0% found this document useful (0 votes)
59 viewsTriangular Thick Plate Bending Elements
ok
Uploaded by
juan carlos molano toroCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
59 viewsTriangular Thick Plate Bending Elements
ok
Uploaded by
juan carlos molano toroCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 25
M 6/5
TRIANGULAR THICK PLATE BENDING ELEMENTS
P.P, LYNN,
Department of Civil Engineerin
University of Colorado, Boulder, Colorado,
BS. DHILLON,
Ken R. White Co., Denver, Colorado, U.S.A.
ABSTRACT
The present paper establishes a triangular finite element which includes
transverse shear as well as bending deformations and could be used to predict
the behavior of the nuclear reactor circular end plates with circular penetra-
tion holes with sufficient accuracy. Three conforming triangular displacement
finite elements are developed for thick plate bending analyses. Formulation
of these elements are based on Mindlin's higher order plate theory which in-
corporates the effects of rotatory inertia and transverse shear on the flex-
ural motions of elastic plates.
By a comparative study of the three displacement elements, a practical
9 degrees of freedom triangular model is proposed. Since the non-positive-
definiteness of triangular elenent stiffness matrices for thick plate anal-
yses have been encountered by several investigators, the positive-definiteness
of the proposed triangular element stiffness matrix is both analytically and
numerically examined. The use of this model has been demonstrated numerically
for static bending as well as dynamic response problems. The proposed trian-
gular element predicts the bending behavior of thick plates with good accu-
racy and efficiency, and is recommended for practical use.
1. INTRODUCTION
The finite element method has been applied to plate bending analyses with
considerable success. However, these solutions have been mostly limited to
thin plates following Kirchhoff's hypothesis, and thus fail to predict the be-
havior of thick plates involving significant transverse shear strains. The
circular end plates of a cylindrical nuclear reactor is one such case.
Test results(1] indicate that the response of an end plate of a pressur-
ized nuclear reactor is consistent with circular slabs with restrained bound-
ary. A two-dimensional axisymmetrical finite elenent analysis of the problem
over estimates the stresses and displacenents(2]. Furthermore, due to the
presence of circular holes in the end plate, the geometry is not axisyn-
metric, The deviations from axisymmetry are by no means trivial. Although,
the use of three-dimensional finite elements would be ideal, the amount of
computational work is prohibitable. As an alternative method of analysis, ~ 366 -
a thick plate bending element may be employed to advantage.
Several investigators have considered the formulation of thick plate fi-
nite elements which include transverse shear deformation. Based on Marguerres
shallow shell theory, Utku(3] developed a curved triangular element which can
be reduced to a flat elenent which includes the transverse shear effect.
Herrmann{4], on the other hand, formulated his thick plate element by means
of Hybrid variational principle. One of the interesting approaches in the
construction of thick plate elements is due to Martin[5] in which a sandwich
triangular element is formulated.
Unfortunately, the finite elenent stiffness matrices proposed by Utku
and Martin cease to be positive-definite, if there is an obtuse angle in the
triangle. Later Utku and Melosh{6] devised certain numerical schenes to elim-
inate this difficulty, but none of them are theoretically acceptable.
Wempner, Oden, and Kross[7] proposed a curvilinear quadrilateral thick
shell element represented by simple bilinear polynomials in the curvilinear
surface coordinates. The Kirchhoff thin plite hypothesis is introduced to
Assure that the Kirchhoff theory is approached in the limit and that the con-
tinuity of element displacement fields is fulfilled. The method is thus
called a "Discrete-Kirchhoff" approach. A triangular "Discrete-Kirchhoft"
bending element was formulated by Dhatt[8]. This element consists of three
sub-triangular elements and basically it has 15 degrees of freedom.
In 1968, Clough and Felippa(9] presented a fully compatible quadrilateral
Plate bending element. In this element, a simple shear distortion mechanisn
is added to Clough and Tocher's thin plate element(10]. The resulting quadri-
lateral element is formed as an assemblage of four partially constrained
linear curvature compatible triangular elements. The quadrilateral element
has basically 19 degrees of freedom.
Independently, in a quite different manner from the above approaches, a
fully compatible rectangular thick plate element with 12 degrees of freedom
was developed by Greimann and Lynn{11], and is based on Mindlin's higher order
Plate theory[12]. Extension of the method to the formulation of triangular
thick plate elements will be described here. Three different such models are
doveloped and numerically studied. Among these elements, a 9 degree of free-
dom fully conforming triangular element is developed by means of imposing a
constant transverse shear constraint along each side of the triangle. This
constraining technique differs from the "Discreto-Kirchhoff" approach employ-
ed by other investigators. Numerical investigation has proven that the 9
degree of freedom element is well suited for thick plate bending analysis by
virtue of its simplicity of formulation, versatility of application, eff-
ciency of computational effort, and accuracy.
