6i
It takes at least a complete quadratic to pass through the 6 points needed to form N 4 (x,y).
~=~:~::::~~:::~::~::::::::::::
Finally,
6.15
ProbleDl 2. Plane stress square Consider the plane stress square with the special coordinates shown. Find the shape functions for it.
(-1,1) .: 4
Y~
3~
(1,1)
(-1,-1)
Solution Four nodes in two dimensions require a quadratic form, though probably not a complete quadratic (only four generalized coordinates). Use the product form.
c(
@.
1-y=O
\;'ll
4
0
C=! 4
NI(x,y)= !(1-x)(l-y)
II
L
!:-..,
~
I
~
II
1 1+y=O
2
~:-
By similar reasoning:
N 3 (x, y)
N 4 (x, y)
=! =!
(1 + x)(l + y)
(1- x)(l + y)
ProbleDl 3. Constant strain triangle Develop the shape functions for the threenoded, constant strain triangle shown.
y
(2.,2.)
-::
(1.,1.)
(3.,1.)
~-----------------x
Solution Establish the equations for the lines that bound the element.
(0,4)
= 1 = C ( 4 -1-1)
C=l 2
><~ ) -~
1'1..3
"
l tx,y=2
;;.....,.......,..,...
rN-c . ) .. ic4. .. . ..
-x-yl
N 2 (x, y) = C(x- y)
N2 (3,1) =1=C(3-1)
C=~
,~~~~.
= C(l- y)
:-.w~~.w~N'_..,.,.,._..,.,.....,.._.
.....,..._....,...,.~._.__.,."'JY'~~
6.17
I I I I I I I I I I
I I I I I I I I
LECTURE7
Stress-Strain Law Strain-Displacement Law Equilibrium Position Virtual Displacement Work Virtual Work
D. VIRTUAL WORK THEOREM E. MODIFIED VIRTUAL WORK THEOREM
POTENTIAL ENERGY
A. VARIATIONAL OPERATOR B. ENERGY-CONSERVING LOADS C. LINEAR ELASTIC MATERIAL D. POTENTffiLENERGYTHEOREM
Comments
PROBLEM SESSION
1. 2. 3. 4. VIRTUAL WORK FOR SPRING POTENTffiL ENERGY FOR SPRING VIRTUAL WORK FOR ROD EXISTENCE OF FORCE POTENTIAL
7.1
!VIRTUAL WORKI
A. MECHANICAL EQUILIBRIUM
There are three interchangeable statements of equilibrium: Newton's laws (principally F=ma) Hamilton's principle Virtual work All are for statics and dynamics. We are interested only in nonrelativistic, nonthermal problems.
7.2
.., }'"'
B. SKETCHES
Elastic problems (including nonlinear cases) are govemed by 3laws: Stress-strain (constitutive) Strain -displacement Load -deflection One should sketch the general properties of these relations in every problem.
(j
~~ds/
u
C. DEFINITIONS
Stress-Strain Law Nonlinear elasticity:
(j
a geometric law can be hardening, softening, or can bifurcate Green's strain tensor often used to characterize
/)~(~
CD-~_../'//
,<_,.,..~-
c
0
Equilibrium Position There is a unique unloaded position I. The static loads are applied and the body moves to the equilibrium position II.
7.4
Virtual Displacement
o
an infinitesimal displacement away from the equilibrium position, from state II to state III must be consistent with geometric constraints causes increment of strain
F
e
o
Never say: ~-~Work= (Force) (Displacement) and then increment: L\Work = (Force)(L\Displ.) + (L\Force)(Displ.)
~
Real work
complementary work
The wrong way to define work creates ficticious work, caused by a change in force. @
7.5
Question: What is the "Misplacement" in the definition of work done by a force on a body? Answer: The displacement is that of the body. A force can do work on a body without the force moving! A person can stand by an escalator and squeeze the moving rail. The frictional force does work on the escalator, but does no work on the man (the man does not move). The energy from the escalator's motor is released in heat. This is typical of sliding forces. Work depends on the coordinate system.
Virtual Work Virtual work is the work done by external and internal forces during a virtual displacement (II ~III). Work done by external forces:
L\
wext = LFi
L\ui
/I--?III
II II--?III
@
7.6
+ LlWexternal
[[~[[[
=0
If the forces are equal, then no net work is done if the door rotates (infinitesimally). @
energy conserving. The system is elastic, but not necessarily linear. One can then show:
Ll W internal
[[~[[[
- Ll U
[[--?[[[
This is put in the virtual work theorem, repeated here: Ll W internal + Ll W external - 0
[[~[[[
[[--7[[[
to yield:
-LlU
[[~[[[
~ W external
[[~[[[
7.7
@) I
~----------~--u
.du
J (stress) (d strain) dV
V II
II~III
_LF. du.
l l
II
II~III~
!POTENTIAL ENERGY I@
A. VARIATIONAL OPERATOR
We now change
our viewpoint from an incremental to a variational operator notation. The increment was ~ , whereas we will now use an operator 8 . Both act only on displacement-like quantities.
~u(x)
~----~--------X
7.8
B. CONSERVATNELOADS
Restrict the class of problems to linear elastic systems with energy conserving extemalloads. Energy conserving loads can be derived from a potential field: {F} = -{V} 'J1) The gradient operator {V} is an ordered set of derivatives with respect to the underlying displacement coordinates. In finite elements, this means with respect to nodal coordinates:
a
a
dUn
= (Forces) 8(Displ)
= 8[(Forces)(Displ)] = 8(- 'J1})
where
II
'J1): -(Forces)(Displ)
II II
'J1/is the "force potential" or "work potential." Forces in Ware constant (in this example). The operation s'J1/varies only displacements.
7.9
= f(stress)(!lstrain)dV
V II II --7 III
= f(stif.fness)(strain)(!lstrain)dV
V II II-7III
=
=
8(1 fcstif.fness)(strain)(strain)dV) 2
V II II
\....._
_./
""'V"
fMJ. . . -:5([1)1
II
)
I II--7 III
W'J=O
II
8(U + '/11) = 0
II II
"potential energy"
IT=
II
U+U'
II II
7.10
8(1})=0
I
I
lThe Pote~ti~-e~f;_l~a~ticl
Isystem exposed to conservative loads is
1 "stationary"
I
L..-~,.-.:
~
1
...,.,..,..-.~~w.''""""'"''",.,.,...,......,,.._..,.."'-."'...,......_.N.w.~v..v..,..-'._...,.......,.....,....,..... . ,...,....w..,.....,.,.........w..-..".-...........-...........................:~
Comments
. ........
The method can be extended to dynamics and for nonconservative loads. The method is useful for convergence proofs.~
@
IPROBLEM SESSIONI@
ProbleDl 1. Virtual work for spring Consider the simple spring with load at end, as shown. No gravity acts. Use the modified virtual work theorem to prove that F=ku without making that assumption in the derivation. The strain energy at any displacement u is of the form
U(u)=-ku
2
In particular,
U(u+t\u)= ~k(u+L\u)
7.11
2
~
~u
II--7 III
"'-----v---'
II--7 III
III 1
II
2 1 2
=02+~u+~kju2-,th>
~ '
F~u = ~u
higher order
(F-
ku)~u
=0
~N~Jit.YoN..............,,..,...J
/@
ProbleiD 2. Potential energy for spring Consider the simple spring with load at end, as shown. No gravity acts. Use the potential energy theorem to prove that F=ku without making that assumption in the derivation. The strain energy at any displacement u is of the form 1 2
U(u) =- ku
2
fJ1) -Fu
This can be checked by
F= -V 11/:
~
~t u
-1u
(-Fu) = F
7.12
11=U+11/
11 =~ku
- Fu
811 = 8 C1 ku2 -
Fu) = 0
Because Bu is arbitrary:
L. . - .....
~
F-ku
il
Sketches of the energy balances are shown. These are typical for a single d. o. f. structure. The potential energy is negative at equilibrium.
7.13
free end, as shown. The body reaches equilibrium. A subsequent virtual displacement of -0.1 mm is then imposed. a) How much total virtual work is done during the virtual displacement? b) How much virtual work is done by the external loads? c) How much virtual strain energy is created?
EA/L = 100 N/mm
f2
Solution a) The question is answered exactly by the statement of the virtual work theorem:
~Wext +~Wint ==0
II-:; III II-:; III
Ill II
= !2 ~u2
II II--:; III
~u II--:; III
=-10 Nmm
Nmm
/@
7.14
ProbleDl 4. Existence of force potential @ Does a force potential exist for the case of the wing shown, when acting with a single degreeof-freedom (pitch) and when acting with two degrees-of-freedom (pitch and plunge)?
~Sa
z
77>77
Single d. o. f.
Two d. o. f.
Solution For the single degree-of-freedom system, the relevant force is pitching moment M, and the relevant displacement is angle of attack a.
M=eL
= eCL qSa
a
fa U{a)
Choose Check:
7.15
For the two d. o. f. system, the relevant force @ and displacement are: { ~}
1} {
{1} . - ~a
dz
'ffla,z)
-~eCLaqSa 2 +
f(z)
U{a,z) =- JL(a,z)dz +const2 = -eCLaqSaz+ g(a) No force potential exists because: - ~eCLaqSa + f(z) =1:- -eCLaqSaz + g( a) The force potential is "not integrable."
2
\..---v----J [ KA]
The aerodynamic stiffness is a "circulatory" matrix, and can lead to flutter. Energy can be extracted from the airstream to cause growing oscillations, until failure occurs.
7.16
LECTURES
GENERAL DERIVATION
A. B. C. D. E. MOTIVATION DERIVATION FOR CONCENTRATED FORCES DERIVATION FOR DISTRIBUTED FORCES DERIVATION FOR INITIAL STRESS AND STRAIN COMMENTS
ROD(TRUSS) ELEMENT
A. STIFFNESS B. EQUIVALENTNODALLOADS
PROBLEM SESSION
1. 2. 3. 4. CONSTANT STRAIN TRIANGLE EQUIVALENT NODAL LOADS FOR ROD STIFFNESS OF HYPOTHETICAL ELEMENT FOUR-NODED TRIANGLE
8.1
IGENERAL DERIVATIO~
A. MOTIVATION
Most finite elements are statically indeterminate. Energy methods work well for such redundant structures. Some problems to be considered later will be nonlinear. Therefore, choose the virtual work theorem. We will predict the displacement field better than the stress field.
8.2
=(8u) =8(u)
II~III
II
f.1
{!}
[G]
[D]
[<l>]
-1
[H]
{q} <=
{ue}
[N] [B]
8.3
I{8e{ {o"}dV = 0
V II-7III II
. nodal dzsplacement
= [B] { u} . .=~:~~-~---
{8} = [B]{8u}
{a}= [G][B][u]
II~III
l II
II
II
[k]{u}= {/}I
..
:....~-......,..,.....~~""""""""'""
::
8.4
f[B]T[G] [B] dV
v
1 1
=0
J s
J
v
-J {De{ {a}dV= 0 v
Relate to nodal coordinates by using:
{8u} = [N]{8u}
{8u{
dV{u~= 0
~
J s
J v
\..--v--/
[k]{u}={f}ext + {f}enl
.
{f}enl+
'I
{f}enl
X
8.6
D. DERIVATION FOR
INITIAL STRESS AND STRAIN
Prestrain -- Thermal stress -- Artifice in nonlinear cases Prestress -- Artifice in nonlinear cases Redo the derivation, using the general stress-strain law:
{ o"} = [Gl({t:}-{t:o}) + { O"o}
BU = I{Bt:{{O"}dV v
BU
enl L
where
enl T
co
enl
ao
~---~""""""'~V..W>.~W.V.~..,.._YN.NY.~V..WNWW,o0AN'oW.VN.MN'o~
j
~
co
{/}~ =-~B{{O"o~--~
8.7
E. COMMENTS
o
o
5 equivalent nodal load types. commercial codes automatically do the integration for many types of volume, surface, line and thermal loads. e. n. 1. concept replaces lumped load concept.
e
e
o
e. n. 1. are also called "work equivalent loads." c0 is used more often than a . 0 automatic integration of e. n. 1. sometimes assumes that "midside" nodes are placed exactly at midside. Do so, as default.
Evaluate:
[k]
= I[N]
v
[D] [G][D][N]dV
{/}
enl
L
frN{{.} dl
1
where:
{(x)} = 0(x/L)
8.8
Hence:
N 1(x)=l-x/L
Likewise:
N 2(x)= x/L
{u(x)} = [1- x/L
x/L ]{~~}
{Cx} = [!}u(x)}
This means:
[G]
=[E]
[B]
[D]=[f.J
x/L]
= [D][N] =
[t}l- x/L
l/L]
8.9
=[-l!L
A. STIFFNESS
[k]
f [B{[G] [B] dV v
1 1
[k] = {[- /
f }E][-l!L
2 2
1 /L ]Adx
EIL]dx
EIL =A o -EIL2
2
-EAIL EAIL
1
-1
1 -1]
J[l- xiLILJ{.o(x/L)}dx
X
J o
L
J(x!L-(x/L) 2 )dx
{/}
enl
L,
= .o o
f(x/L)
3L
2 dx
=..Co 2 L
1 x 2 1x 3 L ----2
0
1 x3 L 3 L2 o
=..Co{~~~}
= .0
L{1 I 3}
2/3
1/3
D&a(o
2/3
total force
IPROBLEM SESSION I
Proble111 1. Constant strain triangle
A unit displacement is given in the horizontal direction at node 1 in a right triangle plane stress element. Find the strain Ex .
y
(O,b)
1 (0,0)
~ -=-,,.,
-----4::.--- X
8.11
Solution
The displacement is in the form of a shape function. That is seen to be: y X
(O,b)
1----==0
N 1(x,y)= C(l- b- a)
1
(0,0)
= Ex):
(i==l-3)
e ==
l
j=l k=l
L L DNkuk J
1:
2
= L DuNj/ 1)
dx 0 [D]=
[N] = [ 1_y_x b a 0
8.12
.J
@
= ( - )(1- - - - ) x dx b a
(}
- --
1 a
The result is constant Ex strain over the element. This illustrates the concept of a constant strain element.
v1
Proble111 2. Equivalent nodal loads for rod Find the equivalent nodal loads for the load cases shown for a two-noded rod. @
Lol=:_:~l
(0
0)
(0
.
L
Ll
1-----L/2=1
...
1-
~L/3~
o)
Dimensions on L 0 = F/L and on L 1 =F. Hint: For the concentrated load use the Dirac delta function 8(x-x0 ), which takes zero values away from the reference point x 0 , but has the unit integral: r~ d
Ju(x-x 0 ) x=1
-oo
{f}enl =
L/2
t [1 ~;{L]
x-X
Ladx
x2 L
2L L/2 2L
2L L/2
_ . {
0
L/8} 3L/8
co
(.........
0)
= LgL{~}
f {f}enl=J
0
[1-x/L x/L J
0
1 8 (x-2x/3)dx
L1
11
e)
=1
{ 1/3} 2/3
(G
.. 0)
8.14
[Hf
=[-~0 g ~] 1 -1
[ l/J] = [1 y X]
[DJ=[-fx]
[G] = [Q]
The element volume is V 0 . a) What is the stiffness matrix for the element? b) Is there a rigid body mode present? @
Solution This problem must have three d.o.f. because [H] is a 3x3 matrix. Both x and y are involved, so the problem likely has a scalar field variable and has 3 nodes. There is a "rigid body mode," as seen from the [<1>] matrix, which has a constant term. The shape function matrix is:
[N] =[ l/J][H]
=
-1
[1 y
0 0] 0 x{-i 1 -11
@
= [1- y X y- X]
8.15
=[D] [N]
=[fxJl-y
= [0 1 -1]
The stiffness matrix is:
[k]
y-x]
=I [B{[G][B] dV
v
dV
=I[ ?][o
v -1
Q -Q]dv
0 Q 0 -Q OJ dV =I [0
v 0 -Q
[k]
= Yo Q
0 0 0] 0 [0
1 -1 -1 1
(0,1) .;
.;: (1,1)
8.16
Proble:m 4. Four-noded triangle A plane stress isosceles triangle has been proposed, including an intemal node.
a) Will the shape functions be linear? Why? b) Propose the simplest possible set of shape functions.
3 (O,b)
c) What are the equivalent nodal loads at node 3 for a constant body force Tx? (T y- 0.) Dimensions on the load are force I area. d) Is the element conforming (displacement compatible with neighboring elements)? Why? Solution a) The shape functions must be quadratic rather than linear. A linear polynomial in 2- D has only 3 coefficients, and the 4-noded element has 4 d.o.f., requiring 4 coefficients.
N 1(x, y) = Cx(x- a)
IN1(x, y) = ~
2a
2 (x- a)
1
~
I
2
N 2 (x,y)= Cx(x+a)
x+a=O
IN 2 (x, y) = L"-~
~~
2 q, ~y~----~J
2 (x+a)
I
8.17
"
N3 (x, y) = Cy(y- b I 3)
N3(0,b) = 1 = Cb(b- b/ 3) =
c 2 b2
3
y =b/3
3 [;3(x,y)= ~ 2b
;r
~
Y2 ~b/;i
y=O
~ ~
N4 (x, y) = Cy(y- b)
N4(0,bt 3) = 1 =
c~ (~-b)=- c~ b 2
y=O
~~-~-
~~------~~-9.y-~~ - -- - - ~sl
!. .
~ N4 (x,y)=- (b-y) ~
..
N3
N4
N4
OJ
c) The equivalent nodal loads are basically given in terms of an area integration:
{f}
enl
AREA
JJ [N]T{Tx(x, y)}dAREA
Ty(x, y)
tf6 .,. fs
In summation convention:
~= fJ
AREA
}:_NjiTjdAREA (i=1-8)
j=l
=0
~ = fJ
AREA
N li 1'1dAREA
f5 =
JJ N
15 (T)dAREA
AREA
f6 =
8.18
If N 16 (Ty) dAREA
AREA
3v ( y - /J I 3) J . .
[N]= [
(2x8)
b a-aylb
2b2
fs=J
y(y-~/ 3 )(Tx)dxdy
2b
0 -a+ay/b
Txab 12
TxA
12
d) The element cannot be conforming because each side is controlled by only two nodes. The displacement field is quadratic and the diplacements on a side cannot be uniquely 1 determined by the information on that side. v
8.19
I I
I I I I I I I I I I I I I I I I
LECTURE 9
BEAM ELEMENT
9.1
9.2
'
~~
The Euler-Bemoulli-Navier approach for flexure concentrates on the deflection of-the centroidal axis under the assumptions: o linear, elastic material e infinitesimal deflections and slopes o plane sections initially perpendicular to the beam axis remain plane and perpendicular to the axis after deformation. e stresses and strains do not depend on z. These assumptions are accurate for beams with length/height ratios of 10:1 or greater. One can analyze thin-walled tubing if the cross section is closed and not more than 2: 1 in depth/width.
B. PHYSICAL MODELING
End moments bend the neutral (centroidal) axis to a circular arc. At a distance y above the neutral axis, a fiber undergoes strain: e (x ) = l(y)-1(0)
X
,y
=--
y R
9.3
From differential geometry, the exact expression for curvature (positive upward) is: y . . , . --.,,..,.,.... 2 l/'\R d v
1
= [l+(tJ]~
dx
J)CJ
f'
~------- x
R-x 2
dx
The cross-section of the beam is so small that the stresses shown are negligible. The case for a y is not obvious, since the upper and lower surfaces <~an be loaded. In practice, howeve~, cry is typicall:y 100 times smaller than crx.
y
z ~~) x
t
l
9.4
: rxy ~
~
~.,...._.~
=:::::
~~
: 1
..
XY!
(bottom
view)
Therefore:
Five components of stress and strain are zero. The strain energy density
[J = ~ [CYxe x + ,/yey + 9'zez +
0 0
reduces to
U =-(Jxx
1 2
All of the energy stored in the beam is in uniaxial stress and strain.The beam acts as a bundle of fibers, maintained in relative position by shear forces, but "locked" in shear strain. The only stress-strain relation of interest is:
rO-~"<-O"'-'>---.--. _,~_,.~.-oo>O<oO<oooo.-~<oO<.<o>O<OOO<~>"--~>Os
. . ..
ICJx(x,y)= Eex(x,y)
9.5
A. C<)ORDINATES
For a un:i form beam with end loads:
EI d v(x) =0 dx4 When integrated, v(x) will be a cubic, with four constants of integration. The finite element must have 4 nodal degrees of freedom:
V
4
AA
VB
A
j
co
. r~"~eA ! ~
_...,.,.....)
i ....,.~$-~
o) j
__.,..,./',/
e
B
0
"rzx
Ex Ey ez 0 Yyz
y,z)}
ql q2 q3
q4
eA VB eB
VA
Because energy:
c~nly
Jx J{v( x)}
@
9.6
2 2
~:
MB
VA
[E] {ax}
[-y 1:2]
<=
[t/J]
q 1 [H]-
VA
<= {ex}
{v(x)} <=
It
:~
q4
<=
~;
(JB
~[N:]~
[B]~
ql
=[1 x x 2 x 3 ] q 2
q3 q4
=v(O)
VB=
v(L)
(JA=dx(O)
dv
(JB
dv =dx(L)
(}A
VB
(}B
1 0 0 0 0 1 0 0 1 L L2 L3 0 1 2L 3L 2
[H]
ql q2 q3 q4
@
9.7
@
3L - L 2 -2 L
[Hrl=_!_
L
0
-3L 2
-2L2 L
0 L3
0 0
0 0
B. SHAPE FUNCTIONS
Shape functions can now be found as an intermediate step to finding stiffness:
[N]= [l/J][H]- 1
1.0 .
The amplitude of the rotational shape functions N 2 and N 4 depends on the length scale and has been arbitrarily chosen here (L = 6). @
9.8
B. STIFFNESS MATRIX
The strain matrix becomes:
[B]
L2
'
Stiffness:
[k]=
0 area
t If
naturally involves the term which we identify as the moment of inertia (second moment of area):
I=ffy2dA
area
~~~---~,--~,--~~--~~--~
~-------(3-9.3-7-in-)--------~1
E/=10 10 N mm2 1000mm
V=0.3
= J[N{{lo}dx
0
5000
5000N
833000 ( Nmm
10
5000N
01 )
833000 Nmm
We must solve: 6000 6000 12 -12 6 6000 4x10 -6000 2x10 6 10 -12 -6000 12 -6000 6 6000 2x10 -6000 4xi0 6
-
0
()A
eB
VA
One partitions out the 1st and 3rd equations and sets them aside to be solved for reactions later. The 2nd and 4th equations yield:
{ 8
A} = { 0. 0417 } rad
-0.0417
(JB
{~~}={-~~~~}
2.4o<r
L_ .~-----------5000 N
9.11
1"'2.40
@
5000 N
IE~I'ENSION TO 3-DI
Include bc~nding about both principal planes, torsion artd axial extension:
4yelem
1
I u5 t
u2i
=
ll
tLx
z elem
/2 /6
!s f 12
Displacerr.tents involved are u 3 , u 5 , u 9 , u 11. The coordinate system is awkward in this plane, leading to an inherent conflict in the finite element and classical beam sign conventions.
dw(O)
lis=-
dx
__ dw(L)
u11-
dx
115
~~;.;~::.W
l*:f"j
~z
0. :
u ------.-....EE-e i.: ~ X
11
-::::i
v-..-.
9.12
-6L
u3
2L2
u5
Ug
U11
6L
4L
2
6L
ocze
;:.
.II!
oee::
. A .t:<<>. . . . . . . . .
~.
EA[ 1-1]{ul}={t L -1 1 u J
7
1}
TORSION
_.A:;:::-~$~
...,.,.,..~->
The beam. element requires an element coordinate system to locate the principal axes of interia.
