Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Non-Prismatic Plates
By
.... (1)
Thus:
1 1
M x + M y = (1 + ) D +
rx ry
or
... (2)
1 1 Mx + My
+ =
= f ( x + y)
rx ry
(1 + ) D
Structural, Electrical and Foundation Engineer, FAC Systems Inc. 6738 19th Ave. NW, Seattle, WA
1 1
1
= =
rx ry rm
.. (4)
and
w
m 2
2
1
=
rn
w 2
1 +
m
3
2
f ( m)
2
0.5 f (m)dm + C
1 ( 0.5 f (m)dm + C )
1
dm + C 2 . (5)
And
wx = w y = m
0.5 f (m)dm + C
1 ( 0.5 f (m)dm + C )
1
(6)
And
1
1
M x = D + = 0.5 D(1 + ) f (m)
ry
rx
1
1
M y = D + = 0.5 D(1 + ) f (m)
rx
ry
2w
2w
M xy = D(1 )
= D(1 ) 2 =
xy
m
.... (7)
0.5 D(1 ) f (m)
1
( 0.5 f (m)dm + C )
2
2 M xy
2M x M y
+
2
= q ( x + y ) = q ( m)
x 2
y 2
xy
.. (8)
or
M x + M y 2 M xy = q (m)dmdm = M T + (m)
D (1 + ) f (m) +
1
( 0.5 f (m)dm + C )
2
3
2
= M T + ( m)
... (9)
or
0 .5 f ( m ) +
1
1 +
0 .5 f ( m )
1
( 0.5 f (m)dm + C )
2
3
2
M T + ( m)
2(1 + ) D
Let:
u = 0.5 f (m)dm + C1 .... (10)
2
2(1 + ) D
1 + 1 u
Eq. 11 is a Quartic equation in u and has four roots, two are for coordinates
(x, y) and (y, x) and the others are possibly imaginary.
And the solution is found numerically by substituting back in Eq. 5, 6 and 7.
An alternative preferred solution for Eq. 11 can be found by expressing in
w& = wx = w y . Rewriting Eq. 6 to
w& = m
u
1 u2
(12)
or
u=
w&
1 + w& 2
1
sin +
1 +
M ( m)
dm .. (14)
tan = T +
2(1 + ) D
Equation 14 has two basic roots in and the rest are differs by 2 multiples.
Thus w& of Eq. 13 has only two real roots for coordinates (x, y) and (y, x).
General Solution for axi-symmetry Case II:
Changing Eq. 2 to a function on x y , thus
1 1
M x + M y = (1 + ) D +
rx ry
or
... (15)
1 1 Mx + My
+ =
= g ( x y)
rx ry
(1 + ) D
1 1 1
= =
rx ry rn
.. (17)
and
1
=
rn
2w
n 2
w 2
1 +
n
3
2
g ( n)
2
0.5 g (n)dn + C
1 ( 0.5 f (n)dn + C )
1
dn + C2 . (18)
And
wx = w y = m
0.5 g (n)dn + C
1 ( 0.5 g (n)dn + C )
1
(19)
And
1
1
M x = D + = 0.5 D(1 + ) g (n)
ry
rx
1
1
M y = D + = 0.5 D(1 + ) g (n)
rx
ry
2w
2w
M xy = D(1 )
= D(1 ) 2 =
xy
n
.. (20)
0.5 D(1 ) g (n)
1
( 0.5g (n)dn + C )
2
2 M xy
2M x M y
+
2
= q ( x y ) = q ( n)
x 2
y 2
xy
or
2
2 M xy
2M x M y
+
+2
= q ( x y ) = q (n) .. (21)
n 2
n 2
n 2
or
M x + M y + 2 M xy = q (n)dndn = M T (n)
D(1 + ) g (n) +
1
( 0.5g (n)dn + C )
2
3
2
= M T (n)
... (22)
or
0 .5 g ( n ) +
1
1 +
0 .5 g ( n )
1
( 0.5g (n)dn + C )
2
3
2
M T ( n)
2(1 + ) D
Let:
v = 0.5 g (n)dn + C1 ... (23)
2(1 + ) D
1 + 1 v2
Eq. 24 is a Quartic equation in v and has four roots, two are for coordinates
(x, -y) and (y, -x) and the others are possibly imaginary.
And the solution is found numerically by substituting back in Eq. 18, 19 and
20.
An alternative preferred solution for Eq. 24 can be found by expressing in
w& = wx = w y . Rewriting Eq. 19 to
w& = m
1 v2
(25)
or
v=
w&
1 + w& 2
M T ( n)
&
dn . (26)
w =
2(1 + ) D
M T ( n)
dn .. (27)
tan =
2(1 + ) D
Equation 27 has two basic roots in and the rest are differs by 2 multiples.