In view of the earlier failures in the positive-definiteness of element
stiffness matrices[5],(6], the proposed 9 degree of freedom element stiffness
matrix is analytically and numerically investigated. It was found that the
Positive-definite character of the stiffness matrix remains invariant with = 367 -
respect to different triangular geometries.
Various static bending and free vibration problems are numerically ana-
lyzed to demonstrate the behavior of the elenent.
2, PLATE STRESS-STRAIN RELATIONS
The plate stress resultants per unit length of a plate (Fig. 1) are de-
fined in the usual manner as
h qa)
Qs y= Jo ©,
xz! "yz)
in which (M,, M,, Myy) are bending monent resultants, (Q,, Q,) are transverse
shear resultants, and the stress components ¢,, ...., Ty, are defined in the
usual sens
According to Mindlin's plate theory(12], the plate displacement compo-
nents u, v, and w in the directions of x, y, and z coordinates are
us 28 (x,y), v= =20 (x,y), w= w (x,y) (2)
where @ and @ are rotations of the plate cross sections, x = constant and
y = constant respectively, and w is the transverse deflection of the plate.
On the basis of Eq. 2, Mindlin derived the following plate stress-strain
relations[12]:
w= DO, - wy)
M, = D0, + 98.)
~ 2p +
wy 7 GaP +o (3)
2,
Q « Ghiw + 8)
2,
= Pon =
ay. Gh Wv °)
where D is the flexural rigidity, G is the shearing rigidity, h is the plate
thickness, v is Poisson's ratio, commas denote differentiation, and the con-
stant «7 denotes w/12.
3, ‘TRIANGULAR FINITE ELEMENTS
One of the most important steps in the finite element displacement for-
mulation is the selection of a suitable displacenent field. Preferably, the
trial displacement field should be constructed such that all the convergence
criteria are met; thus ensurance of convergence is guaranteed. Through the
use of variational principles, convergence criteria for thick plate bending
elements may be formulated(13].
For a triangular element, the simplest possible representation of the
plate displacement field (w, 8, @) is by means of linear polynomials in x
and y as shown in Case I of Table I. Assignment of three nodal quantities, - 368 -
a transverse deflection and two cross sectional rotations, to each vertex of
the triangle leads to a 9 degree of freedom triangular element. Since the
element displacement field possesses the rigid body freedoms, the constant
plate strains, and the continuity of (w, 6, ©) across element boundaries, the
element satisfies all the convergence criterial7],(9], [13]. Accordingly, it
4s said to be a conforming element. However, numerical investigation shows
that the use of this simplest elenent results in a slow rate of convergence
which is of little practical use.
To enhance the slow convergence rate, there are two alternative methods
[11]. Both of these approaches will be investigated in this paper.
From the classical theory of plates{14] or in view of the transverse
shear expressions (Eq. 3), it may be observed that the order of magnitude of
the first degree terms in 6 and » corresponds to that of the second degree
terms in w. Accordingly, the displacement field in Case I, with linear poly-
nomials for w, 6, and %, becomes non-error-consistent approximation. An
error-consistent displacement field may be constructed through the inclusion
of all the second degree terns &, xy, y*) in the expression of w together
With the linear expressions for 8 and 9. This results in Case II, as de-
fined in Table I, which is a 12 degree of freedom element. ‘The requirement
of 12 nodal displacements for each of the triangular elenents involves pro-
vision of some extra nodes beyond the usual three nodes at the vertices of
the triangle. The solution becones feasible with the introduction of three
W mid-point nodal displacements on the sides of the triangle. ‘The presence
of the mid-point nodes is instrumental in including all the second degree
terms of w as well as in preserving the continuity of the transverse deflec-
tion across element boundaries. However, a complication arises therefrom
due to the requirenent of extra storage space in computers. This difficulty
suggests the need to explore some means of modifying the displacement field
80 as to avoid the introduction of mid-point nodes and thus effecting a more
practical displacement model.
We now consider the modification of Case II, that is, the possibility
of eliminating three of the unknown @ coefficients in the w expression. This
is accomplished through a constraining condition based on the assumption of
constant transverse shear along each boundary of the triangular element.
Along the tangential direction to any boundary of the triangular element
(Fig. 2), the transverse shear is given by
a = Pon - %)
t t
or a)
aq - Pon pm + m+ wy = oe]
where t is the tangential coordinate, (4, m) designate the directional
cosines of the boundary outer normal n, and Ve represents the rotation of
elenent edge cross section about the t direction.