NASTRAN codes use two planes defined by the shear center axis and a user-defined vector {v}:
z
elem
Yelem
'~X y
'
9.14
DEEP AND
THIN-WALLED BEAMS
DEEP BEAM
'( . . ... '"'
' (YX/0"~-:hX<O
IPROBLEM SESSION I
ProbleDl 1. Cantilever with end deflections
A cantilever beam is forced (by end loading
not shown) to take a tip deflection of +2 mm and a tip rotation of +0.1 radian. What is the shear force generated by the wall on the beam? Show its direction.
lOOOmm
-1
~-t=~=~::::::~:.I.~
9.15
0.1 radian
u4
6L
c1 u3
L3
11
11
r1
I'
= -2 400 + 60 000
' '
Consider the cantilever beam and rod as shown. They are (~onnected by a pin. A 1.0 00 N load acts at 45 degrees to the pin. Find the horizontal deflection (only) at the pinned joint.
mm
9.16
Solution There are 7 important d.o.f. in the problem. Beam flexure in the xy plane has a 4x4 stiffness matrix. Rod extension has a 2x2 matrix. Four of the d.o.f. are constrained at the walls. The fifth and sixth equations uncouple.
xbeam .
EtA I
LI
0
(12Et/1
LI3
0 0 0
+ E2A2)
0
_ 6Etli L 12 4Etli LI
L2 6Etlt L 12 E2A2 L2 -
0
L12
ll4 us ll6 0
0 0 0
F1 F2 F3 707 707 0 F1
(121/1 + E2A2)u
L13
6E1/1u = 707
6
L2
+4Elh u - 0 - 6E1/1 u L 2 5 L1 61
(300,000+125,000)u 5 =707,000
:~:=~{
;;:j;?'
beam ,./ /
rod //
~~
1 u5
9.17
= 1.66 mm ;
:
---~__.._.,-
A cantile,rer beam is supported at its tip by a spring with constant k = 10,000 N/mm. What is the tip deflection under a load of 5000 N?
yt
..-- - -- - - 2 000 mm- - - 10,000 N/mm
2
E = 68,000 N/mm
Izz= 2x10 mm
Solution The full system has 5 d. o. f. Three of those are "grour1ded" (zero) and their equations are partitioned out. ,.----1 u u _Tu 4 3
1
u2 (._.: .: ;.;,... 0 ~
c >:=~::;:...
usf 6
0 0
t
!.i. . J
1
:.............................._...,_.!
F2
-5000
r..::.E.;:::J
9.18
[l~Ji; k
f[:]{::} ={-5~00}
10000 -20,4oq]{u3} = {-5000} [ 20.4+ -20,400 2. 72x10 u4 0 Use Cramer's rule: -5000 -20400 0 2. 72x10 7
10020.4 -20400 -20400 2.72x10 7 =-0.500 mm
9.19
I I I I I I I I I I I I I I I I I I I
LECTURE 10
CASESTUDY: AUTOHAULTRAJLER
INTRODUCTION
10.1
[NTRODUCTIONI
Open frame trailers carry new automobiles from factory to dealership. Most trailers are welded from cold-rolled tubular steel. The "low boy," trailer is made of hot-rolled structural steel shapes. A major manufacturer is Traffic Transport Engineering of Romulus, Michigan. Their presider1t is Bud Stuart.
It was proposed to modify an existing design for use <ln rail cars, in a "unit train" running from San Francisco to Los Angeles and back.
10.2
The trailers are attached to truck tractor or to a rail car by a "fifth wheel," involving a "kingpin" and a suitable pivot point on the truck or a stanchion on the rail car.
00
00
,f16o
00 . 00
''Low-boy'' trailers.
G)
ffi
~tracks_)
!]
Each side of the trailer is statically determinate. If one pin is pulled out, the structure collapses. (The driver is not asked to drive two pins because of misalignment problems.) Automobiles are raised by hydraulic jacks:
t hyPraulic Jack
Once in place, the hydraulic jack is released and a pin carries the entire load, during travel.
10.5
The rear suspension was first modeled with a rotatioil constraint, but later was modeled as a "whiffle-tree:"
The trailer travels "over-the-road," is lifted onto the railcar, and travels on the railcar. Loads are given by American Association of Railroads, Standards: "Field Impact Tests of Loaded Open-Top Cars." These standards require a test of a loaded car by rolling it into a parked gondola car full of lirr1estone rock, at 4, 6, and 8 mph, (1.8, 2.7, 3.6 m/sec), forward and backward. For design purposes, static equivalent loads were chosen for the lifting and railcar loads: load case support system load direction
1
lgJ lgJ
2g~
2
3
C. RESULTS
Sidewall deflections for the comer lift:
II
\-----r~r--
1 I
aMAX
10.7
= 18,300 psi
Belly droop
= 0.5"
-\
D. CONCLUSIONS
Sidewall deflections are acceptable, for both lifting and specified loading on the railcar.
A high stress region in the front of the trailer
10.9
KINGPIN ANALYSIS
RAIL USE
A. PI-IYSICAL MODELING
assembly kingpin
The head of the kingpin is plate-like and the shaft is beam-like. Loading c~ases are specified by: American Association of Railroads Standards "Field In1pact Tests of Loaded Open-Top Cars"
and
SAE Jl3~3 "Fifth Wheel Kingpin Performance" The critical (SAE) load was 210,000 lb r::::::::;=r=-= (934,000 ~N) at the kingpin tip. _..,_JJ.
10.10
JJ
@
~
\I
'
/
~
,/
J
-"'""
~-
.. 105,000 lb.
I~"""= ~.X'J
10.11
The sec,ond design involved a longitudinal reinforc: i ng bar over the head of the kingpin.
20,000 lb
l_.
--
t. .. '
\I
'
y
~
\
~
....... ~
"
/
85,000 lb.
......-
I I I -.
"
105,000 lb.
"
..
I I I
I I
10.12
tz
r-- 3"~
1"
3"
cover plate
1"
3"
thick plate
tx
Cover Plate
I -1-
10.13
steel tube
I i
!
::::::;:::;::::;:;:::;::::J
I
I
_I
plate doubler
1:::::::::::::::::::::::::
I
apron plate
!
l
-~- kingpin
C. CONCLUSIONS
The redesign of the kingpin area was needed. Design
#4~
it is a very stiff structure embedded in a flexible structure. (A new design for the kingpin has been relea.sed by the manufacturer in which an integral flcmge is added.)
l
~
10.14
LECTURE 11
PROBLEM STATEMENT
PHYSICAL MODELING
DATA
RESULTS
REFERENCES
1. Oden, T., and Ripperger, "Mechanics of Elastic Structures," 2nd Ed., McGraw-Hill Book Co., 1981, p. 228. 2. Roark, R. J. and Young, W. C. "Formulas for Stress and Strain, 5th Ed., McGraw-Hill Book Co., 1975, pp. 96. (Deflection at tip of beam: 1.031 mm).
11.1
ROBLE
f
j
STATEMENT
Find the stress an deflection in a steel cantilever beam with rettangular cross section:
5000N
(1124lb)
~ 2000 N
1
1 1
449. 6 lb)
~-y 150mm
~
IPHYSICAL MODELINGI
The beam is slender, with length/height= 30 (need 10 or more to qualify) . Use Euler-Bernoulli theory: no shear flexibility. Principal planes of inertia are xy and xz planes. Loads are in xz plane--displacements will also be in xz plane. It is difficult to predict whether the tip will move up or down. Guess!
G)
v = (1,1,0)
z
b astc coordinates
4.
y
X
GA
11.3
GB
The vector V and be beam axis establish Plane 1. Plane 2 Passes through the beam axis and is orthogt nal to Plane 1: Z elent Y elem
Plane 2 lane 1
Constrain d.o.f. 12~456 at node 1. Constrain d.o.f. 12f6 at nodes 2,3,4. Add forces to nodes 2 and 3. I
~......pc!_CD_t----4:!~~ooo ~
1
X
()
t2000N
The sign conventiods for finite elements and beam theory are in f onflict. Consider intemal forces and moment~ in a single beam element:
I
~. .lv1I . . . ttv2
~ ' ~ ~
. . . . . . . -!:::;.-..-.....
it .
MzB
YzB
'~=r-~~ B V YB
Both the FE (red) an~ classical (blue) forces must be interpreted in a given problem.
11.4
To obtain stresses in the BAR element, give the coordinates of 4 points on the cross-section: (-75,50) c : (75,50)
(-75,-50)
~~---~(75,-50)
IMSCLNASTRAN DATAl
Use 3 BAR ("simple" Euler-Bemoull~) elements. A flle 111anagement statement is:
asslgn output2='LAB2.0P2' ,new,unit=12,unformatted
I
STRESS=ALL ELFORCES=ALL SPCFORCES=ALL SUBCASE 1 LOAD= 50 SUBTITLE= 5000 N DOWN AND 2000 N UP
GRDSET,,,,,,,12 ~ 6
GRID,1,,
0.0, ~ .0,0.0,,123456
GRID,2,,1000.0, ~ .0,0.0
CBAR,l,29,1,2,l.O,l.O,O.O
CBAR,2,29,2,3,1. ~ ,1.0,0.0
CBAR,3,29,3,4,1. ,1.0,0.0 PBAR,29,12,1.5E+ 4,2.81E+7,1.25E+7,,,, ~ C +P1029 + Pl 0 2 9 , 7 5 . , 50 . , - 5 . 50 . , -7 5 . , -50 . , 7 5 . , :::> C-50. MAT1,12,2.07E+05 \ ,0.3 FORCE,50,2,0,1.0 0.,0.,-5000. FORCE,50,3,0,1.0 10.,0.,2000. END DATA
11.6
@
R
s
T2 .0 .0 .0 .0
LA C E M EN T T3
0
v E C T 0
R1 .0 .0 .0 .0 R2
POINT ID. T1 .0 1 .0 2 3 .0 .0 4
R3 .0 .0 .0 .0
OF T2
.0
CONSTRAINT R2 -1 . 000000E+06 R3 .0
BEND-MOMENT END-B
PLANE 1 PLANE 2
PLANE 1
.0 .0 .0
BAR
ELEMENTS SA3 SB3 SA4 SB4 -4.000E+00 S .. OOOE+OO S .. OOOE+OO 3.725E-15 7.450E-15 7.450E-15
4.000E+00 -8.000E+OO
-8.000E+00 -3.725E-15
-4.000E+00 8.000E+00
8.000E+00 3.725E-15
2 -S.OOOE+OO
3 -7.450E-15 -7.450E-15
-7.450E-15 -7.450E-15
7.450E-15 7.450E-15
The displacements and stresses are all exact, and will be plotted later.
11.7
C DATA
TITLE BIL ANDERSON, JIM BROWELL TITLE 18 UNE 1993 TITLE LAB RATORY 2: BEAM SIZINC}, 600000 ELEMENTS,S ELASTIC SHELL SECTION,3 END
POST
6
11 12 13 14 15 16 PRINT NODE
1
11.8
STRESS 1 TO 3
1
1 TO 3 POINT LOAD
$
THE FOLLOWING CARD SETS ARE FROM => <=LOAD ID 67 0.0 -5000.000
2
0.0
3
2000.000
FIXED DISP
1
0.0 1 2 1 CONNECTIVITY
3 3
0.0
0.0
5 2 5 3 5 COORDINATES 2 4
1
1
2
3 4
2
3 4
T 0 T A L
D I S P L A C E M E N T S
1 o.ooooo
ol ooooo
-l.oooooE-28
1.93386E-04 5 80158E 04
4 -6.04332E-11 1 { 6~i4~
s:ao15BE=04
1 2 3 4
11.10
Stresses are:
TRESCA MISES MEAN PRINCIPAL INTENSITY INTENSITY ELEMENT 1 NODE SECTION THICKNESS AVERAGE MEMBRANE STRESS 3.815E-08 MOMENT 6.659E+01 LAYER 1 STRESS 3. 995E+00 LAYER 2 STRESS 3.750E-08 LAYER 3 STRESS 3.995E+00 NORMAL MINIMUM INTERMEDIATE MAXIMUM INTENSITY O.OOOE+OO O.OOOE+O O 1 NODAL PT. COORDINATE= 0.100E+03
3.750E-08-1.250E-08-3.750E-08 O.OOOE+OO 0.000E+00-3.750E-08 3.995E+00 1.332E+00 O.OOOE+OO 0.000E+00 _._ ->:_ :.::::: ......~
.9 9:=~ :> o:-;
TRESCA
MISES
MEAN
PRINCIPAL
VALUES
INTENSITY INTENSITY ELEMENT 2 NODE SECTION THICKNESS AVERAGE MEMBRANE MOMENT 1.333E+02 LAYER 1 STRESS 7.997E+00 LAYER 2 STRESS 2.500E-08 LAYER 3 STRESS 7.997E+00 ELEMENT 3 NODE SECTION THICKNESS AVERAGE MEMBRANE MOMENT 7.584E-05 LAYER 1 STRESS 4.550E-06 LAYER 2 LAYER 3 STRESS 4.550E-06
NORMAL MINIMUM INTERMEDIATE MAXIMUM 1 INTENSITY 2 NODAL PT. COORDINATE= 0.100E+04 0.000E+00 0.100E+03
1.333E+02 4.443E+01 0.000E+00 O.OOOE+OO 1.333E+02 1.333E+02 7.997E+00 2.666E+00 0.000E+00 O.OOOE+OO 7.997E+00
7.99 7 E~ O a
2.500E-08 8.333E-09 O.OOOE+OO O.OOOE+OO 2.500E-08 2.500E-08 7.997E+00-2.666E+00-7.997E+00 O.OOOE+OO O.OOOE+OO 3 NODAL PT. COORDINATE= = 0.100E+03 0.200E+04
- 7 . 997E ~OO
0.000E+00
-7 ..
534f~-G 5
A~
~#:, ~.: 0 f~ - - C; ~;
11.11
OS DATA
Use 3 BAR elemen~s.
I
,..,.., ................_
z.~.v
BEGIN BUJ~K GRDSET 11 1111 124 GRID 1 1 11 0.0 10.0 10.0 11 123456 GRID 12 11 1 000.0 10.0 10.0 GRID 13 11 2000.0 10.0 10.0 GRID 14 11 3000.0 10.010.0 CBAR 1 l 129 1l 12 1 l.0 1l.OIO.O CBAR 12 129 12 13 1 l.0 11.0 10.0 CBAR 13 129 13 14 1 1.0 11.0 10.0
t
I::>
FORCE 1 50
I
12 10 11.0 1 . 0. 1 -5000.
1
I I I I I
0 . 2000 .
I
11.12
O.OOOOOE+OO
4.71845E-16 4.50886E-01 1.03060E+00
B A R ) I N E L EM E NT s S T R E s S E s BAR AXIAL SA4 SAl ELEMENT SA3 SA2 STRESS SB4 ID. SBl SB2 SB3 1 4.000000E+00 4.000000E+00 -4.000000E+00 -4.000000E+00 -O.OE+OO -8.000000E+00 -8.000000E+00 8.000000E+00 8.000000E+00 2 -7.999999E+00 -7.999999E+00 -3.000000E-06 -3.000000E-06 3 7.999999E+00 3.000000E-06 7.999999E+00 -O.OE+OO 3.000000E-06
F 0 R C E S
I N
B A R
E L E ME N T S
PLA..~E
( B A R )
SHEAR -
AXIAL
11.13
RSULTS
The deflection is,
and
1-D~::AS:
.....
-
sing MSC/NASTRAN
-
-
............ , .
........ --
,, .....
........
...
-
........
-
X ------------------- :=-~
: . - : .=.:. -
Displacements, in
wl
m, are:
w2
w3
w4
- .327E-17
.831E-08
Intemal moment
106
d shear:
s.
MSC/NASTRAN
!vlA.RC
~ z~ o 1-----+\- + - - - - - f - - - - + - - . ~~-1000 ~ I
r/J
-2000 -3000
r-
t-----+----' :
I
11.14
4 MPa
-4 MPa
n-_,.
-----u->-4 MPa
a
X
= _
Mz
/2
(0,0,0)
11.15
NASTRAN usage:
CBAR
+15
I
1
1256
Thin-walled tubing ,lso tends to collapse, causing "c~rippling." Nature is perverse in that the walls will move a way to reduce the area moments of inertia!
} ending
)> torsion
11.16
LECTURE 12
MODIFICATION OF EQUATIONS
12.1
SOLUTI NPROCESS
Topological layou (mesh) [K]{u}= {F}
.,
[K]{u}= {F}
Apply SPC's:
v
II II
[K]{u}= {F}
ode.-<> :: \..,~A ~j.,.
t~)J:)tirrdze
~;,.
()1h ~~~rcl\.;~:t11~c;J1t:
l)a.t1(lvliclth .
v
[K]{u}={F} 44,994
[K]{tt} = {F}
li
44,994
-1
{ F} ,{cr} ,{c}
12.2
...
ISTANDARD FORMI
We desire the standard form:
1]{111}- {2} 2 2 6 2 3 [5 1 2 7 u 4
u2
-
where [K] is real, symmetric and positive definite, all the forces are known, and all the displacements are unknown. Unfortunately, most physical problems are "mixed boundary value problems" where some forces and some displacements are known:
5 2 1]{ u .... [ 2 6 2 U'> 1 2 7 4
1} -{2 3}
-
F'3
[ ~ ~ ~]{~~}
127 4
={ ; }
_ f /3
[ ~ ~ ~]{~~} = {~ ~ g~~!~}
0 0 1 u3 4
12.3
or by a "penalty" pproach:
[ ~ ~ ~ ]{~!} u;
1 2 7x108
={
~ 7x4x10
/;3
8)
or by Lagrange m ltipliers:
= { ~} [ ~1 ~]{~!} 7 Zl;
A,(u 3 -4)=0
Partitioning allows solution of the smallest set of equations. Lagr~ge multipliers are good for nonlinear effect (e.g. contact problems).
I
These are used sing y to represent an inclined plane or slot, in gro ps to represent rigid body behavior, etc. Example:
Uzi-Uzz=O
12.4
How can one MPC relation remove two variables (one force and one displacement) from the problem?
K 21 ' 22 K 22' 22
K 21 ' 21 K 22 ' 21
Each MPC implicitly implies a second, hidden relation that allows no work to be done by reaction forces (a holonomic constraint):
{u}T {F}= 0
Example:
A. REFLECTIVE SYMMETRY
A body has a reflective plane if the half of the
body behind the plane is identical to a mirror image of the front half.
In structural mechanics, one can exploit the symmetry if one has symmetry of: geometry material displacement boundary conditions
12.5
It is not necessary~r the loads to be symmetric. They can be deco posed into symmetric and antisymmetric com onents, and two problems (half-size) solved. ~
antisymmetric load
+
B. ROTATIO AL SYMMETRY
The body has a strai ht axis of symmetry. A fundamental region an generate the entire body by n replications, eq ally spaced about the axis:
/~
n=2 n=3 n=4
C. DIHEDRAL SYMMETRY
Combine rotational and reflective symmetry. One only needs to model half of one cyclic region. There are multiple reflective planes.
D.TRANSLATIONAL SYMMETRY
The fundamental pattern is repeated on a straight line:
~ ~ ~'-----1~'-----1~'------..<
E. AXISYMMETRY
A subcase of rotational (cyclic) symmetry:
o o
e
o
straight axis of rotation. body is swept out by rotation of one segment. lathe pieces are examples. can have skew due to winding process for shells, and material anisotropy.
12.7
t3
!
!
t.:P'Y
~X
3
1____,__..,.
~
'
j
, ~--------+-~1----'t
Symmetric B. C.
T1 = 0 F2 = 0 F3 = 0
Antisymmetric B. C.:
B. STRUCTURAL ELEMENTS
~---t"'lr.
Symmetric B. C.
~~~:-... . ..
(01'! !Jo)i.~~:\>7~W.:
causes no strain energy. An elastic body undergoes neither direct nor shear strain. In contrast, straining modes cause strain energy to be developed. Straining modes can, however, include some rigid body motion. Elastic bodies in three dimensions have 6 rigid body modes, 3 translations and 3 rotations: surge (skid) roll sideslip pitch plunge yaw
12.9
There are two tran lations and one rotation. If the receiver falls off, one has a "mechanism." A broken spring cause a 1nechani a dial phone, would also
The axisymmetric c4se is special and must be handled carefully. he NASTRAN codes embed axisymmetry into a 3-D space. Other codes restrict the sp ce, typically to the xz plane. If the element is a s ecial axisymmetric element restricted to the xz lane, there is only one rigid body mode---transla ion in the z direction.
Hoop stresses prevent both x translation and rotation in the xz plane as rigid body modes.
I
\
\ z
@
12.10
B. RULES OF THUMB
BODY CONSTRAINED TO LIE IN:
3-D 2-D
SUB CASE
3
1 1
1-D
Codes vacy widely in how two-dimensional cases are handled. Some codes (SAP) treat 2-D with special elements. Other codes (NASTRAN ) embed plane stress and plane strain into 3-D.
z
1
-~ x
Node 2 (beam/plate elements): (easy) I ~-----T~ 2 _,T_3 ~'-R~ 1 ,_R~ o __~~--~] 1 -,T~ 2 _,R~ 3 ___
12.11
Node 1 (solid): Node 3 (solid): Node 4 (mem Node 1 (solid): Node 3 (solid):
T 1 , T2, T3
T2 , T3 e): T 2 = 0
=0
=0
(good)
s because it does not This last attempt rigid body mode and it constrain the r from nose to tail! This prevents elastics also "grounds" liv loads in the longitudinal directior1:
3
L1
MEC ISMS
Problem with load r Beams were free to beams w---+-----~ >plates Joining elasticity ele ents to structural elements causes problems: beam 'truss
12.12
NODAL SINGULARITIES
If no stiffness acts in a degree of freedom at a node, the stiffness matrix is singular. This type of singularity can be "sniffed out" by a computer code, by forming an eigenvalue problem at each node and searching for principal directions of stiffness, and their values. In the NASTRAN codes, there is an AUTOSPC feature which removes these singularities.
6!A
3
0
!7
l2 /
5 v/
4
~--~
IPROBLEM SESSION!
Proble111 1. Rigid body and elastic Dlotion Solve for the elastic deflections of a beam which is "guided" at its two ends. The guides allow free translation, but no rotation. Neglect the force of gravity.
4000N
2000N
El= 1013 N/mm
I I I
L=1000 mm
12.13
Solution There is a transla onal rigid body mode present. Unless r oved from the static elastic solution, it will de troy the solution procedure. The body is in equ ibrium, so the rigid body mode is not an im ortant part of the solution. Use a half-model, sing the vertical plane of symmetry at mids an, and assuming the center is fixed in s ace (removing the r.b.m.).
u4 . --~~- u2 (/r.~ _ ___,___-i~........ / ~- - - - - - - - - - - - - - '::;:;:;.....
01 + :.
4u3
\.:::::::.,...
= 0.0167 mm
(up)
12.14
Proble:m 2. Syn1:metric quadrilateral ele:ment A hypothetical, two-dimensional, linear, elastic finite element has four nodes. It has a reflective plane that passes through the x axis.
2mm
i>-
I
~ x
200 N
I.--~~;$:" 1 mm
Nodes 1, 2, and 4 are held stationary, and node 3 is loaded with 200 N at 45 as shown. Displacements (shown) result. Find as many kij values as possible. @
Solution The 2-D quadrilateral has 8 d.o.f. It could be plane strain, plane stress or axisymmetry (not stated). The experiment elds:
ot
xe:: :
. xu..
fl
~
0
0
/3
We have 2 equati, ns in 3 unknowns. Need more information, perhaps using symmetry. Consider a menta experiment, with the same boundary conditi ns and a horizontal load . at the right node: .=jl ~ u6 There will be no v rtical ~ : ---displacement u 6 :
k 55 56 k 56 66
I
us
0 0 0
0 0 0 0 -
ft
!2
!3 !4
!s
0
!1
fs
k 56 u 5 = 0
~)
k55(1)
~) = 141.4
k66(2) = 141.4
k 55 =141.4 N /mm
k 66 = 70.7 N lmm
= k61 = 0
12.16
6f
k2 = ks2 =0
Two children are sitting on a teeter totter. They weigh 150 N 250.N and 250 N. Make a finite element model to find the exact stresses and deflections in the 1200 l-2000mm beam. mm mm Solution The model will need only three elements for an exact solution. A rigid body rotation must be
~1 ..