Thus w& of Eq. 26 has only two real roots for coordinates (x,- y) and (y,- x).
General Solution for Case III:
Changing Eq. 2 to a function on ax + by , thus
1 1
M x + M y = (1 + ) D +
rx ry
or
... (28)
1 1 Mx + My
+ =
= h(ax + by )
rx ry
(1 + ) D
w
w
(ax + by )
w
=
=a
(ax + by )
x (ax + by )
x
w
w
(ax + by )
w
=
=b
y (ax + by )
y
(ax + by )
and
2w
2w
2
a
=
x 2
(ax + by ) 2
.. (29)
2w
2w
2
b
=
y 2
(ax + by ) 2
2w
2w
= ab
xy
(ax + by ) 2
2w
a
t 2
2
1 1
+ =
rx ry
2w
b
t 2
2
2 w
1 + a
t
3
2
2 w
1 + b
t
3
2
And
2
2
2 w
2 w
b
a
2
2
1
1
t
t
M x = D + = D
+
3
3
r
r
2 2
2 2
x
y
2 w
2 w
1
1
b
a
+
+
t
t
2
2
2 w
2 w
a
b
1
1
t 2
t 2
.. (31)
M y = D + = D
+
3
3
r
r
2 2
2 2
y
x
2 w
2 w
1
1
a
b
+
+
t
t
2w
2w
M xy = D(1 )
= D(1 )ab 2
xy
t
2M y
2 M xy
2M x
2
a
+b
2ab
= q (ax + by ) = q (t ) .. (32)
t 2
t 2
t 2
or
2
a 2 M x + b 2 M y 2abM xy = q (t )dtdt = M T (t )
2
2
2
2
w
w
w
w
b2 2
a2 2
a 2 2
b 2 2
t
t
t
t
+ Db 2
Da 2
+
+
3
3
3
3
2 2
2 2
2 2
2 2
2 w
2 w
2 w
2 w
1 + a
1 + b
1 + b
1 + a t
t
t
t
2 D(1 )ab
2w
= M T (t )
t 2
... (33)
Or
2
2w
4
2 2 w
(a + a b ) 2 (b + a b ) 2
2
t + 2(1 )ab w = 1 M (t )
t +
T
3
3
t 2
D
2 2
2 2
2 w
2 w
1 + b
1 + a
t
t
(34)
or
&&
( a 4 + a 2 b 2 ) w
[1 + a w& ]
2
3
2 2
&&
(b 4 + a 2b 2 ) w
[1 + b w& ]
2
3
2 2
&& =
+ 2(1 )abw
1
M T (t )
D
Let
w& = tan
&& =
so w
&
. (35)
cos 2
[cos
+ a 2 sin 2 ]
3
2
(b 4 + a 2b 2 ) cos &
[cos
+ b 2 sin 2 ]
3
2
+ 2(1 )ab
&
cos
2
1
M T (t )
D
.. (36)
(a 4 + a 2b 2 ) sin
1 + (a 1) sin
2
(b 4 + a 2b 2 ) sin
1 + (b 1) sin
2
1
M T (t )dt . (37)
D
Equation 37 has two basic roots in and the rest are differs by 2 multiples.
Thus w& of Eq. 35 has only two real roots for coordinates (x/a, y/b) and
(y/a, x/b).
General Solution Case IV:
The following is the general solution for vertical loads on the plate without
buckling in the Cartesian coordinates using Taylors and Fouriers
representation.