The requirenent of constant transverse shear along the element edge in-
Plies that = 369 -
2
Me 7 Gh ee HED 0
or 7 )
0 eg + 8m? = (2W yoo
1X Se @ dtm a O yy ~
vay * Sy oy
Substitution of the polynomial displacenent field of Case II (Table I) into
Eq. 5 yields
1 2,
(aq + F Yq) m+ Gl8y - Vg) - Og) tm + (ag - 2 85) 2? = 0 (6)
Letting (4,, m,), (4g, mp), and (45, mj) be the directional cosines of the
outer normals to the three sides of the triangle (Fig. 2), and writing Eq. 6
for all the element boundaries, we have
2 2
my? ym ty +h °
2
my” tama 4(By ~ ¥3) - 857 = 1 0 M
2
ny? tang a, - } By °
The determinant of the coefficient matrix equals
(eymy = tym) (tymg - Lymg) (Lym, - tym) (3)
‘The above quantity is a product of three non-zero terms as long as the trian-
gular element under consideration has a finite area. Therefore, we find from
Eq. 7 that
Og - bg, %5 = 2(BQ - ¥g), Og =o By «9)
Elimination of (eq, a5, &) from the displacement field of Case II by means
of Eq, 9 results in Case III as given in Table I. It should be noted that
the triangular element of Case III belongs to the 9 degree of freedom models.
Next, we investigate the conformity of the triangular element of Case
III, Substitution of the elenent displacement field into Eq. 3 shows that
the requirement of constant plate strains is fulfilled. The presence of the
rigid body freedoms is obvious. To show the satisfaction of the continuity
criterion, we first note that, due to the linearity of & and 9, the rota-
tional displacement fields are continuous across the element boundaries.
For the continuity of w, we consider its expression along any side of
the element
we cy + Cot + cyt”
in which C,, Cp, and C, are unknown constants. However, these three con—
stants can be uniquely determined by the two end nodal deflections of that
side and the constraint of Eq. 5. Hence, the element of Case ITI is a fully
compatible element.
It may be observed here that Case III has the same 9 degrees of freedom
as Case I, but it provides a higher order polynomial in w and retains the
error-consistent characteristics of the Case II displacement field. There-
fore, it should be competative with Case ITT in quality of results while = 370 -
maintaining the simplicity of Case I.
Once the displacement fields are established, the formulation of element
stiffness and load matrices, and structural stiffness equations can be car-
ried out by the standard finite element techniques(15]. Some element proper~
ties of Case III are given in Appendix I. Details of the other elements and
computational schemes etc., which will be omitted here, may be found in Ref.
13].
4. BEHAVIOR OF CASE III TRIANGULAR ELEMENT STIFFNESS MATRIX
As a direct consequence of the linear elastic mechanical properties of
elastic bodies, an element stiffness matrix should always be positive-defi-
nite or at least positive-semi-definite(5],(6]. Mathematically, these two
properties of the element stiffness matrix (K] are characterized respectively
by the entirety of positive eigenvalues, and positive with one or more zero
eigenvalues of the [K] matrix. On the other hand, a non-positive-definite
(k] matrix possesses one or more negative eigenvalues; and theoretically it
ie not acceptable, This behavior is directly associated with the non-posi-
tiveness of any diagonal elonent of the stiffness matrix. A negative elenent
K,, of the [Kk] matrix implies that for an applied force F, there will be a
corresponding displacement 8 in the opposite direction of Fy but this is
physically absurd.
In the development of a triangular thick plate element, Utku and Melosh
[6] have observed numerically that their elements becomes non-positive-defi-
nite whenever the triangular element has an obtuse angle. This same behavior
was also detected by Martin[5] in his triangular sandwich plate element.
Subsequently the above investigators have devised various numerical schenes
to eliminate non-positive-definiteness of the element stiffness matrices, but
all of them are indefensible on theoretical grounds.
Since the main reason for this theoretical pitfall is due to the intro-
duction of artificial constraining conditions, the behavior of the element
stiffness of Case III vith respect to the triangular geometry will be closely
examined, The ensurance of the positive-definiteness of the [K] matrix will
be discussed in two ways, analytically and numerically.
Analytical Proof. Let us investigate the positiveness of all the diago-
nal elements of the ([K] matrix as given in Appendix I. Note that the stiff-
ness matrix (K] is presented in two parts; one is the bending stiffness (Kk?)
and the other is the transverse shear stiffness (K°], From Appendix I, it
can be shown that the (K] matrix is symmetrical and its diagonal elements
are all positive. However, the positiveness of some diagonal elements in the
(x*] matrix is not evident.