_JJ
-. .. ..-.. .:::::::
rem~d~~~u-1
........: : ,: :~ -.
_____
u_~~~~r-3___u_6~ (~7
\
::::::.-.-.....
..... ;~::::::...
Arbitrarily set u 4 = 0 to constrain the r. b. m. This centers the teeter totter horizontally.
,!:\.,\
12 L3
X X
1
-o
12 L 3
X
1
6 L2
X
1
0 0
X X X X X X
0 0
X
X
0 0 0 0
0 0 0 0
X X X X
X X X X
0 0 0 0
X X X
X
12.17
0 0 0 0
X X X
X
u1 u2
0 0
0 0
X X X
X
us
u6 u7
0 0 -
Ug
-150 0 F3 F4 -250 0 0 0
/
@
ProbleD14. m.etric loads A bracket is forme from two beams. It has a reflective plane of ymmetry passing through the x axis. Explai the "traditional" use of symmetry, includ g a) proper b. c. at e reflective plane !f b) how to handle e load c) recovering resul sin the ''image'' Solution \ The loading is antiJ metric with respect to the lane. We will model the top half of the bracket, d apply antisymmetric b. c. at the cut edg . The problem is planar, so out-of-plane displa ements are constrained. @
a,b) In a general3- imensional problem, the antisymmetric cond tions at the cut edge would Q>e: +!f/2
Tx =
= \ 0 F~=Mx=M=O
1'z
R~
The forces are interi r ones that are released wh n the structure is cut, and don't account for external oads. The displacement co dition at the cut edge is: u 4 =0 The interior force F 5 d moment M 6 are zero, but the exterior live lEad causes: F5 = -!!{2
I
12.18
c) To recover the displacements and stresses in the image half, use the results in the real half (upper) and change the signs on all quantities. Interpret results in upper half, and then imagine the results as reflected in the image. This process is equivalent to drawing the image half, using left-handed coordinates, and using the results of the real half with signs changed. ~ This process is error-prone; many engineers refuse to use symmetry concepts and instead model the entire structure! An additional problem is that graphics systems often cannot present the image results, so there is no automated way to visualize the answers. /
12.19
I I
I I I
I I I I I I I
I I I I I
I
LECTURE 13
GAUSS ELIMINATION
A. B. C. D. E. PROCEDURE DECOMPOSITION FORMULA FORWARD SOLUTION BACK SUBSTITUTION EXAMPLE
13.1
ITYPES OF SOLVERS
o o
o
Static stress:
[K]{u} = {F}
Each of these can be linear or nonlinear. Transient problems are solved by marching forward in time. We consider only static stress problems in this lecture series.
13.2
r iKi{~i-- iFi ..
~
---~---J
[K]{~
{p}]
-\ .. --
"
~__j
{p}l
.nn--
x~
stress-A= =: :-
analysis
trivial {u}=O
v:
~::
where [K] is nxn, real, symmetric and positive definite. {u} is unknown. {F} is known. Positive definite means: {u}T [K]{u} > 0 for all {u} :t 0 Any displacement causes strain energy. The general concept for solution is:
{u} = [K]- 1{F}
G)
A. DIRECT METHODS
Elimination Cramer's rule Gauss elimination Cholesky decomposition
[K] [K]
=[L][L]T
Gauss-Doolittle decomposition
=[LI][D][LI]T
Wavefront Dissection
B. ITERATION METHODS
Jacobi Southwell relaxation
Ne~on-Raphson
Gauss-Seidel Chaotic
GV
IGAUSS ELIMINATIONI
A. PROCEDURE
Consider a linear static stress problem:
[K]{u} = {F}
where [K] is an nxn, real symmetric matrix and is positive definite. Decomposition T
~v....~=--
Xb ...
>..........
[ K]
=[L
] 1
[D] [L 1]
dll
where:
[Ld=
1
121 131
0 1
132
0 0 1
0
d22
. . . .
[D]=
0 0
Forward Solution
~~~ -<1(
: . . . . . :..< . . . . . . . . .
.. :.....
= {F}
[L 1]{X} = {F}
Solve for {X}. (Easy!) Back Substitution Use the definition for {X}:
[D][L 1]T{u} ={X}
(Easy!)
13.5
B. DECOMPOSITION FORMULA
K11 K12 K13 K21 K22 K23 K_31 K_32 K_33 : ..
0 121 1
131
. .
. .
. .
0 ... d 11 d11121 d11131 ... 0 ... 0 d22 d22132 ... 0 0 132 1 ... d33
. . .
This a matrix equation, with (n+l)n/2 independent relations. Unravel by looking at the first column of [K]:
K11 = d11
K21 = d11l21 l21=K21/d11
K31 = d11l31
l31=K31Id11
K ~1/d1"~1-~~,.....-~c-i -2 ~)Ym~,
l ' ' '
i~
~~w.J
The solution proceeds column-wise from the left. Storage space is saved by writing the newlyfound d ii and 1ij terms on top of the [K] matrix.
13.6
'-1
l
wm1
~
This is the expensive part of the equation solver. It requires on the order of n 3 floating point operations (flops) for a full matrix.
C C C C C C
SUBROUTINE DCOMP(N, K) THIS SUBROUTINE DECOMPOSES AN NxN SYMMETRIC, POSITIVE DEFINITE MATRIX [K] INTO THE PRODUCT [L1] [D] [L1]AT. THE MATRIX [11] IS RETURNED TO THE MAIN PROGRAM IN THE LOWER LEFT PORTION OF [K] . THE MATRIX [D] IS RETURNED ON THE DIAGONAL OF [K] . REAL K(100,100}, SUM INTEGER I, KK, K1, J, N DO 50 J = 1, N DO 50 I = J, N SUM = K(I,J} K1 = J - 1 IF (J .EQ. 1) GO TO 20 DO 10 KK = 1, K1 10 SUM = SUM - K(J,KK) * K(I,KK) * K(KK,KK) 20 IF (I .NE. J) GO TO 40 IF (SUM .LE. 0.0) WRITE (6,30) 30 FORMAT (' DECOMPOSITION HAS FAILED. NEGATIVE TERM 1 ON MAIN DIAGONAL') IF (SUM .LE. 0.0) RETURN K(I,J) = SUM GO TO 50 40 K(I,J) = SUM I K(J,J) 50 CONTINUE RETURN END
13.7
C. FORWARD SOLUTION
[LI]{X} = {F}
1 0 12I 1 13I . 132 0 0 1
1n3
. .
. .
0 0 . . ~3 .
XI x2
FI
F2
-
F_3
. .
1ni
1 n2
1 xn
Fn
'!V'!NtN'."J'oN'tt.Y't,...~~NJ'...,.,Y.WN.VN.-.v.V'NNV'NV'tl'~...
w,.,..,.,.,/'.N'S
l-I
! X.=
~
~
F.- Ll.kXk
l
k=I
(i=l,2;,n)
I
-~
D. BACK SUBSTITUTION
[D][LI]T {u} ={X} dii dii12I
d22
0 0 0
dii1ni ~21n2
dn-I,n-I1n,n-I dnn
UI u2
-
XI x2 xn-I
xn
0 0
un-I
un
IUn = Xn/dnn l
t .:.;..._..-;.;.:..;..x.:.::;.;.:.:.;.: -oto:..:..
-.:.:.-.e:o :~-=~
13.8
,.....,,_,_,................MNo"o"NhV'V"""'hVJYUo"V"WNhW.Y........,,Wo"N""~
I Xi-diik~!fiuk I ui= du
l
(i=n,n-1,-,1)
::t!YVN',..v.w ..-.v....,..,.....,..~,...,,.,..,.N,..,.,.,,..~~..-.-.wNN.w~.,..,..,..~,~,...,..,.,.,..,~..,~<l'lfV'.,W.W..,..,.,,,..,.._.,..:v_.N,...,.,.,,.........,....,.w,.,.,,.,.J
C C C C C
c c
SUBROUTINE SOLVE(N, K, F, U) THIS SUBROUTINE SOLVES THE LINEAR ALGEBRAIC SET OF EQUATIONS [K]{U} = {F} WHERE [K] IS AN NxN, REAL, SYMMETRIC, POSITIVE DEFINITE MATRIX AND {U} AND {F} ARE VECTORS WITH N COMPONENTS. THIS SOLVER PRESUMES [K] HAS BEEN DECOMPOSED AND STORED AS IN DECOMP
II II
10 20
30 40
REAL U(100) F(100), K(100,100), SUM INTEGER KK, K1, K2, I, N DO 20 I = 1, N SUM = F(I) K1 = I - 1 IF (I .EQ. 1) GO TO 20 DO 10 KK = 1, Kl SUM = SUM - K(I,KK) * U(KK) U(I) = SUM DO 40 I1 = 1, N I = N - I1 + 1 SUM = U(I) K2 = I + 1 IF (I .EQ. N) GO TO 40 DO 30 KK = K2, N SUM= SUM- K(KK,I) * U(KK) * K(I,I) U(I) = SUM I K(I,I) RETURN END
I
13.9
E. EXAMPLE
PROBLEM
SOLUTION
[~ ~
1 0 10
d33
-;..:.~...-..-
......:<-".-;.o"".OO:;..:.o:~x:ex~:.;..-.-;.;:::;
1 z31=
13.10
10 =
l31dlll31
+ l32d22l32 + d33
[L1]{X}
= {F}
The forwar[id?ohgb].o{1r}eld{:i}
1 -2 1
x3
31
'original loads
rxl. sl
&..:~~~:
X2 = 12- 16
~
1(8) + (-2)(-4) + x3 = 31 X3 = 31 - 16
13.11
[g
! f]{~~}={1~}
u 1 + (2)(2) + (1)(3) = 8
{u}={i}
GAUSS-SEIDEL ITERATION
Treat the problem as an "input-output" type.
[K]{u} = {F}
13.12
b;!ACOBI ITERATION!
[K]{u} = {F}
Iteration methods might not converge. They are good for nonlinear problems. The most popular is Newton-Raphson.
IPROBLEM SESSIONI
ProbleDl 1. Cholesky decon1position Solve the set of equations
13.13
2 = 121111 = 121(1) 6=
J[
Il
1Xl+OX2=5
2X1 +.J2X2 =14 2(5)+.J2X2 =14
I Xl= 5! LNAYA~V.W'-'J
!x~ --2~l
..
l . . . . .,_..~~~.v .v....-..v.vN.~
j
13.14
2} 12 1]{x1} {5 5 0 [2 1 0 10 12
x2 -
x3
~ X1
~
--1 -
5x2<1)
(1)
=5-(2)(-1)=7
IxJI) = 1.4
~
10x3 =12-(1)(-1)=13
!x3
~
(l)-
1. 3
0 o]{x
10
1}( ) { x2 =
x3
(2)
X1
x2
X3
= }
{-2.100}
1.840
1.410
x2
xl
x3
-7.738 4.095 10 5.206 20 -10.514 5.593 30 -11.482 5.728 40 -11.819 5.775 50 -11.937 110 -11.999887 5. 799955 It takes about 42 iterations to get
13.16
LECTURE 14
OVERVIEW BANDWIDTH FACTORIZATION FLOATING POINT OPERATIONS BANDWIDTH OPTIMIZATION WAVEFRONT COMPUTER COSTS PROBLEM SESSION
1. BANDWIDTH FOR 2-D PROBLEM 2. BANDWIDTH OF A STAIR RAILING 3. WAVEFRONT OF A RING STRUCTURE
REFERENCES
1. Melosh, R. J., and Bamford, R. M., "Efficient Solution of Load-Deflection Equations," Joumal of the Structural Division, ASCE, Vol. 95, 1969, pp. 667-676. 2. Irons, B. M., "A Frontal Solution Program," Int. Joumal for Numerical Methods in Engineering, Vol. 2, 1970, pp. 5-32. 3. Zienkiewicz, 0. C., and R. L. Taylor, ''The Finite Element Method," Third Edition, McGraw-Hill, N.Y., 1988, pp. 89-90.
14.1
IOVERVIEW)
Storage optimization is actually more important than solution optimization. Floating-point operations cost 1/3 of total. o Occupancy of high-speed memory costs 2/3. We need to work harder on storage than on solvers! Finite element matrices are "sparse:"
.. .. .. . .. . ..... . .
. . . .... .. .. .. ... . . . . . ... ....
sparse
. .. .. .
. ..... .
"skyline"
14.2
banded
B = (D+ 1) ndof
greatest difference in nodal numbering ____/ within a single element
\___
Triangular factorization doesn't spoil bandwidth: . .. ... .. . 0 ..-::-. :-. 0 0 .. > 0 9'\}: = ~::. 0 0 =:\
IFACTORI~TION I
~ ~J~~Xl~~
0 0 0 0 0000
0 0 0
000000
oxooo
14.3
lj{(1 d11
(= 0)
(= 0)
J.. = - - - - l)
,./- l "kdkk/Lk l{
d ..
]]
. 1
- ...''
''
@,
'-.p.
NB terms _____,/
:ij
kru
=
k= 1 J ~
i lkdkk lik
dii
~B products
if banded
0~ ~
14.4
xi=
N terms
Fj-
L
k=l
lik xk
~ B products, if banded
.l:hS
N
(i = 1,2,. , N)
.F~l.. (.J1--\S o :
L
11
lkiuk
Nterms~
J.~i+l
\:
(i= N,N-1,-,1)
B products, if banded
(j)
IBANDWIDTH OPTIMIZATIONI .
First nodenumbering scheme:
1
~
==
6
X X
6-----1
X
X
~'
X X X 0 X 0 X X X 0 0 0 X X 0 X X 0 0 X 0 X X X 0 0 X 0
14.5
0 0
X~ X X 0
X
0
X
X X X 0 X
0 0 X 0
X X
3
:
5
::
7
:
If this had been a large problem, the CPU time would have been proportional to NB2:
CPU2 =
CPU1
N2 B~ N1B 1
= 4 2 = 0.44
6
For such small problems, however, "overhead" (input-output, assembly, etc.) would make the problems indistinguishable.
@
14.6
IWAVEFRONT OPTIMIZATION
The method was due to Irons and Melosh. It is basically Gauss elimination in a preferred order, based on element topology. Wavefront methods are useful on computers that have dynamic storage allocation. @
~ k?.t =..:
!3.'~l
=
>~
-: ~
~-:
..__~~
~-~
11
~
r1 LJ
f""'j
!""'!
n n l--1 -'
L1 Ll
< j i :..-......>
ul u2 u3 u4
)"""~
LJ
i..............,~
)"'
Fl F2 F3 F4
"'
~
::
~~
CD
l-~
G)
~
~::.
@
~
8
element
5
active
dropped
1 2
3
3 ;
7
~
4
.o:;.
CD
4:~
~
-~
:
G)
~:-
8
element 1
2 active 1 ,2,3,8
dropped 1,8
2 3
2,3,5,7 4,5,6,7
2,3 4,5,6,7
ICOMPUTER COSTSI ~
A. EFFECT OF DIMENSION
Compare CPU costs for 1-D, 2-D and 3-D, at the same accuracy.
CPU TIME oc NB 2
oc84
00100100000000010010 00100100000000000000
oc 1,866,312
14.8
00000000000010000000 00100000000000000000 00000000000010000000 00000000000010000000 00000000000010000000 00000000000010000000 00000000000000000000 00000000000010000000 00000000100000000000 00000000000010000000 00000000000000000000 ODOODOOOOOOOOODOODOO ODOOOOOOOOOOIODOODOO OODODOOOOOODOOOOODOO
oooo~oooooeooooooo
In 3-D:
T
ex:
oc
In order to maintain the same accuracy, the CPU time jumps as 102, 106, 1010! For large problem sizes, the CPU cost could jump by a factor of 10,000 for each jump in dimension. One solution is to use fewer, isoparametric elements. Also, one can lower accuracy. @
CPU TIME
oc
In 1 - D :
f CPU
B= 2
~~A~.O.X~----
0000000
TIME
oc
N1
<-X .a'! .:(J
0000000000000
00000000000000000000
In 2-D:
~
B=.JN
oc oc
CPU TIME
~ CPU :-
N(-JN)
N i
:-:
&..-...:e:-)IC!o;.:....-.:~-~<0~---...x-:-J
TIME
2~
t---.JN ~
ooooooooo . ooooooooo
00000000000
00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000 00000000000000000000
000000000000000000~ 000000000000000000~
14.9
In 3-D:
CPU TIME
oc
B=N3
2
N(N 3)
1 ~1
-~
r '"
--
TIME
I CPU ~
. 0:: . .
oc
N3
. . :s. .
:oo:oo:o::: ....
. . . . .
~ ::::::::::~.~ 3 ~J N
oooeoooooo~
0000000000 0000000000
ooeo ooeo
j_
oooeoooooo
359 d.o.f. 4load cases 1364 d.o.f. 2load cases 4836 d.o.f. 6 modes
28% 4
2
18% 10
6
2% 0 0
4% 4 4
6%
42% 4 3
78
85
Commercial problems today average 35,000 d.o.f. and decomposition is over 95o/o of time.
14.10
IPROBLEM SESSIONI
ProbleDl 1. Bandwidth for 2-D probleDl
Find the minimum bandwidth of the plane stress problem shown. The material is sheet aluminum, with a triangular hole (shaded). Express your answer in detailed notation.
Solution o There are 11 nodes. o Number through the "thin" direction, if any. o Keep track of the largest difference, D. 0 Plane stress has 5 ~-----~--o 11 2 d.o.f. per node.
~w.w.v.v.-.-.~~-.-.v.
10
ProbleDl 2.
Band~dth
of a stair railing
with circular spindles at regular intervals. All joints are considered clamped, by virtue of embedding the rail and spindles in the surrounding structure.
Consider bending of the railing for the case of a human hand pushing down at an arbitrary position on the rail. An exact solution in the xy plane is to be found. The worst case of deflection and stress are required. a) What is the minimum bandwidth of the total set of assembled equations (before single point constraints at the wall and floor are applied)? Explain your reasoning. b) What is the minimum bandwidth of the "reduced" set of equations (after the single point constraints are applied, the equations are partitioned and only the active degrees of freedom remain)? Explain your reasoning. fori\
@
14.12
Solution The joints between rail and spindles require 3 d.o.f., as do the joints at the floor. There may be an additional node due to the loading, probably at .the last bay toward the free end .
. 4 - ...
ff~-~
2 ;1 ; 14
ir~
1~ 4
i
h.;
~~ ~:f~i-4--Cf~ cf~
l=.
1.
.il?:
a) The beam system is embedded in a 2-D space. The largest difference D in nodal numbering is 2.
@
@
B=(D+On
= (2 + 1)3
B= 9
(detailed notation)
b) If the nodes at the floor and wall are partitioned out, the remaining structure is a simple 1-D string of elements:
B=(D+1)n
=(1+1)3
B=6
14.13
(detailed notation) /
ProbleDl 3. Wavefront of a ring structure The finite element model of a ring-like structure is to be solved by a wavefront solver. The problem is plane stress. What is the minimum rms wavefront in compact notation?
Solution It is not obvious how to number. For best wavefront, one usually tries to make a continuous wave pass through the structure as in our first suggested scheme:
4 5 6 7 8
4 6 6 6 6 6 6 4
0
2 2 2 2 2 ' 2
This scheme has a repetitive pattem, with 2 extra nodes carried along. The rms wavefront:
wrms
=~N~/ri ~=
.!(4 2 +6 ++4 8
14.14
2 )
= 5.57
A second numbering scheme, nu~bering clockwise around the body, gives the same wavefront table as the first scheme.
wrms =
5.57
14.15
LECTURE 15
FLEXURE
ASSUMPTIONS SIGN CONVENTION FOR S1RUCTURES SIGN CONVENTION FOR PLATES PLANE SECTIONS ARGUMENT REDUCTION FROM 3-D TO PLATE THEORY FORCE RESULTANTS Shear Resultants Moment and Twist Resultants Summary of Force Resultants G. MOMENT-CURVATURE LAW H. EQUILIBRIUM DERIVATION I. PLATE EQUATION SUMMARY J. BOUNDARY CONDITIONS K. STRESS RECOVERY L. STRAIN ENERGY A. B. C. D. E. F.
REFERENCES
1. Timoshenko, S'., "Theory of Plates and Shells," McGraw-Hill Book Co., New York, 1940. 2. MSC/NASTRAN User's Manual, Vol. 1., The MacNealSchwendler Corp., Los Angeles, 1992.
15.1
FLEXURE I
A. ASSUMPTIONS
shearing deformation) ~
15.2
(not possible)
The viewer is in the first octant and looks toward the origin. The surfaces of a rectangular prism which are seen are "positive" surfaces, upon which positive force resultants cause positive displacements, slopes and curvatures. Positive moments cause compression in the top fiber of beams and plates. This is Timoshenko's sign convention for beams and the NASTRANconvention for plates. This differs from classical elasticity where positive components point along the positive axes.
15.3
z
A
Positive shears cause positive displacements and slopes. They also cause compression in the top fibers.
I
Positive moments cause positive displacements and slopes. They cause compression in the top fibers.
15.4
Positive twists cause positive displacements and slopes. They cause compression in the top fibers.
~
(_
. - ~r-IT--
;J
~
w...w..-.>A..-..,..,....,_..,..,......"'_.v.,..,.,N<,.,.>:.....w-.- ..-.-.w...w...,
.:. ..........~
15.5
( ) z vx,y,z-
-- aw0ay (x,y)
The lateral displacement due to bending is large compared to the expansion or contraction due to Poisson ratio effects:
w(x,y,z)= w 0 (x,y)
-v -v E E E -v _l -v E E E -v -v _l_
1
0 0
0
0
0 0
~z zx
E 0
0 0
"-
E
0 0
E 0
0
1
0 0
1
0 0
0
-
0
0
~
G 0
G 0
1
..,/
[C]
15.6
dx
Bx By
a
0
()y
a
0
0
0
rxy - a a
Yyz
cz
ffz
rzx
\....
dy dx
dz
iz
0
v
dx
a ay a
[D]
@
Yzx=O
(j_
=0
ex ey ez Yxy 0 0
-v -v E E E 0 -v 1 -v 0 E E E -v -v _l_ 0
E 0 E 0 E 0
0
1
0 0 0 0
ax ay 0 rxy !"yz
'l'zx
0 0 0
1
0 0
0
0
0
0
G
0
0
-
a
0
0
ax
0
dy
0
- a a
dy
a dz
0
()
~
dz
~
dz
dj
dx
e = -z()2wo(x,y)
e -- ()2wo(x,y)
x -
()x2
~------------------------~~
The stress-strain law becomes that of plane stress! Each layer through the thickness of the plate acts in plane stress and az = 0.
E Cx v cy - - E
0
1
Y
W0
()y2
_v
E 0
-1
0 0
-1
Gx Gy
'!xy
Yxy
v
1
0
15.8
G(1- v 2 )
0 0
ex cy
Yxy
F. FORCE RESULTANTS
Shear Resultants The shear force on width ~y is found:
Vx =
-h/2
12
rzx~ydz
v r2
uy
-h/2
Likewise:
vy =
r;yzdz
-h/2
15.9
Moment and Twist Resultants Mx is found by integrating crx through the thickness.