We start by rewriting the partial differential (Eq. 100 pp81 Timoshenko and
Woinowsky-Krieger book using Eq. 37, 35 and Eq. 102 pp81) as follows:
2
x 2
1
1 2
+
+
ry y 2
rx
1
4w
1
1
2
(
1
)
= q ( x, y ) . (38)
+
2
2
rx
x y
D
ry
2 wx
x 2 x 1 + wx2
+
+ wy
y 1 + w y2
2 w y
y 2 y 1 + w y2
3 wy
+ wx
x 1 + wx2
3 wx
1
+ (1 ) 2 + (1 ) 2 = q ( x, y )
y x
x y D
10
(39)
w = p jk x j y k .. (40)
j =0 k =0
Thus
r
wx = jp jk x j 1 y k . (41)
j =1 k = 0
r
w y = kp jk x j y k 1 .. (42)
j =0 k =1
Let:
g ( x, y )
wx =
1 [g ( x , y ) ]
h ( x, y )
wy =
1 [h( x, y )]
(43)
(44)
Then
r
wx
g ( x, y ) =
1+ w
2
x
h ( x, y ) =
1+ w
2
y
jk
x j 1 y k
j =1 k =0
1 + jp jk x j 1 y k
j =1 k =0
wy
jp
kp
jk
(45)
x j y k 1
j = 0 k =1
r t
1 + kp jk x j y k 1
j =0 k =1
. (46)
Now express wx, wy, g(x,y), h(x,y), q(x,y) and in Fouriers representation over
the interval 0 x a and 0 y b for a rectangular portion of the plate in
Cartesian coordinates as follows:
wx = wxmn cos
m = 0 n =1
m x
n y
. (47)
sin
a
b
11
g ( x, y ) = g mn cos
m =0 n =1
m x
n y
(48)
sin
a
b
m x
n y
. (49)
cos
a
b
m =1 n =0
m x
n y
(50)
h( x, y ) = hmn sin
cos
a
b
m =1 n = 0
w y = w ymn sin
q ( x, y ) = q mn sin
m =1 n =1
m x
n y
.... (51)
sin
a
b
Where:
4 a b
m x
n y
sin
dx dy . (52)
wx ( x, y ) cos
ab 0 0
a
b
4 a b
m x
n y
g mn =
sin
dx dy ... (53)
g ( x, y ) cos
ab 0 0
a
b
m x
n y
4 a b
wy ( x, y ) sin
w ymn =
cos
dx dy . (54)
ab 0 0
a
b
4 a b
m x
n y
hmn =
cos
dx dy .. (55)
h( x, y ) sin
0
0
ab
a
b
4 a b
m x
n y
qmn =
sin
dx dy . (56)
q ( x, y ) sin
ab 0 0
a
b
wxmn =
Even though these are not a full representation of a Fourier series these
functions will give exact values in the intervals 0 x a and 0 y b .
By substituting Eq. 47 through 51 in Eq. 39 and differentiating when it is
required yields:
2
3
2
m 3
m n
n
m n
g mn
hmn +
hmn +
g mn +
a b
b
a b
m =1 n =1
a
2
2
m x
n y qmn
m x
n y
m n
m n
=
sin
sin
sin
+ (1 )
w ymn sin
wxmn + (1 )
a
b
a
b
a
b
D
a
b
m =1 n =1
.. (57)
Equating terms result in the following equation to be satisfied:
12
n
n
m
(g mn + (1 ) wxmn ) +
g mn +
b
a
b
3
3
2
q
m
n
(hmn + (1 ) w ymn ) = mn
hmn +
a
b
D
.. (58)
Equation 58 is to be equated term by term for every m = 1, 2, 3, 4,. and n
= 1, 2, 3, 4,.. Thus the solution can be achieved by selecting a set of
coefficients pjk for j = 1, 2, 3, ., r and k = 1, 2, 3, ., t and for an
acceptable set of n and m then integrate Eq. 52 through 56 numerically with
an acceptable accuracy and see if Eq. 58 is satisfied for every n and m. If it is
not satisfied then updated pjk with an acceptable conversion algorithm such
as Newton Raphson Method with the Jacobian matrix. As we said before
there is one deflection per load thus the coefficient pjk must be unique, thus
there is one root for the solution of Eq. 58. The initial vector pjk can be taken
by letting g(x,y) = wx and h(x,y) = wy so that the result gives the solution for
small deflections as a start (see Timoshenko Eq 101, 102 pp81) this makes
Eq. 58:
m 3 n 2 m
+
a
b a
wxmn +
b
wymn = mn .. (59)
It is seen that this becomes a matrix inversion for pjk if jk = mn and the
matrix to be inverted is an (mn) x (mn) inverted matrix. All of the
coefficients of the left of Eq. 52 to 55 can be evaluated exactly without
numerical integrations. The proposed general solution promises to be exact
for an acceptable accuracy for the deflection. In the real world depending on
the application there always a defined acceptable tolerance set by the
engineer.
To control this tolerance it is best to represent Eq. 40 as follows:
j
r
t
D
x y
w
p jk . (60)
=
4
b
a b
j =0 k =0
When b > a, this is done in order to have a more conversion series with a
good representation of the deflection for 0
13
x
y
1 and 0 1 .
a
b
Boundary Condition:
The boundary condition of the plate at the edges or internally can be
achieved by selecting a conversion series for Eq. 60 that satisfies the
boundary conditions for any set of j and k. This will become evident the
following application examples.