In order to show the positiveness of all the diagonal elements in (K°]
matrix, we consider a typical diagonal element, K°(2,2), whose explicit ex-
pression is -3m
2,
w(2,2) = £59h {r2(ad - be)? + 8(b - dd (ad - be)(a + ©)
oy
+ - a)? Mase)? + + ar? + @? +b? 4 0? + a1}
where 4 is twice the area of the triangle and (a, b, c, d) are coordinates of
the triangular nodes as shown in Fig. 2.
We can show mathematically that, indeed, K°(2,2) > 0 for all values of
a, b, c, and d. This proof is carried out in Appendix II. In view of this
proof, we conclude that the [K] matrix, derived on the basis of the displace-
ment field given in Case III, is a positive-definite matrix.
Numerical Proof. We calculate numerically, a1 the non-zero eigenvalues
of the [K] matrix with respect to different triangular geometries. For this
investigation, the triangular element is oriented as shown in Fig. 3. The
nodal coordinates a, b, and d are fixed as a= 4, b= 0, and d= 4. Varia-
tion of the ¢ coordinate from c = 2 to c = 6, then gives the element different
goemetries (Fig. 3). The parametric variation of eigenvalues of the [kK]
matrix with respect toc is presented in Table II. It is clear from this
table that all the non-zero eigenvalues are positive regardless of the element
geometry.
5. NUMERICAL EXAMPLES
To demonstrate the validity of the finite element models in Table I,
solutions to some typical plate bending problems are presented. The use of
symmetry, whenever possible, is employed and the finite element solutions are
compared with known analytical results. The numerical analyses are performed
with the following objectives in view:
4) To investigate the convergence characteristics of the finite elenent
solutions ;
44) To make a comparative study of the three displacenent models;
4441) To show the capability of the present finite elements in incorporat-
ing the effect of transverse shear deformation,
A simply supported square plate under uniform load qq is solved for
thickness to span ratio, h/L © 0.1 and Poisson's ratio, v= 0.3. Types of
mosh sizes used are shown in Fig. 4, The central deflection, bending moment,
and transverse shear are chosen as the measures of accuracy of the approxima-
tion, Numerical results on these quantities, obtained from the three differ-
ent models, are given in graphical form in Figs. 5 through 9.
As an application to dynamic response problems, the finite element model
of Case III is used to analyze the free vibration of the square plate. The
convergence of eigenvalues corresponding to the first two symmetrical modes
Ay, and yg are shown in Figs. 10 and 11.
Finally, a circular plate with a central hole is chosen to illustrate
the versatility of the triangular element to represent curved boundaries and
demonstrate the use of general shape triangular elements (other than right
triangles) in solving thick plate problems. The finite element of Case IIT = 972 -
is used for this purpose. Figures 12 and 13 represent respectively the
finite element idealization and the deflection along the inner edge of the
circular plate.
6. DISCUSSION OF NUMERICAL RESULTS
From Figs. 5 and 6, it is observed that the rates of convergence for
Cases II and III are almost identical, but the convergence rate is much
slower for Case I. This is not surprising, since a non-error-consistent dis-
placement field is employed in Case I. On the other hand, both Cases II and
III yield approximate solutions rapidly converging to analytical results which
are obtained from Reissner's plate theory(16]. The small discrepancy in de=
flection may be attributed to different assumptions in Reissner's and Mindlins
Plate theories. The square plate solutions of Ref. 7 are algo superimposed
on Fig. 5. These solutions (shown in dotted lines) were obtained by the use
of conforming thick plate rectangular elenents proposed by Wempner, Oden, and
Kross[7]. The slow rate of convergence is again due to the error-inconsistent
approximations.
Case II has the drawback that it requires additional midpoint nodes and
thereby increases the computer storage requirement as well as the computa-
tional time. The comparison of structural stiffness matrices of Cases II and
III gives a fairly good indication of the increase in storage requirement.
For example, for the mesh size 8, Case II requires 3.7 times more storage
area than Case III, Similarly the increase in computing time is evident. It
is interesting to note that the time ratio of Case II to Case III increases
sharply from 1.33 for mesh 2, to 3.05 for mesh 8.
The deflections and moments along the centerline of the plate for Cases
II and IIT are shown overlapping in Figs. 8 and 9 as the differences between
them are not depictable. These results indicate that the constraint employed
to arrive at the displacement field of Case III from Case II, does not affect
the solution accuracy. Identical conclusion may be drawn from Fig. 9, where
the transverse shear along the plate edge is shown.
The preservation of positive-definiteness of Case III element stiffness
matrix is tested by the circular plate problem (Figs. 12 and 13), where ob-
tuse angles in the triangular elements are present.