Mx=Likewise:
r~xzdz
. h/2
-h/2
My=- Jcryzdz
-h/2
Sumn1~
of Force Resultants:
h/2
Vx=
J rzxdz
-h/2 h/2
vy = Jryzdz
-h/2 h/2
Mx =- J crxzdz
-h/2 h/2
M y =-
Ja . y zdz
-h/2 h/2
Mxy =-
Jrxy zdz
15.10
-h/2
G. MOMENT-CURVATURE LAW
At this point, we understand the shear-stress and the moment-stress relationships. We can now develop moment-curvature relations, but have to leave shear-curvature relations until later. (They are found from equilibrium.)
hf2
Mx =- Jaxzdz
-h/2
Mx
r d~o + vd2wo] h3 -~ dx 2 dy 2 12
Ti'
v2)
[()2wo J2wo]
[J2wo
J2wo]
H. EQUILIBRIUM DERIVATION
Free body diagram with differential forces on an infinitesimal plate section:
Vy+dVy Mxy+dMxy My+dMY
L Fz = 0
( da~ dx )dy+ (
I
~
x
aa;
a
ay
-v-w.w.,..r.w..w~.-.-~.-
Y + p(x,y)
Y./-,J',..tV"hVh.-.-.v.~V.-J'-.r.l"tt'~.-.
ax
=0
..w..r..v...-..,.,J'...,.,.,.-.-h.r...-.,.,.vaa.,.,.J'a._,.y,..,..,.,..-,.,.,.,..,.,.-.v.J'""J'.V.h:
15.12
~:paM~-:;--~~ - aX ay y j[ [j
....................w.-..-..-.......-....-...-.' " "".-."."""""" " "
We now have three equations for plate equilibrium in terms of resultant forces.
Let us develop equilibrium equations in terms of the deflection of the middle surface w 0 (x,y). Solve for shears in terms of moments and use the moment curvature relations: At this point, we will follow conventional practice and drop the subscript on w 0 (x,y).
V = _ dM:x _ dMxy
X
ax
ay
(d w
dw )
a3 w
15.13 .
=-
aMy
ay
---
aMxy
ax
avx avy
4 4 a a -D w_D w
dx + dy + P = 0
ax4
4 4 a w a w -D -D +p=O 4 2 2 2 2 ax ay ay ax ay
This equation describes the infinitesimal deflection of plates, with no midplane stretching. @
15.14
(a
a4 w a4 w) al = p(x,y)
D
Bending rigidity:
Eh 3 ( - v 2) 12 1
4 v =(;x: +
2 a::ay2 +
;;4)
Lagrange, ~2
1811
~
p(;,;)! ,.w.~J
J. BOUNDARY CONDITIONS@
y
~----------===
~/consideration
edge under
~-------------~~--x
On x=a, consider:
c ss f
0 0
0
dx (a,y)
()2 D~dx2 + v dy2 w(a,y) ()2 -D dx ()x2+ dy2_ w(a,y)
aw
()2]
d [ ()2
-~(twist)
shear
K. STRESS RECOVERY
One typically solves a plate problem in terms of displacements. One must then find the stresses from this displacement information:
zE O"x--1- V2 dx 2 +
~ v
xextreme
-+h -
w 0+v~ w0]
Likewise:
~a
~
;:
6M .,
Y extreme - -
-+
h2
Y ~
~
:;
15.16
Shearing stresses rxz and ryz are assumed to be parabolic through the plate thickness. Their maxi.mum value is at the middle surface:
~ t" - A -; \IX ~~ ~y~ ~ extreme 2 h extreme - 2 h I -- - _ .,_ . - . . -.. -- - - _ .,..,-- . . .. . .,_ .-. - - - . _,.,.,....,~-="-"W J
XZ YZ
jj
.......,,.,..,._,.,,....
L. STRAIN ENERGY
Strain energy is needed for potential energy solutions, such as Rayleigh-Ritz. There are two ways to find the strain energy expression, integrating stress-strain, or integrating moment-curvature. Use stress and strain; integrate over volume.
U
i J{o}T
v
2
{e}dV
()2w
=1
-zdx2 d2w
-z ay2
dV
-1+ vdxiJy
15.17
zE
()2w
- zdxdy
2 ()2w
U=
1 -
2 ~A 1Jz 12
3
2 z2 dz JJ
\..._
_______
=-y-
[(a2 )2 + ./. ::l....:..~- ;2 -:.) ( a2 w ) --.....::..::......::._. _~ ~ + a 2 , . . " y V2 a 2 + 2(1- v)(a 2 w 'f dxdy axay)
W
. . ., .~- .
("'l
.....;~ . . .:. .
u ...,.
Vj
,~
We need to make a perfect "square" by working on the blue term. Add and subtract a term. @
U = 1 Jf D
2
A
+ 2(1- v)(d 2 w
___________
/
twisting terms
It can be shown that the twisting terms vanish if the plate has zero deflection (simplysupported or clamped) on all boundaries.
15.18
The same expression for strain energy can be found by integrating moment-curvature:
IsTRETCHING I
z
A
l }
Nx+dNx
N yx
This is a plane stress problem. Boundary conditions involve displacements and forces:
u(x,y) v(x,y) Nx(x,y) NyCx,y) Nxy(x,y)
Under this infinitesimal deflection theory, one can superpose the plate stretching solution and bending solution.
IDIRAC DELTAI
Concentrated loads are hard to handle without the concept of the Dirac delta function. It is a hypothetical function that goes to infinity at a certain point, yet has a unit integral.
B(x-x 0 )
{D(x-x 0 )dx= 1
0
IPROBLEM SESSIONI
Proble111 1. Plate with concentrated load
A steel plate measures lm x lm x 5 mm. It is simply supported on all four edges. A 50 N load acts downward at its center. What is the maximum deflection? What are the bending stresses at the center? The answer can be approximate (+30%).
E = 208,000 MPa v = 0.3 D= Eh3 12(1- v2)
SON
5 mm
= 2.38x10 6
N mm
Solution Use a one-term potential energy solution. Assume: w(x,y) = Csin ~x sin~ Strain energy:
U= ~ fn[(V 2w(x,y)f+
A
2 ""\, /
+2(1-v)~(~l=.~if-w<x, Y))] dA
~~7
~0
dx2
()y2
U = 1 JD(()2w(x, y) + ()2w(x,y))2 dA
2A
dx2 ()y2
a4
\..
V'
\..
V'
Dn4c2
a/2
a/2
--2a2
Potential energy:
IT=U+U'
f"'MN'.Y .MI'.w.v.'WN'"'"'v
>MM
I IT=
8
:-:-x .. ..
4. ..
,.w......_....._.._.,,.......,,....,.....,"'"''.vMN>.-..a
n C2 2a2
FC
..
I
ot>-Jt.
L (Dn4 c2-Fc) = o
dC
2a2
n4Dc- F=O
a2
=- 0.216
mm
d)
Use the stress definition directly in terms of middle surface deflection: 2 a _ _ zE [ d w0 d2 w 0] x (1- v2) Jx2 + v dy2 The maximum ax (and cry) will be at the extreme fiber. At the top surface: 2 2 a _ h E [ d w0 d w0 ] xmax-- 2(1-v2) Jx2 + v Jy2 There will be compression on the top fiber and tension on the bottom.
ax
max
Proble111 2. Plate with distributed load @ Consider the same plate as in Example 1, but with a distributed load over a "footprint" of 200 mm by 400 mm, as shown. This is a common load, as in the case of a vehicle wheel.
The load totals 50 N as before. How are the maximum deflection and stress affected by the load distribution?
t 400mm
t
300mm
15.23
'
'
Solution Use a one-term potential energy solution. Assume: w(x,y) = Csin ~x sin'; Strain energy (as before):
4 2
U= 2a2
Work potential:
Dn C
'l1/ = .
fJp(x,y)w(x,y)dA
A
=-
400300
00]00
1C
1dxdy
1po
6jo
400
sin
-
dy MPa
~
'
300
~'
=-6.25x10-4 C 1 =-46.01C N
2:
~ ( c~s 0 . 6 1r -
cos 0 .4 1r)
(-1.1756)(-0.6180)
2 C=-46.01a2= -146.0N(1000mm) =-Q. 19 Smm n 4D (2.38x10 6 )n4 From this, both displacement and stress are 1 8% lower than for the concentrated load. v @
15.24
ProbleDl 3. Plate bending, Galerkin solution A two-term Galerkin solution is to be used for a uniform plate with concentrated load. All four edges are simply-supported.
2 w(x' y) =c 1sin nxsin ny+c 2 sin 2a nxsin 2a ny a a
~-----3~~4----~~~/
~4
_L_
"
__________ /
z
~1~~---------a----------~
Find the deflection under the load and show which assumed term contributes the most.
D\7 4 w( x, y) = p(x, y)
=- F8(x- 3 t)8(y- :)
.
:
sin ~
a
(j= 1,2)
0 0 i=l a/2 a
F8(x-
(j= 1,2)
0 0
( a4 +2 a2
ax4
Hence:
2
r2.
0
=- F
(j= 1,2)
i 3 n --v-(0 . 707 , -
1)
sin~= 1 2
where:
J
0
!!:_
8 ..
V
=4 a
Dc1 (
=F
The equations have uncoupled because the assumed modes are orthogonal.
25n4 Dc 1 =-0.707F 8a2
0.125 a2 F
0.00128 a 2F D
=0.707c 1 -c 2
=
O. 707 ( _
F)
0.00164 a 2F D
The first mode contributes 28% more deflection than the second mode.
I I I I I I I I I I I I I I I I
I I
LECTURE 16
REFERENCE
1. Zienkiewicz, 0. C., "The Finite Element Method," 3rd Ed., McGraw-Hill Book Co., 1977. 2. Hughes, T. J. A., "The Finite Element Method," Prentice-Hall, Inc., New Jersey, 1987 (pp. 2-9).
16.1
IOVERVIEWI
The Kirchhoff-Love and the Mindlin theories are the dominant plate formulations. The Kirchhoff-Love theory Oust reviewed) is the generalization of Euler-Bemoulli beam theory, with plane sections remaining plane and perpendicular to the neutral surface. The Mindlin theory allows shearing deformation. Continuity required for the neutral surface at inter-element boundaries is C 1 for K-L and Co for Mindlin. We will use Kirchhoff-Love theory to illustrate a plate element because it is the dominant theory used today for linear problems and because it is far easier to teach!
16.2
Use moment resultants and curvatures as "generalized" stress and strain. This allows calculation of strain energy as an area integral rather than a volume integral. Use 12 d.o.f.:
FI
Mx Ml yl
[G]
Mx My
[.l)]
F2
XX
{==
Mxy
Xy Xxy
{==
{w(x,y)}
8y4
My4
Generalized stress-strain:
() { Xy Xx} C>
IJxy Xxy
16.3
=~;(xi ,yl)
=q2+2q4xl+qsYt+3q7xt+2q8xlyl+q9yl +3qllxl + Y1
BY!=-~; (xl'yl)
As in beam
theory, the yz plane has a better sign convention than the xz!
6(A ..l
._,~~~.,....~~>...
-(~ ---~~-------
16.4
/f?Ovl .
~11
By4
- [
:~
ql2
The symbolic inversion of [H] was done by Zienkiewicz and Cheung. We now have all the mappings!
[k]
=I [B]T[G] [B] dA
A
= [Hf T
The stiffness for an orthotropic, 4 noded, rectangular plate (MZC) element is:
[K]
[L~
D1=
VD
16.5
Dxy = (1- v) D
60 0 0 30 0 20 30 0 15 0 0 0
60 0 0
(symm.)
15 0
JS=
:2
2
20 0 0
30 -15
(symm.)
0 15 0 0 0 30
0 K3= -30 15 0 0 0 0 0
84
-6 8 8 6 0 -84 6 -6 -2
K4=
-6 84
0
6 6 8
0 8
(symm.)
-6 84 6 6 84
-8 6 -6 84 -8 0 -6
0
6
2
84 -6
-6
0
-6
-2 -6
0 2
8 6 84
6 -84 -6 -6 -84 6
2
0
-6
-6 -8 0 -6 -2 0 6 8 6 0 -2 6 0 -8 -6 0 8
16.7
IPLATE STRETCHINGI
The in-plane stretching of plates is modeled by plane-stress elements. -
ITHIN SHELLS I
Examples: Automobile fenders Aircraft skins Pressure vessels Piping Many FEA programs have thin shell elements. (e.g. QUADS in MSC/NASTRAN). Analysts often prefer to use many flat plate elements (e.g. QUAD4 in MSC/NASTRAN) to create a faceted surface that represents the curved shell. Lateral loads are converted to membrane loads at intersections:
Q})
16.8
ITHICK SHELLS I
Most FEA programs suggest using solid elements. The 8-20 noded hexagonal element is best. Even the 8 noded (brick) element allows shear deformation:
r-----------,
c :
:l
For good accuracy, one needs about 7layers of 8 noded solid elements to model a thick shell, or 4 layers of 20 noded sold elements
16.9
IPROBLEM SESSIONI@
ProbleDl 1. Triangular plate bending elem.ent
Consider the creation of a triangular plate bending element that uses a complete quintic polynomial. How many nodes would you use, and what degrees of freedom would you use at each node? Solution The quintic (5th degree polynomial) has 21 terms:
+ qsxy+ q6y
3
3
4 4 5
+ ql7x y+ ql8x Y + q19x Y + q2oxy + q21Y One way is to use a 7 noded triangle with one interior node to be condensed out. Another way is to use an extra twisting d.o.f. at each of the vertex nodes (or each of the midside nodes). Olson and others used the latter. 3 3 2
+ql6x
=
dW 4 CtCJY ,
16.10
Problem. 2. Constraints on cantilever plate Consider a plate with stiffener. The plate elements use Kirchhoff-Love theory and the beam elements use Euler-Bemoulli theory. What are the single point constraints required at each node? No automatic single point constraint capability does this work for you. The wall acts as clamp. The stiffener is modeled with a standoff / distance from the plate; the plate nodes are adequate to z model the entire X 9 3 6 system. Solution The wall constrains all 6 degrees of freedom. The beam elements provide stiffness in all six degrees of freedom, and the stiffener cannot become a mechanism. The plates have the drilling singularity.
~
1 2
NODES
1
X
4
X
7
X X
X
X
X X X X
X X X X X
3
4
X X
5 6
X
16.11
I I I I
I I I I I I I I I I I I
LECTURE 17
LABORATORY 3 PLATE
PROBLEM STATEMENT
PHYSICAL MODELING
DATA SETS
RESULTS
17.1
IPROBLEM STATEMENTI
Consider a flat plate, cantilevered from a wall, with a support at midspan. The support is hinged to the plate, but the plate is continuous.
lOOON (225 lb)
t ~~========~~========~'
nt:n
17.2
..
JT
500mm (19.69 in)
mm
y{~~~-x----------------~~ ~
>,;.,..->
l--soo --1-soo -1
mm
IPHYSICAL MODELINGI
The material is thin and flat--use KirchhoffLove plate theory. o Loading is normal to the plate middle surface. o Assume small deflections and slopes. o Assume the material remains linearly elastic. ___..._The response will therefore be pure flexure (no tension/ compression of middle surface). e B. C. at the wall constrain all d.o.f. 0 B. C. at the midspan support constrain all d.o.f. except rotation about y axis. o There is a reflective plane, with antisymmetric loads.
a
17.3
IF. E. MODELINGI
Use 6 quadrilateral (QUAD4) elements. We won't exploit the reflective plane.
.:
~-
:::
3
t.:...,
9
:-:
12
:
5
X
8
7
11
10
~1
...
4
=
.'-:-~
Constrain d.o.f. 6 (drilling d.o.f.) at every node. Constrain 123456 at wall (nodes 1,2,3).
There is a "defect" in the Kirchhoff-Love theory; the plate has no stiffness in the drilling d.o.f. (rotation about axis normal to plate). 6
1drilling :o.
~
.,..f._ _ _ _ _..,..
file
manageme1~t
state111ent is:
a~re:
GRID,1,, 0.0, 0.0,0.0,,123456 GRID,2,, 0.0,250.0,0.0,,123456 GRID,3,, 0.0,500.0,0.0,,123456 GRID,4,,500.0, 0.0,0.0,,12346 GRID,5,,500.0,250.0,0.0,,12346 GRID,6,,500.0,500.0,0.0,,12346 GRID,7,,750.0, 0.0,0.0 GRID,8,,750.0,250.0,0.0 GRID,9,,750.0,500.0,0.0 GRID,10,,1000.0, 0.0,0.0 GRID,11,,1000.0,250.0,0.0 GRID,12,,1000.0,500.0,0.0 CQUAD4,1,25, 1, 4, 5, 2 CQUAD4,2,25, 2, 5, 6, 3 CQUAD4,3,25, 4, 7, 8, 5 CQUAD4,4,25, 5, 8, 9, 6
CQUAD4,5,25, 7,10,11, 8 CQUAD4,6,25, 8,11,12, 9 PSHELL,25,17,10.,17 MAT1,17,68950.,,0.30 FORCE,67,10,0,1.0,0.,0.,-1000. FORCE,67,12,0,1.0,0.,0., 1000. END DATA
RESULTS FOR MSC/NASTRAN .... .... v- .
~ ~ -
0 P 0 R C E S POINT T2 ID. Tl .o 1 .0 2 .0 .0 .0 3 .0 4 .o .0 .0 .0 5 6 .0 .0
S I N G L E - P 0 I N T
C d N S T R A I N T
R2 1.289496E+05 7.275958E-11 -1.289496E+05 .0 .0
.o
R3 .0 .0 .0 .0 .0 .0
17.6
.o
T3
R1 .0 .0 .0 .0 .0
.o
.o
.0 .0 .0 .0 .0 .0 .0 .0 .0 .0
.o
-4.479750E-14 1.056213E+Ol
The maximum displacements are at the tips under the loads: +10.56 mm.
S T R E S S E S ELEMENT
I N
QUADRILATERAL ( QUAD 4 ) PRINCIPAL MAJOR 5.188910E-02 5.605145E-01 5.605145E-01 5.188910E-02 5.183512E+00 2.797917E+01 2.797917E+01 5.183512E+00 2 .. 9301S9E+Ol 2.880151E+01 2.880151E+01 2 .. 930189E+01 STRESSES MINOR -5.605145E-01 -5.18891t>E-02 -5.188910E-02 -5.605145E-01 -2.797917E+01 -5.183512E+00 -5.183512E+00 -2.797917E+01 -2.880151E+Ol -2.930189E+01 -2.9301S9E+Ol -2.880151E+01
E L E ME N T S
FIBRE ID.DISTANCE NORMAL-X 1 -5.0 -3.912503E-01 5.0 3.912503E-01 2 -5.0 3.912503E-01 5.0 -3.912503E-01 3 -5.0 -1.873651E+01 5.0 1.873651E+01 4 -5.0 1.873651E+01 5.0 -1.873651E+01 5 -5.0 -8.426877E-01 8.426877E-01 5.0 6 -5.0 8.426877E-01 5.0 -8.426877E-01
VON MISES 5.881782E-01 5.881782E-01 5.881782E-01 5.881782E-01 3.089876E+01 3.089876E+01 3.089876E+01 3.089876E+01 5.031964E+01 5.031964E+Ol 5.031964E+Ol 5.031964E+Ol
17.7
IMARC DATAl
TITLE BILL ANDERSON, JIM BROWELL TITLE 18 JUNE 1993 TITLE LABORATORY 3: PLATE SIZING 400000 ELEMENTS 75 SHELL SECT 3 END
POST
14,1111
11 12 13 14 15 16
17
1 1 1 1 1 1
1
-~~'l ' On
A1ises stre,ys
11 12 13 14 15 16 17
3 3 3 3 3 3 3
~~-
POINT LOAD
2
COMMENT 0.0 10
THE FOLLOWING CARD SETS ARE FROM LOAD ID 0.0 -1000.000 0.0 0.0 0.0 1000.000 0.0 0.0
67
0.0 0.0
THE FOLLOWING 0.0 1 3 4 3 1 COMMENT THE FOLLOWING 0.0 0.0 1 2 4 3 4 5 6 CONNECTIVITY 0.0 2 2
6
COMMENT
CARD SETS ARE FROM BC ID 0.0 0.0 0.0 5 6 CARD SETS ARE FROM BC ID 0.0 0.0 0.0
6
1
2 3 4
5
6
75 75 75 75 75 75
1
2 4 5
5
6 8 9
2
3 5 6 8 9
5 7
8
7
8
10 11
11 12
COORDINATES 3 12
1 2
3
4
5 6
7 8 9
10 11 12
0.0 0.0 0.0 500.0000 500.0000 500.0000 750.0000 750.0000 750.0000 1000.000 1000.000 1000.000
0.0 250.0000 500.0000 0.0 250.0000 500.0000 0.0 250.0000 500.0000 0.0 250.0000 500.0000
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
17.9
ISOTROPIC
1
17 68950.00 1 TO GEOMETRY
1
0.300000 6
ELEMENT TYPE REQUESTED************************* NUMBER OF ELEMENTS IN MESH********************* NUMBER OF NODES IN MESH************************ MAX NUMBER OF ELEMENTS IN ANY DIST LOAD LIST*** MAXIMUM NUMBER OF BOUNDARY CONDITIONS********** LOAD CORRECTION FLAGGED OR SET***************** NUMBER OF LISTS OF DISTRIBUTED LOADS*********** STRESSES STORED AT ALL INTEGRATION POINTS****** TAPE NO.FOR INPUT OF COORDINATES + CONNECTIVITY NO.OF DIFFERENT MATERIALS 17 MAX.NO OF SLOPES MAXIMUM ELEMENTS VARIABLES PER POINT ON POST TP NtU~ER OF POINTS ON SHELL SECTION ************* OPTION FOR TERMINAL DEBUG********************** NEW STYLE INPUT FORMAT WILL BE USED************ MAXIMUM NUMBER OF SET NAMES IS***************** END OF PARAMETERS AND SIZING ************************************************* *************************************************
75
6 12 0 33
3
5 5 33 3
10
17.10
T 0 T A L
D I S P L A C E M E N T S
0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 1.25360E-02 1.56905E-02 1.25360E-02 1 ,,, r . l Ll. .:.4.29441E-02 .. u .1..... 0 '-- "" 11-7.37754E-08 4.18335E-02 c.. ru 4 1 ; 1 c} 0 4.29441E-02 f" .,; LJ
0.00000 0.00000 0.00000 4.53590E-05 7.13134E-12 -4.53590E-05 2.80873E-02 2.78225E-10 -2.80873E-02 2.85027E-02 -2.28548E-10 -2.85026E-02
@ REACTION FORCES AT FIXED BOUNDARY CONDITIONS, RESIDUAL LOAD CORRECTION ELSEWHERE FZ MX MY 1 -501.20 -3.5800 1.25286E+05 2 -7.62939E-05 -93.081 2.34375E-02 3 501.20 -3.5800 -1.25286E+05 4 1395.2 -2658.1 2.3203 5 3.95203E-03 -47564. 0.67969 6 -1395.2 -2657.0 -0.86719 7 9.94263E-02 -2.4609 -10.259 8 -1.86768E-02 -8.7617 -0.12891 9 -3.29590E-02 -6.2891 8.6796 10 -0.19495 -11.791 -12.568 11 1.39008E-02 -21.048 -0.80469 12 0.15045 -9.2317 9.5460
17.11
l :_, ___ .J _ _ ; :
''1
--
--.. -<.,J..,J
_,
("\ '"\
,,..'.
.1
-=
''
.t.
IASTROS DATA)
Use 6 QUAD4 elements. ~I'he solution control data set is:
ASSIGN DATABASE LAB3 SHAZAM NEW SOLUTION TITLE=PLATE PROBLEM 3. QUAD4 ELEMENTS. ANALYZE BOUNDARY STATICS ( MECH = 67 ) LABEL = QUAD4 ELEMENTS WITH => C::1000 N FORCES. PRINT DISP =ALL, STRE=ALL,~ <= FORC=ALL, SPCF=ALL END
17.12
BEGIN BULK GRDSET I I I I I I I 6 GRID 11 11 0.0 1 0.0 1 0.0 11 123456 GRID 12 11 0.0 1 250.0 1 0.0 11 123456 GRIDI31 I 0.01500.010.011123456
GRID 14 11 500.0 1
GRID 1 5 11 500.0 1 250.0 1 0.0 11 12346 GRID 1 6 1 ,500.0 1 500.0 1 0.0 11 12346 GRID 1 7 1 1 750.0 1 0.0 10.0 GRID 1 8 11 750.0 1 250.0 1 0.0 GRID 1 9 11 750.0 1 500.0 1 0.0 GRID 1 10 11 1000.0 1 0.0 1 0.0 GRID 1 11 11 1000.0 1 250.0 1 0.0 GRID 1 12 11 1000.0 1 500.0 1 0.0
CQUAD4 1 1 1 25 1 1 1 4 5 1 CQUAD4 1 2,25 1 2, 5 1 6 CQUAD4 1 3 1 25 1 4 71 8 CQUAD4 1 4 1 25 1 5 1 8 9 CQUAD4 1 5 1 25 1 7 1 10 1 11 1 CQUAD4 1 6 1 25 1 8 1 11 1 12 1 PSHELL 1 25 1 17 1 10. 1 17 MAT1 1 17 1 68950. 11 0.30
1
1
2 3
5
6 8 9
FORCE 1 67 110 1 0 11.0 10. 1 0. 1-1000. FORCE 1 67 112 10 11.0 10. 1 0. 1 1000.
ENDDATA
17.13
(?:)
"'..
- ...
T2 O.OE+OO O.OE+OO O.OE+OO O.OE+OO O.OE+OO O.OE+OO 1.6E-17 1.4E-17 1.5E-17 3.1E-17 3.1E-17 3.1E-17
T3 O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO -3.78832E+00 -6.82565E-13 3.78832E+00 -1.09757E+Ol -1.67155E-12
1~09757E+01
R1 O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO 1.33112E-02 1.68439E-02 1.33112E-02 4.45431E-02 4.27327E-02 4.45431E-02
R2 O.OOOOOE+OO O.OOOOOE+OO O.OOOOOE+OO 1.08191E-03 1.45457E-15 -1.08191E-03 2.86563E-02 3.64726E-15 -2.86563E-02 2.84602E-02 4.12344E-15 -2.84602E-02
,..,....... ..
~
%~~.9!
FZ
MY
1 -4.5139E+02 -8.5391E+01 1.1290E+05 2 -6.7188E-ll -2.2201E+03 1.2180E-08 3 4.5139E+02 -8.5391E+01 -1.1290E+05 4 1.3672E+03 -6.4551E+01 0.0000E+00 5 1.1959E-10 -3.9540E+04 O.OOOOE+OO 6 -1.3672E+03 -6.4551E+01 0.0000E+00
17.14
PRINCIPAL STRESSES MINOR MAJOR 5.43595E-02 -5.87200E-01 5.87200E-01 -5.43595E-02 5.87200E-01 -5.43595E-02 5.43595E-02 -5.87200E-01 6.05720E+00 -2.97383E+01 2.97383E+01 -6.05720E+00 2.97383E+01 -6.05720E+00 6.05720E+00 -2.97383E+01 -2.93527E+01 3.12423E+01 -3.12423E+01 2.93527E+01 -3.12423E+01 2.93527E+01 -2.93527E+01 3.12423E+01
!RESULTS
17.15
lz <-y 'x
PLATE PROBLEM 3. QUAD4'S. 1008 N TWIST FORCES
MAXIMUM DEFLECTION
W12
10.6 9. 15
10
'. l1
12
17.16
.._,
~ 1397
11395
l 1367
MSC/NASTRAN
MAR<~
128,900 N mm 125,286
.,.,. _ ...... ,... , .. _,. .. ~ ~ ../ ~,~~~ }~ ~
' . .:-. ,N,.,"~ ........ , ... .... J>
3204 2657
Zo~
- ~53,630
47,564
125,286
.112~9()()
65
MSC/NASTRAN
Iv1t\RC
17.17
From the MSC /NASTRAN solution and PATRAN postprocessing, the direct stress on the lower surface is, in MPa:
9.79
. 653 . - .653
-1.96
LECTURE
IS
COORDINATE SYSTEMS
OVERVIEW
EXAMPLES
COORDINATE SYSTEMS
A. B. C. NODAL COORDINATES ELEMENT COORDINATES MATERIAL COORDINATES
18.1
loVERVIEWI
FEA codes have a default coordinate system which is a right-hand Cartesian system. Users create additional coordinate systems to conveniently: o enter geometric data e apply boundary conditions e prepare material properties e exploit special dimensional properties
~.
~t,
(axisymmetry,
z
I
IEXAMPLESI
'
X
'z
Rectangular (Cartesian)
Cylindrical
~
Spherical Curvilinear
IcooRDINATE SYSTEMS I
A. NODAL COORDINATES
A nodal coordinate system is required to:
o
Each node has 6 d.o.f. at the outset. There are several universal notations:
u1, u2, u3,
e1' e2, e3
z
!
3,6
1, 2, 3, 4, 5, 6
~... 1,4
2,5
~:X
defined by the user for parts of the structure exploit special geometry features (e.g. holes) Cartesian (rectangular) Cylindrical z Spherical
z
+
basic
~... x
~y
X
local, rectangular
The "global" coordinate system is the union of all the local coordinate systems (ip. green).
l
0
i
~
{u}
global
18.4
B. ELEMENT COORDINATES
Each element type needs a special coordinate system to represent stresses, forces and section properties.
Yelem
C. MATERIAL COORDINATES
The material coordinate system is aligned with principal material directions in nonisotropic materials. It can be defined in relation to the basic coordinates or special local coordinates. NASTRANusage: material coordinates are defined on connection or property cards. y material HEXA defaults to element coordinates. QUAD4 and QUADS default to system aligned 't with Gl-G2 e side of element. G2
18.5
--.1.__
B. OVERVIEW
Notation: We use matrix notation, which suppresses the base vectors: {::} The full notation does not hide base vectors: ii = ul el + u2 e2 Translations: A translation of the base vectors does not change Cartesian components ui. Rotations: A rotation of the base vectors does change components ui in Cartesian, cylindrical and spherical components. @
18.6
C. ROTATIONAL TRANSFORMATION
l.t 1 ,
J5! {}t)t~.!
,c<;>s a
11 2 ~, ~ ..;.,~~n
,5 t(,. - , .1 (
In matrix form:
u {u2
1}
_
elem
[
\....
c~sa
-stna
v
sin a ]{u1}
cosa u2
/
global
[t]
This is done, node by node:
utA u2A ulB u2B
~.
t(aA)
- - - - r - - - 0
~
0
t(an)
~
~-
- - _..,.
t.
T
ulc
0 ~ - - - r
s.
global
18.7
{u}elem
= [T]{u}global
1
Forces are physical vectors that transform the same way as displacements:
{f}elem = [T]{J}global
E. STIFFNESS TRANSFORMATION
Consider equilibrium in element coordinates: {j}elem = [k] elem {U}elem Use the transformations for {f} and {u}:
[T]{J}global
= [k ]elem[T]{u }global
-1
= [T]T[k]elem[T]{u}global
' I
(8)
tY
Find the global stiffness matrix for this truss: u 2{lem. \..,.,.~u
lelem
Solution check The truss should be stiffer in the horizontal than the vertical direction:
~....,..~V.v.N'V'A..,..,...,.,..,..~
..........,...,.,..,,.....,.,.......,.....,...
-:::::-
knglobal
= T (0. 75)
EA
k22global
EA =-r(0 25 )
This checks our intuition. The solution and the constraints are in global coordinates. To recover element results: {u}elem
= [T(30)] {u}global
18.10
a difficult topic angle changes do not commute in 3-D there exist singular directions the problem is related to flight mechanics, where one compares body coordinates to a reference coordinate system z
o
e
t /y
~X
IPROBLEM SESSIONI
ProbleiD 1. Deflection of a derrick
A derrick is used to lift heavy loads. It is
modeled as a planar truss structure. Each boom has the property: EA = 10 10 N What is the elastic deflection of the tip of the derrick due to the tip load?
10 m
18.11
Solution Number the trusses so that their common node is the last one. The u2 included angles then fit the rule for mapping from element to global stiffnesses: The first element has elastic forces:
ul u2
u3
u4
10 8
0 0 0 0
us u6
Fl F2 F3 F4 0 -10,000
[ 7.286 5. 701
5 - 7.286 5. 701
= 0 . 241
===
mm
3 0 0 0
!Y !
l l t. .............~
18.13
matrix [t] relating the element system to the basic (underlying) system. B) Find the diagonal stiffness term k 11 at node A referred to the basic coordinate system. Solution
''ltl'lh'YV'....v.-.w.A..........~.........w.v.
affected by a rotation of the coordinate system 2 in -D: [cos a sin a OJ [t]= -s~a co~a ~ B) The stiffness is:
[k]b aszc . = [T{ [k] ezem [T]
'e:!J
~25
0 0 0
0 0 0 0
0 0 0
18.14
= T21 k22
= 10.85
elem T21
(1;)
tY
,___-~~x
==
~~3=.-.--
u3
y
L ...............~.
u2
~~38
~
18.15
Solution The stiffness in basic coordinates is related to the stiffness in element coordinates, where the parameter EA/L appears:
cos 2a cosa sina -cos 2a -cosa sina E4. cos a sina sin 2a -cos a sina -sin2a [k] . JL~ -cos 2a -cos a sin a cos 2a cos a sin a baste o. o 2 2 -cos38 stn38 -stn a cos a sina sin a
=9.28x10 6 N/mm
/@
18.16
LECTURE 19
GAUSS INTEGRATION
NUMERICAL INTEGRATION
A. TRAPEZOIDAL RULE B. SIMPSON'S RULE
RADAU AND
LOBATTO INTEGRATION
19.1
NUMERICAL INTEGRATION
Consider the integration of a real function of a real variable f(x): f(x) ~
Replace the common (Riemann) integral with a finite sum: /sample points
f(x)dx = .J wi f(xi)
i=l
19.2
(~~~~
"-weights
2) Differences f(xi) - f(x j) 3) Derivatives 4) f(xi) at specially chosen xi (not equally spaced Gauss integration is the fourth type.
A. TRAPEZOIDAL RULE
f(x)
Use n evenly spaced sample points:
Xn
+ + [f(xn_l) + f(xn)]
B. SIMPSON'S RULE
f(x)
x1
Xn
x2
xn
!GAUSS INTEGRATION!
e
e
,.,
b
;-
-1
ft(~)d~
_,..,.
-1
19.4
+1
+b
fc /;)d/; +c fc /;
-1
)d/; + ...
-1
(V
Use n sample points to get 2n (total) weights and abscissae as parameters. Use "defining equations" to find weights and abscissae. Get 2n equations in 2n unknowns.
---+1
-1
fcl)dS = w1JCS1)
-1
2 = w 1(1)
-1
-1
fc~)dS = w1JCS1)
0=
w1;1
The solution w 1 = 2,
~1
=0
leads to:
ft(S)dS
-1
=2f(O)
l-tef~~t~ss
!C S) = 3 + S
Kc,y:::
-1
= 2(3)
= 6 ......,_Gauss, exact
19.6
-1
The one-point formula underestimates! Conclusions For n sample points: 1) get exact results up to degree 2n-l 2) get approximate results for polynomials above degree 2n-l, for transcendental functions, and for experimental data.
... . . . . . .'(I(:O"""
" )" '"" ")'" ")O "<)'" " )IX,J( " )O "" ~~
@
@
f(~)
f(l)d~
-1
-1
-1
f(~)
19.7
-1
fg 2 dg = w g? +w2g; 1
1 2 J= 2 J= 2 j (3) ~ 3 = w1'='1 + w2'='2 ~
~-=-->.< .>1!
11!1! <~-
f(~)
< . . . . . . . . . . . . . . . !.
-1
-1
= 2;1Cw 1 -1)
Hence:
~=
w 2 =1
g? +(-g1)2
g2 = -0.577350~1
19.8
~~-;1-=-~-~-0--.5-7-7"""""""'350269
Hence:
E. NUMERICAL EXAMPLE
Consider Use Gauss integration to find the exact value of:
-1
I( S +i S
3
~ S)dS
1.0
0.5
0
f(~)
~~~~~~~ :~~:~~~~
~2
~1
-0.5
i<O. 577350)
ic -o. 577350)
_{c~3+l~2-~~)d~ =~4+A~3-?6~2l:
=i
There is no algorithm error in the two-point formula applied to a cubic, only round-off error.
........
-1
..,....,--
----
_ _. __
--,..,..
..,....,
ll\TONSYMMETRIC FORMI
Consider:
ff(x)dx
a
-~
x=
b; a+ b 2a~
= -1
x=b, ~=+1
-1
when x = a , ~
ff(x)dx
a
= Jf(x(~))b-ad~
_1
....., ,.
.........
XI
~
x2
19.11
11
i =1
@
e e
+ g~ +
;:3
+ q1J +
get 3 defining equations per sample point need 2 sample points for quadratic, 4 for cubic, 5 for quartic, 7 for quintic not done this way in practice
B. EXPEDIENT PROCEDURE
Use 1-D ideas repetitively. An example is:
JJcg
-1-1
+ 1]2 )dgd1]=
-1
J[ wi(g[ + 1] )]d1J
2
i=1
~I -~-~ I
~
= ~ wX ~ Wi(g?+ 17])]
J=l
2
I
1
~~ I
1
z=l
- _,- - t-
r--
2 L L wjwi( ;i2 + 11 j)
j=1 i=1
4-noded quadrilateral
4-noded rectangle
solid, hexagonal
@
19.13
IPROBLEM SESSIONI
Problelll 1. Two-point Gauss integration A polynomial
f(x)=l+x 3 -3x
4
is defined over the region [2, 10]. You wish to get an approximate value of the integral
J~(x)dx
2
Solution The solution will not be exact because it would take 3 sample points to exactly integrate a quartic (5 constants involved). The formula for the unsymmetric domain is:
f b; b;
a
f(x)dx=
a+
Xi=
b+a
2 +
I
b-aJ=
=6+4(-0.577350)
e I
I I
=3.690599
e
I
6
19.14
10
= -13,727.378
f(x 2 ) = 1 + 3. 690599 - 3(3. 690599)
3 4
= -505.28789
We now have all the required numbers:
jf(x )dx = 4(-13,727.378- 505. 28789)
2
= -56,930. 665
10
2
~ 4 3 5 jj(x)dx=x+lx -sx
2
=-57,476.800
-1
The defining equations need exactly integrate only the quadratic, cubic, quartic and quintic terms.
-1
-1
f(~)
-1
J~3d~=wl~i +w2~~
: ... : ....:. ... . . :. . '3 ':. :. : ""' . .. 3" "8 I0 = w1g1 + w2g2 I (2)
~ ~
-1
~
r
~
w1g14 + w2 -
~24l
~.VNN.J
(3)
-1
~~-. ~~il~~2~i
(4)
Assume symmetry:
19.16
Equation 1 yields:
(also w 2 = 9 )
.j
@Y
Hence:
f( -0. 7745967)
over the interval [-1, 1] using Gauss integration and 3 sample points. Is your answer exact or not? Why? Solution
1
The three-point solution will not be exact; four points are needed to exactly integrate a sixthdegree polynomial. @
19.17
~ ff
tt t
19.18
--~
Solution a) The shape functions for the two-noded Euler-Bemoulli beam are cubic. The stiffness integral is of the form:
Since N(x) is cubic, two derivatives are linear and the integration has a quadratic integrand. number of points= (polynomial degree+ 1)/ 2 = 1.5 (decimal) = 2 (integer) b) Because N(x) is cubic and the loading is quadratic, the integrand is quintic and 3 Gauss points are needed. f r
J [N]
0
{L(x)}dx
v'
ProbleiD 5. Lobatto integration You are asked to exactly integrate a series of cubic functions on the domain where ~ ranges from -1 to +1. Use Lobatto integration and develop the approach from first principles. Compare your results with Simpson's integration rule. f(~) Solution __. A cubic has 4 constants, hence +l 4 defining equations -1 are needed.
~-.. ..-..V.""-~*.JII."'.A.J'I'.YAWJ'....
....._---
s
@
2 3 /(S)=cl +c2S+c3S +c4S Lobatto integration uses both endpoints as sample points.
19.19
0=-w1
~=
+w3
+w3
w3
wl
2nd and 3rd equations give: 2nd equation gives: 1st equation yields: integration formula is:
=1 I 3 w1 = 1 I 3
w 2 =41 3
Llf(~)d~ =
LECTURE20
ISOPARAMETRIC ELEMENTS
20.1
=f(x 1) + f(x)(x-x 1)
I
=f(xl)
These two shape functions can exactly interpolate an arbitrary linear function f(x):
f(x}, ..
f(x)
_
N1
___..._~..____
7,
_
>.::,,.
,.,~.,: '<
N2
___,;,;,;,__ X
xl
x2
B. INTERPOLATE X
Consider the outrageous idea of interpolating the independent variable x:
x = N 1 (x)x 1 + N 2 (x) x2
= x(x 2 =
X
x 1)
x2- xl
20.3
Con1ments: ""
..
o o
o
obtained identity not as crazy as it seems concept generalizes good for cases with 2 sets of independent coordinates (mapping) this example is "isoparametric" can also get superparametric: geom. > field subparametric: geom. < field
I 1
~ ~
Define~:
J(1 + S)u
20.4
coord.:
fl
X]
B(x)]T [G(x)][B(x)] A dx
= f [B(x@){[G(x@)][B(x@)]IJIAd ~
-1
where:
1
[G]
= dx =
d~
- xl
+ X2
= X2- xl = L 2 2 2
= [E]
=
[D]
[A_] =
dx
G)
Shape functions:
= [i~J[~o- ~) = [Finally,
[K] =
1 J
~C1 + ~)]
i iJ
1 L [E]
-I
[-v. ] [-l
YL
-1
l] ~ Ad~
= EA[
L
1 -1]
1
20.5
where:
El( 1- v2)
0
0 G
[G] = Evi(J-v2) 0
[D]
= ~
ax ~
20.6
1
dx
dy
Invent a
(~,
II.;.
'-:-
Interpolate x, using
x
~,
11 as parameters:
=i~Nf(~, 1J)xi
=i'!:INi(~, 1J)Yi
N1 (1, 1)
=0 N 1(-1,1) = 0
20.7
N 1(x,y)
const = 4
NI(~,7J)=!(l-~)(1-7J)
1 0 4 (1 - ;)(1 - 17)
(1 +
~)(1 0
11)
xl
Y1
(1 +
~)(1 + 11)
0
(1- g)(l + 11)
:2
x2
Y4
These are the "serendipity" shape functions. Field variables are interpolated the same way: ux = LNi(~, 1J)ui
i=I,3,5
Uy =
LNi(;, 17)Ui
i=2,4,...
20.8
into:
-1-1
cc1r1st~ (l.ertv~lttves
We need
!x and Jy:
ax ag ax a11 ax
ag - dx ag + dy ag
a _ a ax a ay
d _ d dx + d dy
d1i - dX dri
dy dij
t~ =!gC;1Nicg.77)xi)
= .Lijtxi
z=1
~
4 ()N.
20.9
Hence, if
[J]-l exists:
~X = [1(g, ry)]
T-l
t,t;
d~
ay
a11
0
-I-1
hi dg + 122 dry
-1 Ta
-1 T
-1
121
dg + 122 dry
0
-1 Ta
1u dg + 112 dry
-1 Ta
-1 Ta
l (1 - g)(l - ry)
o
l (1 + g)(l - ry)
o
lo-g)CI-ry)
0
(1 - g)(l + 17)
20.10
...l
Numerically integrate:
Integrand(~' 11)
11
,... ""
IPROBLEM SESSIONI
Probletn 1. IsoparaiDetric Dlapping A rectangular finite element in two dimensions is given. It is to be numerically integrated using a mapping to the double unit square. A) To what (-1,3) ~ ~,~ (1,3) point (x,y) does y
c;.11) =co.s,o.2)
correspond? B) What is the value of I J I ?
x ~
(0.5,0.2)
(-1,-3 ) .w-~-~': (1,-3)
20.11
L Ni(;, 1J)xi
i
y(;, 17) =
L Nf(;, 1J)Yi
i
Hence
J
l
(1 +
~)(1-77)x2
11)X4
x(0.5,0.2) = 0.5
y(~, 17) =
l (1- ~)(1l
1J) y1 + (1 +
3
~)(1- 17)y2
= 0.6
(0.5,0.6)
A .....,..___
~- ~ ~
G--:18&---
yj
20.12
B) The IJacobian I has the interpretation of the area ratio (A 01ctl~ewl, which corrects for the change in coordinates in the integration. By inspection, this ratio is 3.0. Let us calculate IJ I in a formal way. dx dx
-dn
III=
~;
a;
d:
(Jij
dg =7dfxi
dx
dNi
(-1)(1-7])xl +
l
1
(1)(1- 1J)x2 +
(1 )(1 + 7])x 3+
=1+1+_.1_+_.1_ 4 4 4 4
=1
Next:
1 1 1 1 = -- (1-77)( -3) +- (1-77)( -3) +- (1 + 77)(3)-- (1 + 17)(3)
=
4 0
20.13
an -7 an xi
=! (-1)(1- g)x 1+ ~ (-1)(1- g)x2 + ~(1)(1+g)x3 + ~(l)(l+g)x4
=0
ax-~ aNi
ay
an=
=3.0 Finally,
7an Yi
g)y4
aNi
= !(-1)(1-g)yl+ ~(-1)(1-g)h
+ ~(1)(1+g)y3+ ~(1)(1+
III=
>
20.14
for this quadrilateral. The functions need only need be cubic (no higher). Solution A) No, the suggested displacement function has 10 generalized coordinates, and we only need 8, because there are only 8 nodes. One could arbitrarily drop out the q 6 and q 9 terms, or altematively, drop q 7 and q 8 . B) Consider a midside node. If one multiplies the equations for the three red lines shown, one has a product form for the shape 8 function.
C=)
@
This gives: N8 ( g, 77) = C(1- 7J)(1- g)(1 + 17) The constant is easily found to be 1/2.
Likewise
N7 (g, 71) = ~ (1 + g-77)(1-:- g)(1 + 77)
I
7
"'\=
11
I
0
/
:;.
II
~
I
~
~----~s~----~~~
7J+1=0
20.15
points
s
\ . .... .~: ;:,.,
0
'e
6-
2.4 3
,..~~v/1
0
-1
s ~
s~~nn
A) The goal of isoparametric mapping is to project a complicated element onto a simple domain where the shape functions are easy. Hence, map the point x = 2.4 onto ~ = 0.0, as well as mapping the endpoints to + 1.0.
x(~) = N1 (~)x 1 + N2 (~)x 2 + N3 (~)x 3
~c;-1)
......
<-'
So-
L,
20.16
. + O.
= 2.844
5773
2
(0.5773 + 1)(3. 0)
/
e c e o
-1
o~
20.17
I I
LECTURE21
SOLID ELEMENTS
HEXA ELEMENT
PENTA ELEMENT APPLICATIONS NEEDING CONSTRAINTS SUBSETS OF GLOBAL DISPLACEMENTS MULTIPOINT CONSTRAINT (MPC) CONCEPT RIGID ELEMENTS IN MSC/NASTRAN PROBLEM SESSION
1. 2. ENERGY IN SOLID ELEMENT CUBE MADE OF POROUS MATERIAL
21.1
TETRAHEDRON
WEDGE
HEXAHEDRON
TETRA (4-10)
PENTA (6-15)
21.2
HEXA (8-20)
{f}
SYM
Gss
Gs6
G66
SYM 0 0 0
0 0 0
0 0 0
G44
0 0
G55
G66
G)
v v
0 0 0
1-v
SYM
v
0 0 0
1-v
0
0
dx
0
0
dY
0
()
[D] =
dY dx
0
a a a
0
dz
0
()
di dY
dz
9x
21.4
e o
Four faces Four nodes Can be developed as an analog to Tumer's triangle. u(x,y,z) = q 1 + q 2x + q 3y +
v(x,y,z)
q4 z
The polynomials are "complete" linear forms. The elementary F.E. )heo:ry applies:
= j [B]T[G] [B] dv
vol
L Vi ui
21.5
D. HEXAHEDRON
Six faces o Eight node version o Eight- Twenty node version o lsoparametric formulation and numerical integration are common o Need enough integration points to avoid zero- energy deformation modes (Cook). o Need to recover shear stresses carefully to avoid error as solid approaches shell (MacNeal). e Bruce Irons' 20-noded isoparametric formulation is widely adopted.