The above observations of the numerical results lead to the conclusion
that the 9 degree of freedom element of Case III represents the best proper-
ties of the other two cases, that is, efficiency and quality of results. It
shows superiority over Case I in the results of approximations and outweights
Case II with considerable savings in computing time. Consequently, Case III
becomes the most efficient and practical finite element among those studied
here.
The simply supported square plate eigenvalues (natural frequencies) 4)
and 4,3, obtained by the use of the Case III finite element (Figs. 10 and 11),
converge rapidly to their exact answers given by Mindlin. Comparison of ),,
and 4,3 with those from classical plate theory reveals the influence of
rotatory inertia and transverse shear on the eigenvalues. = 373 -
7, CONCLUSIONS
‘The formulations of three fully compatible triangular thick plate ele-
ments are described. These models are compared with respect to the conver-
gence characteristics, the quality of results, and the computation efforts
involved. As a result, a practical 9 degree of freedom model (Case III) is
established for the analysis of thick plates in flexure. Both by analytical
and numerical means, the positive-definiteness of the element stiffness matrix
of the proposed model has been carefully examined.
The use of this model has been demonstrated for static bending as well as
free vibration problems; and its capability to incorporate the effects of
rotatory inertia and transverse shear deformations has been illustrated. Due
to its simplicity of formulation, versatility of application, computational
efficiency, and sufficient accuracy, the proposed 9 degree of freedom tri-
angular element is recommended for practical use. In particular, the method
is well suited for the analysis of the nuclear reactor thick circular end
plates with circular penetration holes. = 374 -
REFERENCES
fa]
(2)
(31
4]
(s]
6]
C7]
(8)
(9]
fio)
fil
fiz]
(13)
Cia]
fis)
fie]
Paul, S. L., et. al, "Strength and Behavior of Prestressed Concrete Ves-
sels for Nuclear Reactors", Vol. 1, Structural Research Series No. 346,
University of Illinois, Urbana, Illinois, July 7
Corum, J. M., and Krishnamurthy, N., "A Three-Dimensional Finite Element
Analysis of a Prestressed Concrete Reactor Vessel Model", Proceedings
of the Symposium on Application of Finite Element Methods In Civil Engi-
Searing Meshvilie, temnesses, ng, Nashville, Tennessee, 1963, pp. 3-94.
Utku, S., "Stiffness Matrices for Thin Triangular Elements of Non-zero
Gaussian Curvature", Journal of the American Institute of Aeronautics
and Astronautics, Vol. 5, Sept. PP. 1667.
Herrmann, L, R., "A Bending Analysis of Flat Plates", Proceedings of the
Conference on iatrix Methods in Structural Mechanics, Wright-Patterson —
, Ohio, 66-80, Nov. PP. 577-602.
Martin, H. C., "Stiffness Matrix for a Triangular Sandwich Element in
Bending", Technical Report 32-1158, Jet Propulsion Laboratory,
California Institute of Technology, Pasadenss Californias Feb, 1968,
Utku, S., and Melosh, R. J., "Behavior of Triangular Shell Element Stiff-
ness Matrices Associated with Polyhedral Deflection Distributions",
Technical Report 32-1217, Jet Propulsion Laboratory, California Insti-
ute of Techsolegyy Pasadena, California, Fob 1968,
Wempner, B. A., Oden, J, T., and Kross, D. A., "Finite Elemont Analysis
of Thin Shells", Journal of the Engineering Mechanics Division, Proceed-
ings of the Anerizan Soolety of Civil Englaccre, Bees 1865, pp. y of Civil Engineers, Dec. 7; PP. Y27S——
Dhatt, G. S., "Numerical Analysis of Thin Shells by Curved Triangular
Elements Based on Discrete-Kirchhoff Hypothesis", Proceedings of the
Tih Civil Engineer:
Symposium on Application of Finite Element Methods in gineering,
Vanderbilt University, Nashville, Tennessee, Nov 1960, pp-2Se2 78
Clough, R., and Felippa, C., "A Refined Quadrilatural Element for Analy-
sis of Plate Bending", Proceedings of the Second Conference on Matrix
Methods in Structura1 Mechanics” Wright-Patterson AFB, Ohio, Oct, 1888,
pp. 399-739.
Clough, R., and Tocher, J., "Finite Element Stiffness Matrices for the
Analysis of Plate Bending", Proceedings of the Conference on Matrix
Mothods in Structural Mechantes~ Wright-Patterson KFS; OhTosAFFDL-
‘'TR-66-80, Nov. 1865, pp. 515-545.
Greimann, L. F., and Lynn, P. P., "Finite Element Analysis of Plate
Bending with Transverse Shear Deformation", Nuclear Engineering and
Design, Vol. 14, No. 2, Dec. 1970, pp. 223-2800
Mindlin, R. D., "Influence of Rotatory Inertia and Shear on Flexural
Motions of Isotropic, Elastic Plates", Journal of Applied Mechanics,
Vol. 18, March 1951,’ pp. 31-38,
Dhillon, B. S., "Triangular Finite Elements for the Bending Analysis of
Thick Elastic Plates", Ph.D. thesis sutmitted to the University of
Colorado, May 1970,
Timoshenko, S., and Woinowsky-Krieger, 8., Theory of Plates and Shells,
2nd Ed. McGraw’Hi11 Book Co., N. Y., 1967.
Zienkiewicz, 0. C., The Finite Element Method in Structural and Continuum
Mechanics, McGraw iii Book Co-, N. ¥.> ;
Salerno, V. L., and Goldberg, M. A., “Effects of Shear Deformations on
the Bending of Rectangular Plates", ' Journal of Applied Mechanics, Vol.
27, March 1960, pp. 54-58. - 378 -
APPENDIX I, ELEMENT PROPERTIES FOR CASE III MODEL
On referring to Case IIIof Table I, we define the element nodal displace-
ment vector [6] as
r
(0) = fw, 0 0; wy, 9 & Wy % 95) ay
and the generalized coordinates vector {a) as
(aJ™ = Cay, a, a5, By, Bys Bye Yas Yor «1-2)
They are related by a matrix [A]
{6} © (ay7? {a} (1-3)
where the [A] matrix is defined by the coordinates of the element nodes (a,
b, c, d as shown in Fig. 2)
1 0 0 oO oO o oO 0
oo © 1 0 0 © o 0
o 0 oO 0 Oo oO 1 o
roa dv 0 & Bo 8
fale o 0 0 1t a b 0 ° ° , (1-4)
co) 0 o o oO 0 1 a b
Me tep apenont 200s mone comes
o 0 o 1 c a 0 ° 0
© 0 0 0 0 © 1 e« 4
Equation I-3 can be written in matrix form
{o) = (p}(B]{e} = (p}(B](aI"*(5} (1-5)
in which the following matrices are introduced:
fol" = Om, HH, QO]
x Ny xy Ox = 376 -
o 0 0 0 0 0 © 1 ©
o 0 © 0 0 1 © o 0
(8) = WO -0 0. 0 dO 00 8
oO ° z
°
From the definition of the element stiffness matrix, [K], the [K] matrix is
obtained in two parts; the bending stiffness matrix (K°] and the transverse
shear stiffness matrix (k°]. Hence, we have
Ck] = Canty" J ftp, "CD, 5 30, Janay ](472)
(1-6)
+ (a71)"Cf ftp, I? tpy9I(By Jandy ]CA7*] = Cx?) + CKE]
The explicit expressions of Kj, and Ki, are given in Table IIT.
APPENDIX II, PROOF OF K® (2,2) > 0
It is required to show that K° (2,2) > 0, for any a, b, c, and d.
s
since 292 > 0, it is only necessary to consider the sign of (see Table
11).
12(ad - be)? + B(b - d)(ad = be) (a +c)
+ (b> - 4)? [aa +e)? + + a)? 1-1)
Peal.
+ 024 Pac’
Expanding Eq. (II-1) and rearranging the terms in powers of c, we find
g(e) = (6b? + 4ba + 2a”)c? - Gacb + a)%e
2,
ar-2)
+ (20>? + 6274? + daa + 2b4 + 2a4 - 20%a - ba)
It is noted that g(c) is a quadratic equation inc, Thus, if we can establish
that g(c) is positive for a certain c and the discriminant of g(c) is negative
for any a, b, and d, then K* (2,2) is always positive for all a, b, c, and d.