0
E. ZERO-ENERGY MODES
"Elasticity" elements consider only translations (not rotations ) at nodes. Gauss points might lie at unfortunate spots:
~
'
v
I
I
I
LINEAR
LINEAR
PARABOLIC
21.6
[~EXA ELEMENT)
Transform to ~' 17, ~space:
::
==
z =;
L~x
.=:=~/Y
For 8 noded element: 1) Use 2x2x2 array of Gauss points 2) Evaluate shear strain at special shear points. Transfer values to Gauss points.
For 9 to 20 noded elements: 1) Use 3x3x3 array of Gauss points 2) Evaluate shear strains at 2x2x2 array of Gauss points and extrapolate to 3x3x3 array using the isoparametric shape functions for the 2x2x2 array.
>
3x3x3 Gauss points 2x2x2 Gauss points
21.7
"These procedures are necessary to relieve intemal constraints on the strains which destroy accuracy when the element is used to model thin shells.'' (R.H. MacNeal) The MSC /NASTRAN HEXA ELEMENT: 1) can use anisotropic materials (MAT9) 2) has heat transfer capability 3) has differential stiffness 4) provides stresses in the material coordinate system, which can be the element coordinate system (default), basic coordinates, or any other user defined MSC /NASTRAN coordinate system (Recommendation: don't use default!).
IPENTA ELEMENTI
o e e o
a companion of the HEXA has all of the HEXA capabilities has isoparametric formulation with numerical integration is mapped onto a right triangular prism. has two integration schemes 6 point (for 6 nodes) 9 point (for 7-15 nodes)
!'
l
. 1J
21.8
a shear correction is introduced to permit the distance between the two triangular faces to become small:
For the 6 noded PENTA 1) Use 6 Gauss points 2) Use 3 special points for transverse shear strain For the 7-15 noded PENTA 1) Use 9 Gauss points 2) Use 6 special points for Ez, 'Yyz, 'Yzx
@
2) Generalized plane strain
~::::::::~
3) Cyclic symmetry
',tit<
21.9
~~
us
= =}ua- }uf}un Ug
lis are removed by SPC (single-point constraint). lio are omitted by static or dynamic condensation. ur are suppressed to remove rigid body modes.
@ @
Typically, the first d.o.f. mentioned is eliminated in terms of the other d.o.f. Nonhomogenous constraints are created by using scalar points and setting them nonzero.
21.10
.....
Them-set of d.o.f. are eliminated. The n -set define the rigid body motion.
RBE 1 ELEMENT
o
o o
o
a general rigid constraint element allows connection of an arbitrary number of grid points requires ann-set controlling the 6 rigid body modes allows a very general distribution of n-set and m-set d.o.f.; lengthy input.
21.11
RBE2 ELEMENT
o
o o
a general rigid constraint element allows connection of an arbitrary number of grid points then-set is defined at only one grid compact data entry
RBE2 EID GN CM GMl GM2 GM3 GM4 GM5 +NEXT +NEXT GM6 GM7 GM8
RBE3 ELEMENT
o o
a weighted average constraint element specifies up to 6 d.o.f. (may be at different grid points) to be weighted averages of other d.o.f. These 6 d.o.f. are put in m-set (eliminated)
EXAMPLE (Schaeffer) Transfer loads from a reference grid point to grid point around the periphery.
21.12
!PROBLEM SESSIONI
ProbleiD 1. Energy in solid eleiDent An 8-noded hex element has all nodes but #4 held fixed. Node #4 is displaced 1 mm in the x direction only, requiring a live load:
4 Fx Fy 4 -{100} 50 N ..,C::,__---~ -::
F'z4
25
,z
JL!x
~ -= 1
___
2~
element at the equilibrium position? B) What is the virtual work done by the extemal forces during the virtual displacement? C) What is the virtual strain energy absorbed by the element during the virtual displacement? Solution The equilibrium equation is:
k10,10
k 11 '10 k 12,10
0 1 0 0 0
21.13
F1
100 50 25
F24
Solve the lOth, 11th and 12th equations: k 10 ,10 = 100 N lmm
k 11,10 = 50 N lmm
25 N lmm A) The strain energy is:
k 12,10
UII
=; {u}~
0
[k]{u}II
Oil
klO,lO
1
Oil
@ @
= Lo o -o.o1 o oJ 25
= (-0.01)(25)
=-0.25 N mm
F2
II-..:, III
~
II-..:, III
LlU
= -0.25
N mm
Il-7/II
21.14
A cube of porous material behaves as a linear elastic solid with zero Poisson's ratio. An experiment is carried out for the cube, clamped on the bottom surface. The results of the experiment are studied with an 8-noded finite element.
The cube is subjected to the loads shown. The only nonzero displacement is u 15 = 0.001 mm. The other loads shown on the top surface are those reactions needed to hold the vertices at zero displacement. There are no horizontal forces at any of the nodes, including the reactions at the base. Find as many ~j as possible.
lOOON lOON 200N 200N
0.001 mm
@
21.15
0 .001 0 0 0 0
!3
0 1000 0 -200 0 0
@
106 N/mm -10 5 N/mm
= lOON/0.001 mm =
5 k151s = k1s15 =-2x10 ' ' 5 k15 21 = k2115 = -10 ' ' 5 k15,24 = k 24 ,1 5 =-2x10
The zero reaction forces and zero applied loads lead to ki,IS = k 15 ,i = 0
(i = 1,2,4,5, 7,8, 10, 11,13,14,16,17,19,20,22,23)
LECTURE22
HYDRAULIC VALVE
A. B. C. D. E. F. G. H. I. CUSTOMER PHYSICAL PROBLEM COST AND TIME ESTIMATES PHYSICAL MODELING FINITE ELEMENT MODELING RESULTS TECHNICAL CONCLUSIONS MANAGEMENT CONCLUSIONS ADDITIONAL, LATER RESULTS
SCALING LAWS
A. PRESSURE LOADING B. GRAVITATIONAL LOADING
22.1
!HYDRAULIC VALYEI
A. CUSTOMER
HYDRECO
JIM GLIDDEN Vice Pres./Engr. LANSON BECKER Manufacturer of Hydraulic Pumps & Valves
22.2
B. PHYSICAL PROBLEM
/
/ /
/ /
/ /
/ /
o e
e
o
o
Valve housing is a single piece casting, made of gray iron. Housing has 3 bores. Used on heavy equipment (road graders, ... ) Casting is cracking during fatigue test. Fails at mold line (some offset present) Intemal pressure is 3000 psi. Want to extend to 5000 psi.
GOAL: Analyze housing and propose a redesign. Consider stresses and displacements
22.3
C. COST/TIME ESTIMATES
F.E.A.
FIRST CASE CALENDAR 1 mo. TIME TOTAL $5,000 COST ADDITIONAL CASE FIRST CASE
EXPERIMENTAL
ADDITIONAL CASE
1 wk.
$1 ,000
6 mo.
6 mo.
$10,000 $10,000
D. PHYSICAL MODELING
e
a o
e e o
a solid body linear, elastic material small displacements pressure loading 5 planes of reflective symmetry 2 moving planes of reflective symmetry
Moving plane
I ;
~
r <
'9A9h9A9A:
22.4
= 0.3
v = 0.3
o
The problem has been solved by both SAP6 and by MSC/NASTRAN. Explain MSC/NASTRAN. Use 8-20 noded HEXA elements (88). e Use PLOAD4 cards for pressure on exposed surfaces. e Use SPC on GRID cards for 3 fixed reflective planes of symmetry. a Use RBE2 cards for 2 moving reflective planes of symmetry. e Calculate unresolved forces (projected on lllirror image bodies) and apply with FORCE card on d.o.f. remaining in "n" set on RBE2 card.
22.5
22.6
HEXA elements: 88 1364 D.o.f.: 547 Grids: 1700 sec, before reseq. Decomp time: 1279 sec, after reseq. $175, commercial rates. CPU cost:
z
X
top view
~~ r\-
.J ~
..........,
!'...
~~
~,
~~ ~
v
r~~
end view
22.7
22.8
13,000 psi
(90 MPA)
22,600 psi
(156 MPA)
22,400
(154)
14,000
(97)
14,900
(103)
G. TECHNICAL CONCLUSIONS
0
e
e
0
Points of highest stress were found. They coincide with fatigue crack. Stress were higher than expected. Displacements were higher than expected. Certain column-like regions should be straightened:
BAD
GOOD
BETTER
H. MANAGEMENT CONCLUSIONS
0
e
0
Actual cost: $6200 Actual calendar time: 5 weeks Good technology. Hard to convert to design "rules of thumb" Had a good data base for further work.
I
I. LATER RESULTS
Runs cost as little as $450. e Changing to malleable iron doesn't change stresses. a Geometric scaling doesn't change stresses.
0
22.10
e o
Adding material to post helps. Adding material to top and bottom of housing is not cost effective. Some 14 runs were made in total.
ISCALING LAWSI
A. PRESSURE LOADING
Force- L2
~
t---L~
EB Ee
,...
2L
... I
Direct stress:
() =
p A
-22.11
L2 L2
Bending stress:
0'=
Me _ (PL)c _ L2LL L4 I I
B. GRAVITATIONAL LOADING
Force- L3 Direct stress:
cr = _E_ -
L3 L2
L4
C. CONCLUSION:
22.12
LECTURE23
LAB PROBLEM 4:
GEAR TOOTH
PROBLEM STATEMENT
PHYSICAL MODELING
DATA SETS
RESULTS
23.1
IPROBLEM STATEMENTI
Consider a steel spur gear, made from flat steel plate. The outside radius is 110 mm and the "pitch" radius is 100 mm. E = 207,000 MPa v = 0.3
---4~ 10 nun
23.2
What are the extreme von Mises stresses in a gear tooth under the following two loads?
Compare the stresses for the two cases. These load cases correspond to an aligned pair of gears, and a misaligned pair. Also, when gears are. new, they may need to "wear in."
--J~!t~Jl-- --circle
98 .s 96] 94 8 92 ~ 90 88
12 10 8 6 4 2 0
y coordinate, mm
23.3
86
IPHYSICAL MODELINGI
e
The gear tooth is a three-dimensional body, because the three dominant dimensions are almost equal. Treat it as a solid body. loads are in the y direction.
Extend the model into base of tooth, beyond expected high stress region. Consider the base of the tooth as clamped. Assume linear elastic material and small deflections. Check this later.
IF. E.
MODELINGI
Use two solid elements. Estimate accuracy by comparing nodal stresses from adjacent elements.
23.4
Equivalent nodal loads are found by integrating along the line load path:
{!}=
f [N(99.6,-1.8,z{ 6~0 dz
-5
fl
!44
f47
0 0
0 0
-
1000
0 0 4000 0 0 1000
lOOON
4000N
!so f6o
IMSC/NASTRAN DATAl
G12
_..o
__A-
G4
G3
Connectivities:
3,5,7,1, 5,17,19,13, 4,6,8,2, 110,11,12,9, 16,18,20,14 15, 17, 19, 13, 7,29,31,25, 6, 18,20, 14,*2,23,24,21,t8,30,32,26
23.5
GRID,8,,86.50,12.70,0.0,,123456 GRID,9,,93.00,8.70,5.00 GRID,l0,,93.00,-8.70,5.00 GRID,ll,,93.00,-8.70,-5.00 GRID,l2,,93.00,8.70,-5.00 GRID,l3,,99.60,7.80,5.00 GRID,l4,,100.00,0.0,5.00 GRID,l5,,99.60,-7.80,5.00 GRID,l6,,99.60,-7.80,0.0 GRID,l7,,99.60,-7.80,-5.00 GRID,l8,,100.00,0.0,-5.00 GRID,l9,,99.60,7.80,-5.00 GRID,20,,99.60,7.80,0.0 GRID,21,,105.00,5.70,5.00 GRID,22,,105.00,-5.70,5.00 GRID,23,,105.00,-5.70,-5.00 GRID,24,,105.00,5.70,-5.00
GRID,25,,110.00,3.50,5.00 GRID,26,,110.00,0.0,5.00 GRID,27,,110.00,-3.50,5.00 GRID,28,,110.00,-3.50,0.0 . GRID,29,,110.00,-3.50,-5.00 GRID,30,,110.00,0.0,-5.00 GRID,31,,110.00,3.50,-5.00 GRID,32,,110.00,3.50,0.0 CHEXA,l,30,3,5,7,1,15,17,+Al +Al,l9,13,4,6,8,2,10,11,+A2 +A2,12,9,16,18,20,14 CHEXA,2,30,15,17,19,13,27,29,+Bl +Bl,31,25,16,18,20,14,22,23,+B2 +B2,24,21,28,30,32,26 PSOLID,30,31,0 MAT1,31,2.07E5,,0.3
23.7
The data were entered on an Apollo 3500 workstation and the run submitted through a network to a Hewlett-Packard 700 workstation. Results for displacements follow.
SUBCASE 1
D I
S P L A C E M E N T
V E C T 0
13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
6.925300E-03 8.207047E-03 1.078616E-02 1.081868E-02 1.078616E-02 8.207047E-03 6.925300E-03 6.963037E-03 1.053944E-02 1.107019E-02 1.107019E-02 1.053944E-02 1.273195E-02 1.317402E-02 1.293621E-02 1.279123E-02 1.293621E-02 1.317402E-02 1.273195E-02 1.276665E-02
23.8
1.089157E-04 3.226746E-04 5.329577E-04 -3.848918E-17 -5.329577E-04 -3.226746E-04 -1.089157E-04 -7.930044E-17 -6.990795E-07 2.797945E-04 -2.797945E-04 6.990795E-07 1.981639E-04 -9.562496E-05 -1.351982E-04 -1.837723E-16 1.351982E-04 9.562496E-05 -1.981639E-04 -2.255743E-16
SUBCASE 2 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
D I S P L ACEMENT
VECTOR -4.030569E-03 -1.422255E-03 1.032596E-02 7.308815E-03 6.180907E-03 -1.031911E-03 -5.690398E-03 -4.662068E-03 -6.341968E-03 4.727034E-03 3.447913E-03 -6.600303E-03 -6.569131E-03 -2.816546E-03 -2.435198E-03 -7.938867E-04 -1.471236E-04 -3.153600E-03 -6.588367E-03 -6.830011E-03
-3.037014E-03 -9.022470E-04 4.127733E-03 1.983507E-03 -3.469355E-03 -5.559830E-04 2.247461E-03 3.452979E-03 1.629195E-03 1.188325E-03 -7.570790E-04 -2.100891E-03 -1.113537E-03
7.160452E-03 1.387754E-02 2 .. 594546E-02 9.509350E-03 7.340994E-04 3.647954E-03 4.210754E-03 7.545148E-03 1.421410E-02 1.682306E-02 5.926699E-03 5.796329E-03 1.709548E-02 1.744876E-02 1.552363E-02 1.308611E-02 6.501771E-03 . 7.233304E-03 8.689580E-03 1.211220E-02
SUBCASE 1 STRESSES IN HEXAHEDRON SOLID ELEMENTS ------CORNER POINT STRESSES--------ELE GRID NORMAL PRINCIPAL VONMISES MENT
1 15 17
X y
z
X
z
2 15
X
3.247527E+01 -1.019972E+02 1.207871E+00 3.247527E+01 -1.019972E+02 1.207871E+00 -6.527668E-01 -9.447925E+01 -6.475156E+00 -6.527668E-01 -9.447925E+01 -6.475156E+00
A 6.460155E+01 1.725774E+02 B -1 .. 312817E+02 C -1.633880E+00 A 6.460155E+01 1.725774E+02 B -1.312817E+02 c -1.633880E+00 A 1.337905E+01 1.117319E+02 B -1.073635E+02 c -7.622745E+00 A 1.337905E+01 1.117319E+02 B -1 .. 073635E+02 c -7.622745E+00
z
17
X y
Average minimum principal stress: -119 MPa Average von Mises stress: 142 MPa
23.9
SUBCASE2 STRESSES IN HEXAHEDRON SOLID ELEMENTS ------CORNER POINT STRESSES--------VONMISES ELE GRID NORMAL PRINCIPAL MENT
m
1 15 17 2 15 17
X
y
z
X y
z
X y
z
X y
-9.783768E+00 -4.481932E+02 2.662371E+01 5.878863E+01 1.483216E+02 6.972133E+01 -2.824837E+01 -3.729665E+02 4.365237E+01 3.985226E+01 1.231419E+02 5.648651E+01
1.531718E+02 6.305997E+02 2.186662E+02 1.919591E+02 1.170095E+01 4.646441E+01 7.118484E+01 4.959503E+02 -4 . 484840E+02 1.973675E+01 1.923505E+02 2.038985E+02 -4.303397E+01 7.016408E+01
B -5.499731E+02
c -3.455197E+01
A
c
A
c
A
Average minimum principal stress: -499 MPa Average von Mises stress: 563 MPa
IMARC DATAl
Use two #21 solid elements. The nodes are connected in this sequence:
G8 G5
. (}15
(119
Gl2~
.-e
G4
G3
23.10
BILL ANDERSON, BACHAR ALJUNDI 2 JULY 1993 LABORATORY 4: SPUR GEAR TOOTH
400000
21
1
11 12 13 14 15 16 9
PRINT NODE
1
THE FOLLOWING CARD SETS ARE FROM LOAD ID 1000.000 0.0 0.0 0.0 4000.000 0.0 0.0 0.0
21 0.0 0.0
0.0 1000.000 0.0 0.0 0.0 0.0 17 FIXED DISP 1 COMMENT THE FOLLOWING CARD SETS ARE FROM BC ID 100 0.0 0.0 0.0 1 2 3 1 2 4 3 7 5 6 8 CONNECTIVITY 2 21 1 7 3 5 1 15 17 19 13 4 6 ::::> <: 8 2 16 18 20 14 10 11 12 9 17 21 2 15 27 19 13 29 31 25 16 18 ::> c::::: 2 0 14 28 30 32 26 22 23 24 21
23.11
COORDINATES 32 6 1 86.50000 2 86.50000 3 86.50000 4 86.50000 5 86.50000 6 86.50000 7 86.50000 8 86.50000 9 93.00000 10 93.00000 11 93.00000 12 93.00000 13 99.60000 14 100.0000 15 99.60000 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 99.60000 99.60000 100.0000 99.60000 99.60000 105.0000 105.0000 105.0000 105.0000 110.0000 110.0000 110.0000 110.0000 110.0000 110.0000 110.0000 110.0000
12.70000 0.0 -12.70000 -12.70000 -12.70000 0.0 12.70000 12.70000 8.700000 -8.700000 -8.700000 8.700000 7.800000 0.0 -7.800000 -7.800000 -7.800000 0.0 7.800000 7.800000 5.700000 -5.700000 -5.700000 5.700000 3.500000 0.0 -3.500000 -3.500000 -3.500000 0.0 3.500000 3.500000
23.12
5.000000 5.000000 5.000000 0.0 -5.000000 -5.000000 -5.000000 0.0 5.000000 5.000000 -5.000000 -5.000000 5.000000 5.000000 5.000000 0.0 -5.000000 -5.000000 -5.000000 0.0 5.000000 5.000000 -5.000000 -5.000000 5.000000 5.000000 5.000000 0.0 -5.000000 -5.000000 -5.000000 0.0
ISOTROPIC
1
31 206800.0 1 TO GEOMETRY
1
0.300000 2
COMMENT THE FOLLOWING CARD SETS ARE FROM LOAD ID 22 0.0 6000.000 0.0 =>
c::::=o.o
o.o
o.o
15
CONTINUE
21 22 23 24 25 26 27 28 29 30 31 32
-2.23730-03 1.05125E-02-6.17717E-06 3~09846E-03 1.10322-02 2.58904-04 ~3 . ,. c_:::::~:~=E -<1: 61~~ -- 03 1.10322-02-2. 58900E-04 -2.23730-03 1.05125E-02 6.17642-06 -1.28404E-03 1.27212E-02 1.10437E-04 6.02995E-04 1.30845E-02-7.31406E-05 2.07415E-03 1.28662E-02-1.02124E-04 1.81989E-03 1.27444E-02 1.58906E-10 2.07415-03 1.28662E-02 1 .02125-04 ~OP4~~ ~ . .).t~J- rJL 7 31427E - 05 6.02996 -04 J. ~~ -1.28404-03 1.27212-02-1.10438-04 -1.26696E-03 1.27143E-02-2.14446E-10
SUBCASE 2 T 0 T A L 15 16 17 22 23 27 28 29
D I S P L A C E M E N T S y z X ;84 l ~ C~ r)9.85038E-03 }.t:.: -- ;._ ~ 3.48013E-03 2 . ~ 3.61965-03 9.56697E-03 7.12179-03 3.28183-03 1.35515E-03 5.99426-03 3.79613E-03 1.71328E-02 4.99996E-03 2.02774E-03 5.86051E-03 3.56897-03 3.16523-03 1.59602E-02-1.06632E-03 1.69326-03 1.28264E-02-1.07498E-04 9.46777-04 6.82969E-03 2.02396E-04
23.14
12 13 14 15 16 17 18 19 20 21 22 23 24
78.616 111.52 111.52 78.616 42.296 57.485 104.35 95.291 104.35 57.485 42.296 40.576 19.219 37.624 37.624 19.219
-73.415 -24.861 -24.861 -73.41 -40.770 -57.519 -94.746 -91.004 -94.746 -57.519 -40.770 -38.652 -5.0742 -31.129 -31.129 -5.0742
-.78602 3.6316 3.6316 -.78602 -3.3889 -4.1354 -6.6263 -6.6847 -6.6263 -4.1354 -3.3889 -1.1705 2.7987 3.2493 3.2493 2.7987
10.063 98.142 98.142 10.063 5.1412 3.3370 20.486 12.377 20.486 3.3370 5.1412 4.4351 16.831 9.0640 9.0640 16.831
NODAL STRESS-MISALIGNED
mises p r i n c i p a 1 v a 1 u e s intensity minimum intermediate maximum
9
100 .. 39
274~69
-63~610
-150~97 -24~130
29~307
42~877 165~96
10 11 12 13 14 15 16 17 18
19 20 21 22 23 24
-3 .. 3503
44~839
123 ..
102~71
-114~55
-52 .. 805
-32~210
102 .. 23
181~54
-104 .. 10
-216~85
-321~55
356.72
131~40
58~169
43~711
68 .. 181
59~949
-34 .. 320
-20~037
51 .. 904 131 .. 54
46 .. 531
44~186
-16,.171
23.15
19,.034
IASTROS DATAl
Use two IHEX2 elements with connectivity:
GIS GI3
Gil
G5 G3
23.16
GRID,l,,86.50,12.70,5.00,,123456 GRID,2,,86.50,0.0,5.00,,123456 GRID,3,,86.50,-12.70,5.00,,123456 GRID,4,,86.50,-12.70,0.0,,123456 GRID,5,,86.50,-12.70,-5.00,,123456 GRID,6,,86.50,0.0,-5.00,,123456 GRID,7,,86.50,12.70,-5.00,,123456 GRID,8,,86.50,12.70,0.0,,123456 GRID,9,,93.00,8.70,5.00 GRID,l0,,93.00,-8.70,5.00 GRID,ll,,93.00,-8.70,-5.00 GRID,l2, ,93.00,8.70,-5.00 GRID,l3,,99.60,7.80,5.00 GRID,l4,,100.00,0.0,5.00 GRID,l5,,99.60,-7.80,5.00 GRID,l6,,99.60,-7.80,0.0
GRID,l7,,99.60,-7.80,-5.00 GRID,l8,,100.00,0.0,-5.00 GRID,l9,,99.60,7.80,-5.00 GRID,20,,99.60,7.80,0.0 GRID,21,,105.00,5.70,5.00 GRID,22,,105.00,-5.70,5.00 GRID,23,,105.00,-5.70,-5.00 GRID,24,,105.00,5.70,-5.00 GRID,25,,110.00,3.50,5.00 GRID,26,,110.00,0.0,5.00 GRID,27,,110.00,-3.50,5.00 GRID,28,,110.00,-3.50,0.0 GRID,29,,110.00,-3.50,-5.00 GRID,30,,110.00,0.0,-5.00 GRID,31,,110.00,3.50,-5.00 GRID,32,,110.00,3.50,0.0
23.17
ec .;;
y
J.. ;::;
. : .-.