We first consider that g(c) is greater than zero for c = 0. Setting
¢ = 0 in Eq. II-2, we find
(0) = 2a7(o? + 2ad + 3d”) + b( - art + Sco? — a)? as)
Now we have to show g,(b) = b” + 2db + 3d” > 0, for all b and d. Considering
g,(b) as a quadratic equation in b, we have
(0) «347 > 0, qa)
and its discriminant
a? ~ 3a” = - 24? < 0, for any d . ai-s)
Hence, g,(b) is greater than zero for all b and d, and Eq. II-3 is definitely
positive. Therefore we find -377-
g(0) > 0, for alla, b, and d. (1-6)
Next we show the sign of the discriminant of g(c) to be negative. For
this purpose, we consider the discriminant
2 2,
D(a,b,a) = [= 3ab+a)?]? ~ (6b? + add + 247I(2007 + 6a2a? + 407d
+ 24+ 244 - 2n%4 - 2ba°)
(11-7)
= = a2 ap! - 4v9a + 2074? - ava? + 344)
= 4(a® + va® « vat - av9a? - vfa? - v8 + 30%) -
If bd > 0, we rewrite Eq. 11-7 as
D(a,b,a) = = a2[3(b - a)4 + B(ba( - a7]
= at(a? - 03)? + vaca - vy? + 208 - wife? + vat] om
Consider the underlined part of Eq. II-8 as a quadratic equation in a
(2?) = 6?(a?)? - v8ca?y + 20% ar-s)
Based on the fact that
(0) = 2° >0, (11-10)
and the discriminant of g»(d”)
pty? = 4(b?) (20%) © - re <0, ar)
we obtain
gp(d?) = 208 - via? + va > 0, for ari b and 4. (1-12)
It is clear from Eq, IT-8 that
D(a,b,d) <0, for all a, b, and d provided bd > 0. (11-13)
If bd < 0 (or bd = =|ba|), we rewrite Eq. II-7 as
D(a,b,a) = = a2(ab4 + 4b? [bal + 202d? + 4a” [bal + 304)
r-14)
= [4d + add> + av2at + 4 [bal (ad2a? + b8a + vty + 120°]
Let the two underlined parts of Eq. II-14 be
(0) = 4a? + 4a%p + 40% a1-15)
and g(a) = 40a + vas vt. 1-16)
Taking g3(b) and g4(d) as quadratic equations of b and d respectively, we can
show
4 > 0, (1-17)
gg (0) = 44° > 0, and g4(0) = b
and the signs of the discriminants of g,(b) and g4(4) to be respectively
2
(a5)? - aca) (a8) = - 3a! < 0 (11-18)
and 03)? ~ 4(4v?) (v4) = = isv® < 0. 378 -
In view of Eqs, II-17 and II-18, we find
&(b) > 0 and g4(4) > 0 for all b and a. (aI-19)
Hence, we have
D(a,b,d) <0, for all a, b, and d except for bd = 0. (11-20)
If bd © 0, we find directly fron Eq. II-7 that
D(a,b,a) = - a2(ap4 + sat) - aca® + 3n%) <0, ar-21)
for all a, b, and d.
Equations II-13, 11-20, and II-21, enable us to write
D(a,b,d) <0, for all a, b, and d. (iI-22)
Finally, on the strength of Eqs. II-6 and 11-22, we conclude that K° (2,2) >0
for all the values of a, b, c, and d. - 379 -
Table I, Three Conforming Triangular Displacement Fields
case ASSUMED DISPLACEMENT FIELD ELEMENT
WG) 00K *05y
TL |@ =A, +Aex +Asy
O97, +7, KY e
|
WG, 00, X 4 ¥ 94x" +g xy +g y*
wae nayy Bt dy Ay
0 =A, +P, x +BY
ON hah 8,
Aa
1 | o:AeAxeay
OTe Rxely 8 %,
An %
TABLE IT
Parametric Variation of Non-zero Eigenvalues
for Case III Element Stiffness Matrix
Eigenvalues
4 c-4 c-6
M4 4.648 x 10° 3.212 x 10° 1.714 x 10°
rs 4.737 x 10° 4.246 x 10° 3.129 x 10°
Xe 1.179 x 108 1.542 x 10° 2.425 x 10°
dy 2,225 x 10° 2.929 x 10° 5.183 x 10°
dy 1.316 x 107 1,202 x 107 1.181 x 107
ry 1.464 x 107 1,810 x 107 2.665 x 107
Alo Ag 7 Ag @ 0 (rigid body modes) = 380 -
TABLE IIT
‘Triangular Elenent Stiffness Matrix {k] *
dy > ae Kyo » Sgn
a ° (o-d)? + (ena)?
a,2) ° (oa)? 2 = (enayl$ + (oa) gh
a3) o (o-aytd « 2 (ea) - (e-ay? gh
aya) ° (ved) d ~ (e-a) ©
a,8) ° co-ay(“° ~ Ba) + cenar(be - 538
a,6) ° (oa) (38 - *2°) 4 (e-ay(ZBE - 22)
an ° = (o-d) b+ (ea) a
an ° wnt - 2) + ew - $8)
a9 ° (o-ay(F - 88) + ce-ay(ge - BY
(2,2) u(b-d)? + (ca)? & + oeayhit + Fama)?
2,3) = (bed) (e~a) (ev) Cle-a)r, + (o-d)rg]¥y + (va) (ona) 3b
(2,4) o ete foray Ora * Fy
(2,5) ud(d=d) = © (cua) $(- 8+ 9). @ar(- fas. ue, 334)
re road rac rc
(2,6) ue(ona) - va Gora) $ GE- EF) ows BE BE
rob rye
@,7 ° -#- woes
rb, rgbdryad rgb
@,8) Hood) +a ea) 3 GA-2) + woe DE -
ra rgbe rac ga
(2,9) -un(oea) + vo (era) (- $8 + EE) + ran(- Be - BE, 3) TABLE III (Cont.)