CIHEX2,1,30,1,2,3,4,5,6,+Al +Al,7,8,9,10,11,12,13,14,+A2 +A2,15,16,17,18,19,20 CIHEX2,2,30,13,14,15,16,17,18,+B1 +B1,19,20,21,22,23,24,25,26,+B2 +B2,27,28,29,30,31,32 PIHEX,30,31,0 MAT1,31,2.07E5,,0.3 FORCE,21,15,0,1.0,0.,1000.,0. FORCE,22,15,0,1.0,0.,6000.,0. FORCE,21,16,0,1.0,0.,4000.,0. FORCE,21,17,0,1.0,0.,1000.,0. END DATA
ASTROS results:
**SUMMARY OF AUTOMATIC RESEQUENCING** METHOD SELECTED NONE CRITERION RMS WAVEFRONT AFTER BEFORE 20 20 BANDWIDTH 384 384 PROFILE 20 20 MAXIMUM WAVEFRONT 12.000 12.000 AVERAGE WAVEFRONT 13.153 13.153 RMS WAVEFRONT 32 NUMBER OF GRID POINTS 31 MAXIMUM NODAL DEGREE ELEMENTS PROCESSED 2 CIHEX2 2 TOTAL ELEMENTS **NO RESEQUENCING OF GRID-POINT DATA**
23.18
SUBCASE 1 D I S P L A C E M E N T T2 POINT ID. T1 6.87901E-03 13 -2.99627E-03 8.20150E-03 14 1.81333E-04 1.07336E-02 15 3.62362E-03 1.07588E-02 16 3.77 4 2BE - 03 1.07336E-02 17 3.62362E-03 8.20150E-03 18 1.81333E-04 6.87901E-03 19 -2.99627E-03 6.93078E-03 20 -3.07543E-03 1.05023E-02 21 -2.23514E-03 1.10215E-02 22 3.09547E-03 1.10215E-02 23 3.09547E-03 1.05023E-02 24 -2.23514E-03 1.27089E-02 25 -1.28281E-03 1 . 30718E-02 26 6.02412E-04 1.28537E-02 27 2.07215E-03 1.27321E-02 28 1.81814E-03 1.28537E-02 29 2.07215E-03 30 6.02412E-04 1.30718E-02 31 -1.28281E-03 1.27089E-02 1.27020E-02 32 -1.26574E-03
VECTOR
T3 1.00735E-04 3.18485E-04 8.96093E-17 -3.18485E-04 -1.00735E-04 -2.65087E-17 -6.16896E-06 2.58649E-04 -2.58649E-04 6.16896E-06 1.10332E-04 -7.30709E-05 -1.02024E-04 1.41570E-16 1.02024E-04 7.30709E-05 -1.10332E-04 5.08379E-17
SUBCASE 2 D I S P L A C E M E N T POINT ID. T1 T2 7.49445E-03 13 -1.72820E-03 14 8.52527E-04 1.38828E-02 15 3.47678E-03 2 .. 48169E-02 16 3.61616E-03 9.55771E-03 17 3.27866E-03 1.35384E-03 18 -4.61217E-04 3.54839E-03 19 -4.12347E-03 3.96428E-03 20 -2.99420E-03 7.45384E-03 21 -7.41248E-04 1.41999E-02 22 3v79245E-03 1.71162E-02 23 2.02578E-03 5.85485E-03 24 -3.38530E-03 5.69057E-03 25 3.65341E-05 1.68888E-02 26 2.13585E-03 1.73123E-02 27 3.16216E-03 1.59448E-02 28 1.69162E-03 1.28140E-02 29 9.45870E-04 6.82309E-03 30 -6.22668E-04 7.46013E-03 31 -2.23698E-03 8.63793E-03 32 -1.20099E-03 1.21780E-02
23.19
VECTOR
T3 -4.61180E-03 -1.16420E-03
7.11491E-03 5.98848E-03 -9.03632E-04 -6.03393E-03 -5.19854E-03 -6.21537E-03 4.99514E-03 3.56554E-03 -6.56663E-03 -6.54059E-03 -3.05085E-03 -1.06527E-03 -1.07370E-04 2.02217E-04 -3.09845E-03 -6.38599E-03 -6.61780E-03
.,.....,..~-...
{ ?. --:/} SUBCASE 1: STRESSES IN 20-NODED SOLID ELEMENT ( IHEX2 ) MEAN OCTAHEDR.Ai:::.:::..:>' ELEMENT STRESS --CENTER AND CORNER POINT STRESSES ID. POINT NORMAL PRINCIPAL STRESS SHEAR STRESS 1 15 X 8.805312E+00 A 4.310242E+01 3.653325E+01 7.868285E+01 Y -1.113043E+02 B -1.436572E+02 z -7.100762E+00 c -9.045020E+00 1 16 X 1.171149E+01 A 4.200758E+01 3.458708E+01 7.705680E+01 y -1.097193E+02 B -1.400154E+02 z -5.753422E+00 c -5.753427E+00 1 17 X 8.805218E+00 A 4.310225E+01 3.653337E+01 7.868287E+01 y -1.113045E+02 B -1.436573E+02 Z -7.100868E+00 c -9.045044E+00
2 2 2
A 1.281254E+01 B -9.698936E+01
c c
A 3.344736E+00 B -1.015158E+02 A
B
(;}~[~~~) SUBCASE 2: STRESSES IN 20-NODED SOLID ELEMENT (IHEX2) .................. ELEMENT STRESS OCTAHEDRAL --CENTER AND CORNER MEAN POINT STRESSES 1 15 X -3.017008E+01 A 1.629278E+02 1.473530E+02 2.916142E+02 y -4.221561E+02 B -5.377648E+02 z 1.026724E+01 c -6.722198E+Ol 1 16 X 5.033059E+01 A 1.533314E+02 -2.562039E+01 1.027524E+02 y -5.254855E+01 B -9.827264E+01 z 7.907909E+01 c 2.180239E+01 1 17 X 3.595124E+01 A 1.388956E+02 -6.883413E+01 5.493555E+01 y 1.124937E+02 B 4.727376E+00 z 5.805749E+01 c 6.287943E+01
2
2
1.264328E+02 9.834150E+01 2.283589E+02 B -4.115497E+02 c -9.907625E+00 A 1.135053E+02 -1.994841E+01 7.989239E+01 B -8.168916E+01 c 2.802911E+01 A 1.041717E+02 -5.047683E+01 4.296539E+01 B -1.001492E+00 c 4.826026E+01
23.20
IRESULTSI
The mesh is:
Note that the maximum displacement on the aligned load case is not under the load.
23.21
0.01317
0.013fJ8
Good agreement! The maximum displacement is higher for the misaligned teeth by a factor of 1.9!
The von Mises stress for the aligned gear teeth is (MSC/NASTRAN & I-DEAS):
23.22
The von Mises stress for the offset contact is shown (MSC /NASTRAN and I-DEAS ):
-450.58
~99. ~
3:3? . 59
276.1214
~ ~:~: I
@
I
:-::::
~ :~~~
460.66
~99 . ~
:3~7 . 59
I-DEAS provides discrete contour drawings; misaligned case, von Mises stress:
276.~ I'
21-4.49
152 . 94
.: .:JI
23.23
misaligned load
563
357
142
104
The maximum von iVIises stress is higher for the misaligned teeth by a factor of 3. 7! We must not "chase" the singular stress!
23.24
LECTURE24
THERMAL STRESS
OVERVIEW THERMAL EXPANSION IN ROD
THERMAL STRAIN
A. 3-D SOLID B. 2-D PLANE STRESS C. 2-D PLANE STRAIN
NUMERICAL EXAMPLE
A. B. C. D. E. F. 1WO-NODED ROD STRESS-STRAIN LAW EQUATIONS OF EQUILIBRIUM EQUIVALENT NODAL LOADS FREE EXPANSION FULLY CONSTRAINED CASE
PROBLEM SESSION
1. THERMAL STRESS IN ROD 2. THERMAL DISPLACMENT IN SYSTEM 3. EQUIVALENT NODAL LOADS IN TRIANGLE
24.1
ITHERMAL STRESS I
Ten1perature causes: - n1echanical expansion - changes in material properties Ten1perature changes occur in - high speed flight -engines - catalytic converters Treat as initial strain or initial stress:
{a}= [G]({}- { 0 })- {a0}
24.2
Identify {c0} as the strain caused by free thermal expansion. Example: Line element (ROD)
{e0 (x)} ={a(x)~T(x)}
.
COOL HOT
Thermal strain is positive when ilT is positive. Easy! Whether the temperature field causes any stress depends on the temperature distribution and b. c.:
,.,
"rD
~
.;:
rr;:
:r.
...
24.3
!THERMAL STRAIN!
A. 3-D SOLID
For material coordinates in principal directions:
Zmat
Ex 0 E
{eo}=
Yo Ez o rxyo
Xffiat
~z o r zxo
0 0 0
t;r
24.4
k .......~x
ORTHOTROPIC SOLID
Ex 0
{eo}=
cy
a 1LiT a 2 LiT
0
Ymat
~at
Yxy
C. 2 - D PLANE STRAIN
a 1 ~T
= (1 + v) a 2 b,.T
0
This strain is greater than in plane stress. The body is captured in the z direction and thermal expansion "spills" over into x andy directions (Poisson ratio effect).
24.5
ex {eo}= ey
rxy
,...,
fi 1 11T
in the case of plane stress. Get thermal shear strain. This is so awkward that nobody does it!
AlurninuiTI:
2.2lxl0 mm/mmoc
.
24.6
MPal({s}-{2.21 x10
~
})
ailT
C. EQ'NS OF EQUILIBRIUM
/1} =f { f
2 enl
[B]T[G]{0}dV
Eo=
t[-~}E]{aLlT}Adx
150,000
={-EAailT} EAailT
={-150,000}
150,000 N ,..
f__ 1 _ ____,'l
24.7
... 150,000 N
!2
/ 1}
The equivalent nodal loads cause the saine free mechanical expansion as the heat would have.
crx
24.8
The element is fixed at both ends and heated. Solve for mechanical reactions at ends:
{;J
Hence
150,000 N
+ {-
-: ]{
~}
@
....(_)_ _ _ _ _ 1. .
150,000N
Note: The ten1peratures are used at the start of the problen1, to calculate equivalent nodal loads. They are used again, at the end, to recover stresses.
24.9
!FUSION REACTOR!
There are three major ways to control fusion: Pellet fusion (Inertial confinen1ent) Tokamak (toroidal magnet) Hybrid The pellet approach was pioneered by KMS Fusion, Ann Arbor, MI. A 10 m spherical chamber was designed:
@
24.10
Ef
The problen1 became a mechanical wave propagation problem. Wave speed was found. Reflected waves cause tensile forces.
ICOMMENTSI
A. SPECIFYING TEMPERATURE
One can specify ten1perature in elen1ents or at nodes. The result is different:
T
~specified
at nodes
~specified
in elements
1::::::::::::::::::::::::::::1::::::::::::::::::::::::::::1::::::::::::::::::::::::::::1
24.11
B. MESH GENERATION
Does one refine the mesh in areas of high temperature gradient, or stress gradient, or both? Probably need to do both! Always need fine mesh in areas of high stress gradient. The temperature gradient creates a strain gradient, but this may or may not create a stress gradient.
111111111
Analysts often use a thermal load to represent a preload, such as in a pretensioned bolt. Typically, such a bolt is in tension, and applies compression to a relatively stiff structure.
The analyst makes the bolt "cold" to provide the tension. The structure compresses, relieving some of the prestress. An adjustnient in telllperature is needed, in a 2nd con1puter run.
24.12
IPROBLEM SESSIONI@
Problein 1. Thennal stress in rod A rod is constrained by two rigid walls. If the left half of the rod is heated 100 C and the right half is cooled 150 C, what is the stress distribution in the rod?
~
~
+100 C
-150 C
~
~
'"""'' '"'"' ' ' ' ' '"' ' ' '"' ' ' ' ' ' ' ' ' ' ' Hii!iH!i!i'>i'' "' ii!i!i!;!i!i!i!i!i!i!i!i'>!i!i!Hi
E==10 5 MPa
a == 1o-5 1 I C
k == 4. 0 x 10 3 J I mm sec C
Model the system with 2 rod elements, using equivalent nodal loads due to temperature.
ul
~;::::: .
- - - u2
~--0----::::::-:~
EAab..T1 =10 5 x10 3 x10-5 (100)=10 5 N EAab..T2 = 10 5 x 10 3 x 10-5 ( -150) = -1.5 x10 5 N
24.13
6[_~ -~]{~J={~~}+{-~~~}
[_~ -~] {~2}={JJ+{_~.~~~g~}
106
-~~~
0
2:?;6 -10 6 10 6
-~0 6
{2 = {~ }+ ;1~~ x1o 5
2}
F2
10 -1.5x10 5
= 0.125
mm
@
@
Equilibriulll says both rods are at the same stress. Mechanical (total) strain in the left rod is:
{
X
}=
Problem. 2. Therm.al displacem.ent in systetn A beam extends horizontally between two rigid walls as shown. The ends are effectively clan1ped. A rod is used to support the center of the beam; the rod is pinned to the continuous bean1. The rod is first cooled 200 t. The pinned connection is then loaded with a vertical load of 400 N. What is the deflection of the center of the bean1?
EAIL = 200 N/mm EA = 10 5 N a= 1o-s 1/C ~--,,,_
:::;~=:~.
400N
1--
2000 mm-----~1
Solution This is a planar problen1 with 7 relevant d.o.f. The boundary conditions constrain d.o.f. 1, 2, 5, 6 and 7, while symmetry of the system and loading will cause d.o.f. 4 to be zero, leaving only the rotational d.o.f. 3 unconstrained. We need to assen1ble the stiffness matrix, emphasizing d.o.f. 3.
T
3 ilf:
5 :!\~:
;::::::.:......
1 .!~ :
. . .*= ===
3 =~l~=
~-
-400
F3
thermal
e4EI+EA)
L3
u3
EAa~T}
EA a!1T
5
200N
= 10\10- )( -200){
-l}
200N
= 2oo{_f} N
The third equation of equilibrium is:
u3
= -200 I 440
= -0.45 mm
(downward)
200N 400N
j
24.16
Proble111 3. Equivalent nodal loads in triangle A three-noded constant-strain triangle element is pinned at the left node and on a roller at the right. The element is heated 100 C. What is the equivalent nodal load in the vertical direction at the top node due to the heat?
The element is aluminurn, 10 rnrn thick. Its base width 2a is 200 rnrn and its height b is 150 mm.
E = 68, 900 MPa V= 0.3 a=2.21xlo-5 l/C
(O,b)
.::
OJ
Solution Since only one force is needed, we should carefully look at the terms to calculate only the necessary ones. The basic equation is:
{f}enl prestrazn . = V [Bf[G]{C0 }dV
~ ] - ~ dj : : : : : y~b
dx dydx
-
[G]
0 l/b 0
E l-v2 vE l-v2 vE l-v2 E
1-v2
0 0
@
The thermal prestrain is:
1 0 2 0 3 0
Yxy 0 0 All of the terms in the integral are constants, so the integral becomes merely an area integral. There are also a lot of zeroes:
= x By
0 0
= a~T
{1}
1
= hA B 26 ( G21 1 +
G222)
24.18
j6
!6
=215,000 N
>
24.19
I I I I I I I I I
I I I I I
LECTURE25
AX!SYMMETRY
OVERVIEW
A. B. C. D. E. F. GEOMETRY DISPlACEMENTS STRAINS STRESSES FORCES VECTOR RElATIONS
PROBLEM SESSION
1. 4-NODED QUADRILATERAL
25.1
lo~RVIEWI
A. GEOMETRY
0
B. DISPLACEMENTS
For axisymmetric loads:
and:
v (r,O,z) = 0 { u(r,
25.2
C. STRAINS
er(r,z)
{c(r,z)}=
i.:{J(r,:::!
-~~- ~:li()(JI:;:~~
strain
ez(r,z) Yrz(r, z)
D. STRESSES
"H()Qpn stress
E. FORCES
0
Defined as total force around ring. Most codes require constant value around ring (severe limitation).
zt
;. ...........~
F. VECTOR RELATIONS
{!}
25.3
Stress-strain law:
{a}= [G]{c}
where:
1-v
v
1-v
[G]=
(1+v)(1-2v)
v v
1-v
v
0
v
0
0 0 0 1-2v
Strain-displacement law:
{}= [D]{u}
where
dr
1/r 0 [DJ= 0 a
didr
The hoop strain is interesting:
1 EfJ=rur
a a
dz
u6
leads to:
1
rl 0 r2 0 zl 0 z2 0 z3
0
[H]
----
0
-
1
0
1 0 1 0
'i
'3
1 0 1
0 r2
0 zi 0 z2
0 13
0 z3
25.5
B. STIFFNEss
[K]
=J [B]T[G][B] dv
1
=J J[Hf
0 AREA
1
vol ;,,.,. .
~..,
T
[
=2n [Hf T
AREA
We havt e all ingredients. In princ~iple, can integrate. In practice, numerical integration is used. Equivalent nodal loads:
{f} =
enl . . . ~...,..---;:;::- ~a [MJ
J.[Hf
lfJ{{_1(x,z)} dSPACE
@
U s,e isoparametric concept. Map to double unit square. Define shape functions in ~, 1J plane.
25.6
l l (1 + S)(1-7J)
(1 + S)(1 + 1J)
(-1 1) '
:
11
(1 1) '
=
2 '-:
(1, - 1)
u4
jl~:.
. :
w ='LN.(~, 11)w.
l l
u3
~~=:.:~
B. STIFFNESS
[ k]
[K] = 2n
JJ[N]T[D{[G][D][N]I II rdSd7J
in 3-D or 2-D or 1-D in 3-D
1 1
-I -1
{/} = f [N]T{f}dSPACE
enl space
In
g, 1J
enl
<~oord:
1 1
16-NODED TRIANGLEI
Many pe~ople feel the 6noded triangle is currentl:y the best axisymn1etric element. Some commercial codes allow harmonic loading in the azimuthal (8) direction. An alternate, even so, is to use a solid element (hexagonal) and cyclic symmetry, with gen.e ralloading.
25.8
ILAME'PROBLEMI
Consider the ring shown, a plane stress case. Pressures Pi and Po act. The flat surfaces are not loaded. The solution will also hold for the cylinder shown.
...:-...
~ .-.-.......
.......-----..----
:-..---.
.
-----
for the case p 0 = 0. Note that the sum of the two stresses is a constant. Comment: The Lame' solution for plane stress in a ring also holds for a long cylinder with unrestrained ends because plane sections remain plane!
25.9
The Lame'solution for plane stress in a ring also gives the correct stress for a long cylindt:~r with restrained ends. This is a plane ~;train problem!
if
Pi -
0.3
Find the stresses and displace1nents throughc>ut the ring. Assume plane stress (no forces onL faces of ring).
25.10
..
15
16
17
18
19
20
21
==
14
t__.....
X
__......
~--~--~--~~--~--~--~
~~
4
25.11
B. D.ATASET
I
l\.
assign output2='lame.f12',unit=12
~rhe e.xe~cutive
l~SC/NASTRAN
ECHO=B<)TH
@)
DISPLAC~EMENT=ALL
S PCFORC~ES =ALL
SUBCASI~
1
TEST PROBLEM. AXISYMM:ETRIC.
LOAD= 50
SUBTITI~E=LAME'
OTJTI)T.Jl' (PLOT)
.:.
)..} .. J
:]'
../ .:..
:.:.: :
i. ),, : ..... J
n '[' .T
n
,_,
..
n
,_,
"
n
..,;
.t' .LAJ' .
,. ,... _..,
~"I'
D .Ldi~
0 S?T 51
25.12
Tl~e
e1~tries
are:
PARAM POST -1 GRDSET 1 1 1 1 1 1 1 2456 GRID 1 1 1 0 1 5 0 . 1 0 . 1 0 . = 1 * ( 1) 1 = 1 * ( 10 ) 1 == =(5) GRID 1 8 1 0 1 50. 1 0. 125. 11 23456 = 1 * ( 1) 1 = 1 * ( 10 ) =(5) GRID 1 15 1 0 1 50. 1 0. 1 50. = 1 * ( 1 ) = 1 * ( 1 0 ) 1 == =(5) CTRIAX6 11 130 1 1 12 1 3 110 117 1 9 =1*(1) 1=1*(2) 1*(2) 1*(2) 1*(2) 1*(2) 1*(2) =(1)
I= I=
CTRIAX6 1 4 130 1 15 1 8 11 1 9 117 116 =1*(1) 1=1*(2) 1*(2) 1*(2) 1*(2) 1*(2) *(2) =(1) MAT1 1 30 12.07E+5 11 0.3 PLOADX 1 50,100.,100.,15,8,1 END DATA
1
C. RESULTS
TT.r<ER Il\JFr\Pl'v1A'' TIQ'l\T '-.l.;.w.. ,_l\:
\....);.'::) j_
EPSILON 2.1151167E-15
25.13
T2 0.0
T3
LA.-.J
')
~2-c._Q
..L...:
.J
Tl
1
T2
T3 0.0
Rl 0.0
R2 3.92E+7
R3 0.0
The radi<il force resultant equals the entire pressure load on the inner radius of the ring:
Fradial = (pressure )(area )
(/\'\
'(Y_}
1.047xl06 N
2.618xl05 N
~-*::::.....
~-~~:;::.
0.0438 mm
25.14
0.0277 mm
STRESSES IN TRIAX6 ELEMENTS STRESSES IN MATERIAL COORD SYSTEM AZIMUTHAL ELEMENT GRID RADIAL ID ID 1 1 -8.255891+01 1.585935E+02 3 -3.454276E+01 9.180396E+01 17 -3.380558E+01 9.249939E+01 2 3 5 19 5 7 21 -3.314767E+01 -1.173269E+01 -1.101427E+01 -1.082116E+01 5.503397E-01 7.369445E-01 9.240816E+01 6.539439E+01 6.586105E+01 6.579588E+01 5.225233E+01 5. 246037E+0 1
15 1 17 17 3 19 19 5 21
25.15
aro MPa @
-6.5
-1.4 +1.8(T)
;: -12.7
CD
-34.5
=
=
-11.7
" 3
-
+0.5(T)
::.
3 ;) ~ 8"
D
1 6{)" . .3.
~ ~ - 1"'1 j ..
2'1 ~ "
k,.
-11.5
-5.0
# 126
-1.0
lV[SC/NASTRANTRIAX6
t:.
~1.1\.RC~
-125
-100
~
b
~
I~
MSC/NAS~~6 ~
Ntt\RC #126______________]
....
00 00 <1.)
~
-75
00
~
ro ro
-so
-25
0 0
20 40 60 80 100 120
"'C)
-~
Radial distance, r, mm
25.16
150
C\S
+2.4%
:;E
t>
~
~
~
125
<D 100
"
Q)
r:/'1
75 50
25
+2.19(;
::r:
~ 0 0
r~-----~~-C/NAST~~@~RI~6 -~
. 11 MARC #126
-~'
20
40
60
80
100
120
Radial distance, r, mm
IPROBLEM SESSIONI@
ProbleDl 1. Four-noded quadrilateral
Calculate the stresses in a 4-noded quadrilateral element with an imposed displacement field. The element has a uniform radial displacement of 0.001 mm and contracts toward the centerline by -0.0003 mm. Find all six stress components.
z!~
!
i
! f
~---;; --
'
25.17
100 mm
_l_
Solution A 4-nocjed quadrilateral in 2-D will have a "bilinear" set of displacement functions:
w(r,z)
(The con.s tants can be found formally by matching the displacement functions to the nodal displacements.) The strain-displ. relation for axisymmetry:
i
l/r
di
{u(r,
z)}
w(r,z)
25.18
ez
Yrz
CJr v( -6 x 10-6 + 0. 001 I r) CJe _ E v(-6x10- 6 )+(1- v)0.0011r CJz (1 + v)(1- 2v) (1- v)( -6 x 10-6 ) + vO. 001 I r
rn
The direct stresses all have a constant part and a part that diminishes with radius. The shear stress is zero.
25.19
LECTURE26
MODELING CONCEPTS
PHYSICAL MODELING FINITE ELEMENT MODELING SHEET METAL ELEMENT DISTORTION PLANE STRESS PLATE BENDING SOLIDS
26.1
1~!-IYSICAL MODELINGI
When yt>u are first given a proble:m, decide: What type of structure is it? What is it made of? What arte the boundary conditions? What arte the loads? How will the body deform? How will the body fail? Where will the body fail? How can you document the critical areas? What arte special features: symmetries, end releases, layered plates, anisotropic material?
26.2
The most important question is: How can the steepest stress gradients be :modeled? What elements should be used? How many elements should be used? Where can the mesh be coarse? Where must the mesh be refined? What special FEA tricks need to be used? (SPC, MPC, rigid elements, ... ) What resources are available for the model: -manpower -computer -pre/post processing?
You should make a small test run if you are using: -a new element -new loading types -a new solution sequence Try a warm-up problem with 2 or 3 elements. An example, for transient heat conduction, is:
~
bearing
--(~/L
26.3
(SHEETMETALI
SPOT WELDS .,., ..
....
BOLTED~
MATING SURFACES
> :O' IQIOI,)')QIQI'
>v
Some co,mpanies choose to model one surface of sheet metal rather than the centerline, and choose to model the "mating" surface: Physical problem:
FE
modt~l
surface:
26.4
LOAD= 1,000
MSC/NASTRAN V66
26.5
IE:Ij:EMENT DISTORTION!
Consider the following definitions for plate-like elemenLts in the undeformed shape:
aspect ratio, alb skew, 8 taper, d/t~ 2 directions warp, h/a
d
MSC/NASTRAN V66
WAD= 1000
26.6
IPLANE STRESSI
tttttttt t
: . :-:.:-. . . .
. . <> . .
26.7
FIRST PRINCIPAL STRESS :i GPSTRESS OlPTION 10 NODES pgR EDGE MSC/NASTRAN V66
ly
---X z
TENSILE SPECIMEN MODELED WITH CQUAD4 ...
26.8
X
.
..
; ~ ::-: : :. :
TENSILE SPECIMEN
. ..
TENSILE SPECIMEN
FIRST PRINCIPAL
HP1: 8*8*2 MESH 3:1/5:1 RATIO 5*8*2 MESH 1:1/5:1 RATIO
HP2:
. . . .: ..:: : : ::::.<
i
IPLATE BENDINGI
Consider the effect of skew on bending:
"~
26.10
PERCENTAGE ERROR IN DISPLACEMENT ROO QIJA04 DOF/node = 3 using MSC/NASTRAN eo guAD1 i ii- iiiS
'V. ERROR OF DISPLACEMENT
~.m~~.
o Rn TRIA3
....... _
I
~ . --~ -5;-----~~~------~~~~~--------------~
. 0111--=---11110-----------~;;~~:::::=~l .... .
o ..........._
=-=--~
-.........
.........
-10~--------------------~~------------~~~
,.
-15;---------------------------------~~----~
.,.,.,.
0.18 0.20
-20~--~--~--~--~~--~--~--~--~--~--~
0.08
0.08
0.10
0.12
0.14
0.18
0.22
0.24
0.26
% ..... ert.Q.r
. 1~
~~- Q.f::$~-b.t$ lt$r snz.~
26.11
ISOLIDSI
2C
~v
2B~
~-~{=~~
. ..
:
:~:.::::::::::::::::::::::::::::::::::::::.:::::::.:::::::::::::::::::::::::::::::::::::\.::::::::::.: . :::::::::::::::::::::::::
-
ao -- ...................................... ...................
~
~
~
-~~~~~"~~~~~~~,~~
.. ,. ... -.
:: ::::::::::::::::::::::::::::::::::=:::=:::::::::::::::::=::=::::::::::::~::::::::::::::::::::::::::::::::::::::::::::::::~::v.:::: ..:::
................................................................................................................ .. ..... ...........................
.. ............................................................................................................................ ............. ....
- -
..
40 ................... - -
.............................---- ---..::::u:
........................................................................................................................-.................... .
...................................................... ....... ........................................................... ........ .. ....... .. .. .
....
ASPECT RATIO
26.12
MSC/NASTRAN V.66FX
Two 8-Noded HEXA's Warped Interface of 0.58
100 SID 10 70
50 40 30
- -- -
eo -
. . ...
..
........ .
... ... ............... .................... ......... ..... ........ .... ........................... ....... . .............. ......... .................. ............... ...................
-1: . ...: .:. . :--:: :.: : : :: : : : ::. .:.: : : .:.:..: .: . : :.: : : : :- .:: : :: : : :: :.::.:: : : : .:: : : :
:: :::::.:::.::::::\1.:::: . ::::.: ::.:::.: ..:::::::.:..::::::::::::.:::::.:::::::::.:.:::.::::::.:.~:::::...::::::::::::::::::::::::::
-70 -10 ....................................................................................................................................
-~+-~~~~~~~~~~~~
20 10
-eo
ASPECT RATIO
- ...
MSC/NASTRAN V.66FX
Ten 10-Noded TETRA's Warped Interface of 0.58
100 - : r - - --- ----,r--r-r---r---r-~--, 90 .................................................................... .......... ................... ........................... .
.. .. ........ .......... .. .
70 . .................... ...................... ......................... ............ .............................................. . 60 . ........................................ ............. ...... .... ......... ......... ........ ......... ..... .................
.. ..
~
~
~:
:::::::v ::::::.::.:: .: ::::::::.::.:::::... :::::::::: ..::::: ::~: :: :::::::::: ::::::::: :::::::: ::: ::::.:: :.: . ::::::
:~~ .:u : :.::::::::--::::..:--:::::::::::::....:::::::::::::::: . :::.::::: :::::::: :::::::::. ::::::::: :..... --:: :::::::::::: :: ::.:::--::...:: ~:::::..: :::W:::::.:.:.::::: ::::::.:: .:: ::::::::.::::::::: ::::::::: : ::::.: .::: ::::::::::
-50 ... .............................. ......................... ... : ......... .................................... .............. ..
0+ . ~~~~~~~~-4~~
-eo ............................ ....................................................................................... ......... 70 ........ ...................................... .. ............ ..... .. .................................. ..... ............. ........ . -ao ............................................................... ..........................................
-SID - ... ......... ......... .. ..... ..... ........... .. ...... ..... ..... - .. ........ ...... . 100 r o....-rrr--~ ......-.--.- -rr- --.. 'r"TTTT
10 MD tOOO
ASPECT RATIO
26.13
I I I I I I I I I I I I I I I I I I I
LECTURE27
CONVERGENCE
CONVERGENCE THEORY
A. B. C. D. E. F. G. CAUCHY VS. ABSOLUTE DEPENDENCE ON MESH H VS P CONVERGENCE COVERGENCE CRITERIA MONOTONIC CONVERGENCE (MELOSH) PATCH TEST (IRONS) GOAL (SANDSTROM)
27.1
CONVERGENCE
THEORY
A. CP.lUCHY VS ABSOLUTE
D. 0. F.
27.2
B. DEPENDENCE ON MESH
The accuracy of a mesh depends primarily on the number of d.o.f. It is influenced by the average size "h" of the elements, the polynomial degree "p" of the displacement field, and the quality of the mesh. For a sequence of F. E. models, the error in the displacement field is:
E= O(hp+- 1)
In the Tumer triangle, halving the mesh spacing reduces the error by a factor 1 I 4. The error in an elastic element (where stresses are first derivatives of displacement) is:
E=O(hP)
The error in structural elements (where stresses are second derivatives of displacement) is:
E=O(hp-l)
(slow convergence)
The stress in a space frame of Euler-Bemoulli elements has error: E= O(h 2 ) (fast convergence)
27.3
C. H VS. P CONVERGENCE
Element refinement, h
1
2
~
(=
(=0
{=
3
'
+c
e
0
+ 0 +e + .!efoq @efofe9
9 (!o!oo9
~
A(x) = A 0(1+x/L)
I I
'
+ +0, +++o+++'
4 ~:!
5 4:~eeo'
'
@e!ooo9
@
The "end-to end" system stiffness has error, for 2:1 area ratio: Element refinement, h
1 1
2 2
0j(i
3
!-j(i
9()
::
+3.97
c:
+0.121 etc
0
'
3 c~
4 Qo
+0.00368%
0
+0.00011 0 o/o
5 E8[o oo
+O.o102
tro
'
!'
~o!oo' ~
Qioa!ooof
27.4
Consider a mesh refinement (h) case, with G) QUADS'S and MSC/NASTRAN, as in Lab 1:
2 elem. O"x = 274 MPa 20 d.o.f. O"y = -95 MPa
~__,__
ux = 1.133 mm uy =-0.480 mm
.......
16 elem. Ox= 428 MPa ux = 1.267 mm 112 d.o.f. O"y = -216 MPa uy =-0.569 mm 60 elem. Ox= 400 MPa ux = 1.187 mm 400 d.o.f. O"y = -184 MPa uy =-0.530 mm 200 elem. ax= 391 MPa ux = 1.163 mm 1280 d.o.f. O"y = -176 MPa uy =-0.510 mm
A similar mesh refinement with QUAD4's (stresses at centroid) yields maximum values:
8 elem. 24 d.o.f.
ax= 169 MPa ux = 1.097 mm O"y =-29 MPa uy =-0.461 mm ax= 352 MPa ux = 1.131 mm O"y =-146 MPa uy =-0.487 mm ax= 370 MPa ux = 1.148 mm Oy =-159 MPa uy =-0.498 mm
D. C<)NVERGENCE CRITERIA
Zienkiewicz and others state that for convergence to be guaranteed, one should have displacement functions that:
1. allow rigid body motion (with no strain)
3t
ITa.-0
[[]]
[0
E. MONOTONIC CONVERGENCE (MELOSH)
One can ~~uarantee monotonic convergence under certain conditions: If a) ele1nents satisfy the previous conditions on confor.mability, b) a sequence of solutions is made with nested grids, continually refined, and c) a ]potential energy formulation is used; Then dis]placements, strains, stresses and strain en<:~rgy will converge monotonically to the exact answer from below. Comment: as in the global Rayleigh- Ritz method, the finite element model is too stiff.
27.6
The preceding sufficent conditions for convergence (Melosh) are severe. It is now known that less severe conditions in the "patch test" are sufficient. This allows study of nonconforming elements which are convergent. If one creates an arbitrarily-shaped patch of elements with at least one internal node, and if one imposes displacement conditions around the boundary that are consistent with a constant strain field, and if the correct stress field is obtained, tl1.er1 the element will converge to the correct answer. This test is practical, because it is numerical.
Nonconforming elements can pass this test. The physical rationale is that no energy is lost at interelement boundaries. If one passes to infinitesimal size elements in the patch test, tl1e11 the test is both necessary and sufficient. Reference: Zienkiewicz, 0. C., "The Finite Element Method," 3rd Ed., McGraw-Hill, 1977 (pp. 35, 269, 372).
27.7
G. GOAL (SANDSTROM)
I
Problem Stateme11t
Effort Meter
L'
Analysis
50%
100<31
Accuracy Meter
Results
J'
50o/o
100%
I
\
'
I
/
II
III
27.9
EFFEC11VE
ST~SSES:
MODEl.~
AREAl
AREA2
AREA3
AREA4
AREA5
STG. 2
14.13 13.48
14.62
14.90
-
27.10
CONCLUSION
Use of this model is justified for parameter optimization. For a final validation of the design, a fine mesh would be used.
IPROBLEM SESSION I
ProbleDl 1. Hypothetical elen1ent Consider the hypothetical plane stress triangle shown. It is equilateral and has one degree of freedom per node, for a total of 3 d.o.f.
~~~ u3
Is this a useful element for general use? Discuss its good and bad features. Use commonly applied criteria for evaluation.
27.11
Solution This elerrtent is clearly bad. First of all, it will not assernble with neighboring elements to any use:
The nodes would not be able to move--each element would constrain the motion in a different ~direction. With only 3 d.o.f., the element could only represent 3 modes of motion. It appears that the elemt:~nt cannot have any of the required 3 rigid body modes.
The elem~ent could not pass the patch test. If the propc,sed constant strain field required any element to move with rigid body motion, it would in~;tead be locked in place.
t ~X
27.12
in which 6, 10 and 15-noded versions exist. It is named the ''amoeba" element because of its ability to adjust boundaries to its neighbors. It is desired to compare the accuracy of the versions of the element by modeling a certain planar problem. The first mesh uses 100 of the 6 noded elements. The second mesh uses 49 of the 10 noded elements. The third mesh uses 16 of the 15 noded elements. Which model will have the least error in stress? Each mesh uses elements with similar distortion. You may wish to cast the error estimates into nondimensional form for easier comparison.
Solution
The 6-noded element can use a complete quadratic for displacement polynomials: u(x, y) = qo + qt x+ q2y+ q3x2 + q4 xy+ q5y2 The 1 0-noded element can use a complete cubic:
2 2 u(x, y) = qo + qtx+ q2y+ q3x + q4xy+ qsY
We now compare stress error on the basis of @ element size and polynomial degree:
1 ( Ew oc "49
)3
1 oc 343
~~~~~---BEST
The best model uses 49 of the 10-noded elementE;. This was a race between p and h convergence, and neither won!
v1
27.14
LECTURE 28
28.1
RVIEW
Important th problems include thermal stress and heat c duction. The two can be coupled easily in e direction, but the other direction is really ifficult!
rill~:I< ~1 1\l~- PI< ()Jj L.(:t:ivl S
crEIV1l)El~i\'IlJE<_E$ )
I
("'! l-f(""1s-o.;'(l(.,_ N _{~)Nr. [ :J.J~_j ,t _}:
HEi\T
rt.-r
I
We will view the heat conduction problem as a "field" problem. Other field problems are: stress pressure magnetic electromagnetic fluid flow, ..... One can derive field equations from equilibrium concepts energy (variational) methods weighted residual methods Integral formulations are desired because they "smooth" the field variables (derivatives tend to increase error).
VARIATIONAL METHODS
[[(a~~)J +2'P(x)}x
28.3
The variational o~rator 8 : acts only on cdrtain dependent variables usually acts oJ the field variable, but might act on s t me derivatives has the same nroperties as a differential operator in regard to free interchange of integrals, sum ations and derivatives, and in regard t use_of the "chain rule."
J$x
Interchange is ossible because we have proper integral , finite sums and smooth functions.
is equivalent to as ng for an extreme value of F w.r.t. the functi n '. When this is satisfied, the functional iss d to be "stationary" for the particular'* that is found. In heat conductiof . the operator 8 acts only on the temperaturre, and not on the heat flux.
28.4
@
E
U(u)
= LUe(u)
e=l
-LN(x,y,z) J UB u(x,y,z):
Up
=LNu . 1 J
1
A
28.5
the~al
arbitrary:
au -o
f>=O
au :L a
I
auN -
au _ 0
dU:1
(i=1,2,N)
aCLUel= 0
"
The prolJlem has reduced to consideration of one elerr1ent at a til me! Assign concentrat d fluxes F k arbitrarily to one of tl1e element connecting the node k.
28.6
-QLJ\juj }dv+
fQLNj~ dA
s2
fiK(LNjuj-~)~A-Lfkuk
s3
- fQNidS+ fK(Ll\juj-Ta)f\'ldSs2 s3
fi *- 0
Simplify. lnterch
ge
f, L, fx:
uj
~=j~
aue
F {
s3
}~
*0
- fKTaNfdS - fi
s3
We can identify a
aue aue
dllp
:
~atrix form:
r
rk]{u} +
I
duA
[ k ]{u}- {!}
f KNN.Js J
s3
l
/;ie --
28.8
In matrix notation:
{]}=
total t1ux
jLNJ QdV+
v
volun1etric tlux
s2
surface flux
s3convection .
concentrated
flux
au
a~
Finally:
dul
au
auN
[K]{u} + [K ]{u} = {F}enl + {F}enl + {F}enl + {F} Q Q Ta
internal energy convection int1o\v inflo\v int1o\v inflo\\'
outtlov,r
ILINE ELEMENTI
Consider a two-noded, heat-conduction line element:
(o
e)
~ X
u(x) = [1- xl L x/ L
J{::}
= jfka~ aNj Ad J ax ax X
0
28.9
Hence:
k 11 ={kfx( - x/L) x ax 0
~(1- x/L)Adx
=
Ak/L
=tk(-1
0
)(-1/L)Adx
k12 =
k21
=fk(-1 L) (1/L)Adx
0
=-Ak/L
=
Ak /L
28.10
..__ 0.25"
1. 5"
1
4
0.015"
5(10 ,.
2
't
II
*
el
72
~ 1 ---1~~1-~-1"
Thermal conductivities are: k pliers = 6 Oox10- 4 BTU/secFtn
kwire =
5.05x 10-3
"
-kP
-kw
kP
0
-18:~~; 0.~
-0.892
92
19~635
0
u 1 = 89.8 F
u 1 = 286.0 F
28.11
(without pliers) @
The flux: into the iode is found by forward multiplication: F4 == -0.000016 BTU/sec (with pliers) (without pliers) Conclus:i on: One ~an work 6 times longer on a solder<~d joint by\using pliers as a heat sink.
I
F4 == -0.000102 BTU/sec
TIME, t, sec
PROBLE~M
I
SESSION
Problem 1. Solde~ng temperatures Consider a solde g situation where the tip raises the center o a copper wire to 200 F, as shoWil. The is 5 inches long and is 0 oF at both ends. What is the heat flux into the wire?
1.-
5"-~-.1
28.12
CD
2
o
with: where
kw = kA/ L
Solve:
= 6 x 10-4
2.5 in
F2 = (200 <>p)(2kw)
0)
Solution
28.13
=
0
dX dX
28.14
LECTURE 29
PROBLEM SESSION
1. 2. RIGHT-TRIANGLE ELEMENT GENERAL TRIANGLE
29.1
USSION
_LES
In classical heat c nduction , heat flux is often a vector quantity: In finite elelllent~volumetric heat flux Q is a sc ar, positive into a point. It is c led a heat 1 "load." Equivalent nodal fl es Fi are scalar at a node, p sitive inward:
29.2
(+)
{X}
.. f
{rt}
u{:
work/vol
~'------work _ _____.+
tL--..----work
Lwork/space
_+
--------Jt
enl
Equivalent nodal loads were {/} = [N{ {x}dV. These concepts are not as obvious in heat conduction.
I v
G)
{Q}
{Q}
[k]
t ----work----------~~
L..-_
lwork/space
{ !} = [N{{Q}dV.
en1 V
29.3
DIMENSIONS AN
~
UNITS
Fl'
Consiste:n t units oR, oF OK, oc lb N in m sec sec BTU J=Nm BTU/sec W=Nm/sec
XXX
.,Y.N'.WN.
".!.Y.
~.v.-..t.
~..v.w"' .,..,..,..,..;:;.,.._,,~*';J!l,:!;e"e""
u...-.
1...-~~.-,.~S~'$.~:;~~-;s.~~X~~!::,
Symbol
{u}
Stn ctures
displc tcement ve ocity ace e.~ eration stij fness
1
{u}
{it}
[k]
[b]
[m]
{F}
ctrurpmg
II Lass
thermal load
29.4
appli edload
CONSTANT-GRADIENT TRIANGLE
The constant gradient heat-conduction G) triangle is analogous to the Tumer triangle.
A. SPECIAL CASE
y Consider the right ~ . (O,c) triangle shown. Calculate k 33 and find equivalent fluxes for the distrib- Qo~ uted flux. Assume isotropic material.
~~
~ ~~-----------~-(0,0)
(b,O)
J
AREA
[k
x(JX
a~ aNj +k aNiaNjl dA
dx
Ydvy dvyf
and need to find the shape functions, which are plane surfaces passing through 3 points, in each case:
NI(x,y)=l- b- c
X
y
1
~~
N 2 (x,y)= ~ N 3 (x,y)= ~
1
29.5
AREA
kh -c2
khA - 2
c
_ khbc
c2 2
33-
- khb 2c
nodal flux:
{!}enf = ~lfLNl{Q}dS
Q
2
= 1 [LN(O,y)]{Q 0 }dy =
I
Qo
J{l- 0ylc} dy
o
y/c
29.6
r =
fl
The equivalent nodal flux at a node has no vector sense--it is merely positive into the node and negative out of the node.
B. GENERAL CASE
Consider the triangle shown. The isotropic heat conduction matrix is calculated as:
b 2+ c 2 (a+b)
[k]=kh
(ab-c 2) (a+b)
(a2+c 2)
-b
-a
a+b
2c
(ab-c 2)
(a+b) -b
(a+b) -a
See Problem 2 for calculation of a typical term. The right triangle becomes a special case with a = 0.
(-a,O)
29.7
(b,O)
PROBLM SESSION
Problem 1. Rig t-triangle ele111ent
a 3-nod with a l1eat sourc Nodes 1 and 2 re erature does nod
Considt~r
d, constant flux triangle of 100 watts at node 3. ain at 0 C. What temp3 reach after a long time?
~ ,-100 watts
k = 10 J /mm sec C
h=1 mm
0 0
a
~----o---x
('l
1 200mm 2
F }
100
=2~C
29.8
(-a,O)
Solution The question is precisely as to the value of k 21 , i.e., the flux at node 2 is generated by a temperature of unity at node 1, only.
k
21
dx dx
dy dy
-1
c(l +b)
--(y-c+-x) c(l + ~) b
-1
C(y-c-~x)
= 1 = C(O- c- cg)
-1
Hence:
N 2 (x y)' -
c l c(l+bla) 1
!
(y-c-cx) (1+bla) a
29.10
ax ax
1
ay ay
= hkA[(alb
J
kzl =
2c
kh[ab- c
(a+ b)
.j
29.11
I I I I I I I I I I I I I
I I I I I
LECTURE SO
30.1
DUCTION YNAMICS
In struct11ral cs, the displacement field is a func1ion of tim :
{u ={u(x,y,z,t)}
Nodal displacemen
={u(t)}
t
--+------'-'
;1--~ - - -t ~~\1'~~;\dt
A~ v ~
t
forced vibration
"free" vibration
30.3
This is the respon e problem. If damping and e emal forces are set zero, and if displaceme ts are assumed harmonic:
[B] =[C)]
{F}={O}
l"
RESPONSE
;${.
I
I
t;;..
30.4
No one disputes that inertial forces are collinear with acceleration and that damping forces are collinear with acceleration.
.. } Pdv{u
~;;
~
.~-'?
Jlf{it}
.f
Ct~. ~~
J1"{.} U
View the inertia ef ct as an extemal volumetric loading. Use equi ent nodal loads:
{j}e.n.l. = [[N]T{X}dV
=-
I p [N] {U}dV
T
Inserting the
{f}e.n.l.
~ -~[N]T[N]dV{ii}
-[m]{ii}
mass mahix
Damping terms
12 0 1 0 2 0 1 1 0 1 0 2 0 ~0 1 0 1 0 2
Notice that, in total, the terms add to two times the triangle's actual mass--representing the inertia in two directions.
kg
Mg
slug slinch
N lb
Effect
. . ' , j
.
Effect
;
'--
Cause
Cause
st~e~sst~atn
~]
straindispl. {
=
Ux} Uy
Uz
[/)]
~X
!??~~?~~~~~!!~?..~~6
contact problems interference
!19n!&~C!f_1p~teJi~.
problems include:
F-.
0"--------J) .., F
Geome~c :gon..~!~Jariti.~sinclude:
"--~
~ .....
..,
_,. .l~~
.ru contact
30.10