- 381 -
G5 RS Oe Kiyo » gn
(3,3) (ba)? + u(e-ay? a Fa¢e-a)?
@.4) ° He cena. BE
(3,5) ve(bed) = ud(c~a) &(-B ae + (e-ar(- eae = ele + Ese
(3,6) d(b-d) - ue (e-a) 3 GF - “s + (ea) rae Tis - TG
@,7) ° -$- (o-ay(2 + =e
(a8) sated) s urte-ay $B - 2H), ceom(Be ne. oe
(3,9) ~b(b-d) + ua (e-a) £ (B+ ZB) + e-wr(- 2 eas
4,4) ° oF + a?
4,5) ° ~ } (ea? + bo”) + 2B (rye? + ryae)
4,6 ° 3 (aa? + ac?) - 2 (rged + ry07)
(4,7) o - ac - bd
(4,8) ° 3 (od? + ade) - § (rgbd + rybe)
(4,9) o - 2 (ea + ac?) + § (rgad + ryac)
2
(5,5) ud? + 3 (va? + 020?) - (rgba? + rybed) + te
(5,6) (uevded = } (aba? + ave) + 3B (ode + 04”)
+ TE eed + be) - et
6D ° 3 (a + abe) - § (rgbd + ryad)
(5,8) shbd - ac = 3 (v?a? + abea) + 2 (o?a + ba?)
r rgbd
1 (2 2, _ 73
+ qh We + aa%) - TABLE III (Cont.)
= 382 -
ye
b
Ga Rye OD
6,9) stad -vbe
6)
6D °
(6,8) vad ube
(6,9) -bd -bac
aD °
(7,8) °
(7,9) °
(2,8) uv? + a?
(8,9) (u + v) ab
(9,9) vb? + ua?
eon
Kyo ae
r,
ed + abe?) - 52
*1
- yy (abe + acd) + Je
2
re
F (07a? + 0707) - § (rgacd + rya0”) + 3
- ¥ (ava + ae) + 2 (eyed + rac)
r,
$ (abd? + aca) - 42 (add + bea)
ry rgbe
- yy (abe + ade) + $e
x,
~ F (avea + 07c?) + 73 (7a + be?)
r rgac
+ ay ae + ac®) - 3
v?
= 7a + 074) + J (yb? + ryan)
1
Ew
e+ 0%) - % (rpab + rya”)
5 rb?
3 (07d? + 07a?) - 3 (rgba + ryabay + Sy
r
=} (ca + aca) + 42 (ada + be)
mo
+E (ea + abe) ~ 228?
2
73a
F (070? + 0707) - 2 (rpabe + rya%e) + —S-
* In Table III, the following notations are used:
zn?
D- "i
2a)
4 = (ad-be),
= (ate),
2 = (btd), and
1g 7 (ry 2,707 epPee%ea) = 383 -
ZW
Fig. 1 Positive plate stress resultants
Fig. 2 A triangular finite element Fig. 3 Parametric variation of elenent geonetry
ue
MESH 6 0-005
DEFLECTION ovat
i
0-001
°
3
- 385 -
REISSNER'S THEORY.
‘CLASSICAL THEORY
RECT. ELEMENTS (REF 7)
Pig. 5
MOMENT - M/qg(2
2
2
3
0.048}
0.046
0.044]
o.042
0.040!
2 4 6 8 10
MESH
Convergence of plate central deflection (h/t = 9.1,
Ualfare load; and sieply supported square plate)
REISSNER'S THEORY
2 4 6 @ 10
MESH
Pig. 6 —Convorgonce of the naxizun moment - 386 -
¢
v)
‘9 0.005
3
= 0.004 pS
‘
z
2 0.003 c
oO
i 0.002
&
3
0.001
° 0.25 0.5
x/L
Fig. 7 Transverse deflection along the plate center line
MOMENT- M/q,2
0.05
0.04 soe
0.03 *
0.02
0.01
°
°
ip
a
°
a
x/L
Fig. 8 Bending moments along the plate center line = 387 -
SHEAR- Qy/agL
° 0.25 0.50
x/L
Fig. 9 Transverse shears along the plate center line
1.10
CLASSICAL THEORY
MINOLIN'’S THEORY
- 2 [o
EIGENVALUE - A,,=,/2TT' enc
3
®
0.94
ae 2 4 6 8
MESH
Fig. 10 Convergence of 1,1 eigenvalue (simply supported square
plate with h/L = 0.1) - 388 -
ke
«
=
a
S 58P
2
“a