Section 1:
A. Clamped Rectangular Plate at the boundary edge:
By inspection at x = 0 and x = a Eq. 60 must have a root at x = 0 and x =
a to have zero deflections. Also for a clamped plate wx = 0 at x = 0 and x
= a so Eq. 60 must have another root at x = 0 and x = a to have zero
slopes. This can also be said for the boundaries at y and Eq. 60 can be
written as:
2
D
x x y y
w = 1 1
4
b
a a b b
2 r
x y
p jk
a b
j =0 k =0
t
j +3
j+2
k +4
k +3
k +2
r
t
x j + 4
y
y
x y
x
= p jk 2 + 2 +
b
b
a b
a
j =0 k =0
a
.. (61)
Thus wxx and wyy can be written as:
j +2
j +1
j
r
t
a2D
x
x
x
wxx = p jk ( j + 4)( j + 3) 2( j + 3)( j + 2) + ( j + 2)( j + 1)
b4
a
a
a
j = 0 k =0
k +3
k +2
y k + 4
y
y
2 +
b
b
b
. (62)
k +2
k +1
k
r
t
D
y
y
y
w yy = p jk (k + 4)(k + 3) 2(k + 3)(k + 2) + (k + 2)(k + 1)
b2
b
b
b
j = 0 k =0
j +3
j+2
x j + 4
x
x
2 +
a
a
a
. (63)
Using Eq. 62 and 63 we can find the moments at boundary knowing the
slopes are zero at the boundary such that 1/rx = -wxx at x = 0 etc. and the
moments becomes:
14
Mx
x=0
My
x =0
= 2
b4
a2
= M x
k +3
k +2
y k + 4
y
y
2
p
+
0 k
b
b
k =0
b
t
x =0
k +3
k +2
y k + 4
b
y
y
= 2 2 p jk 2 +
a j = 0 k =0
b
b
b
4
Mx
x= a
My
x =a
My
y =0
Mx
y =0
= M x
x =a
j +3
j +2
r
x j + 4
x
x
= 2b 2 p j 0 2 +
a
a
j =0
a
= M y
y =0
j +3
j +2
x j + 4
x
x
2
= 2b p jk 2 +
a
a
j =0 k =0
a
r
My
y =b
Mx
y =b
(64)
= M y
.. (65)
x =b
We can see from Eq. 64 and 65 that the moments on the boundaries are
different functions with different coefficients pjk and Eq. 61 is a
representation of a clamped plate.
B. Clamped Rectangular Plate at the boundary edge and symmetric
load:
For a symmetric load we first translate the deflection to the center of the
plate and look at the deflection we have:
r
t
x 2 1
D
w = p jk
b4
4
j =0 k =0
a
j+2
y 2 1
4
b
k +2
.. (66)
a
b
and y = satisfied the boundary conditions. When
2
2
j+2
y 2 y
b b
15
k +2
.. (67)
x =0
My
x =0
b4
= 2 2
a
= M x
My
x =a
x=a
= 2
k +2
x =0
4
Mx
y 2 y
p0 k
k =0
b b
t
b
a2
= M x
p
k =0
y y
b b
2
0k
k +2
... (68)
x= a
x 2 x
= 2b p j 0
j =0
a a
r
j +2
My
y =0
Mx
y =0
= M y
y =0
x 2 x
2
= 2b p j 0
j =0
a a
r
My
y =b
Mx
y =b
= M y
j +2
(69)
x =b
j
j +1
2
x 2 x
x x x
p jk ( j + 2)( j + 1)2 a 1 a a + 2( j + 2) a a
j =0 k =0
y 2 y
b b
b2
w yy =
D
k +2
k
k +1
2
y 2 y
y y y
x 2 x
a a
j +2
(70)
Substitute x = a/2 and y = b/2 in Eq. 70 yields:
16
2b 4
Da 2
1
p jk ( j + 2 )
4
j =0 k =0
2b 2
w yy =
D
1
p jk (k + 2 )
4
j =0 k =0
wxx =
j +1
j +2
1
4
1
4
k +2
k +1
b4
2 Da 2
1
p jk ( j + 2 )
4
j =0 k =0
r
b2 r t
1
p jk (k + 2 )
=
2 D j = 0 k =0
4
j + k +1
j + k +1
(72)
Since wx = wy = 0 at x = a/2 and y = b/2 the moments at the center becomes:
b2
M x ==
2
1
b2
p
jk 2 ( j + 2 ) + (k + 2 )
j =0 k =0
4
a
j + k +1
b2
M x ==
2
b2
p jk (k + 2 ) + 2 ( j + 2 )
a
j =0 k = 0
4
j + k +1
(73)
(74)
For a square plate a = b then w is the same for the coordinate (x,y) and (y,x)
then j=k and we can write pjk = pkj = pjj = pj and substitute in Eq. 67 yields:
r
r
x 2 x y 2 y
D
=
w
p
j
b4
j =0 j = 0
a a a a
2 j+4
. (75)
(a,b)
x a/2 y b/2
1 = 0
+
Fig. 1- Clamped Rectangular Plate with a Clamped ellipse Built in the center
17
D
x x y y a x 1 b y 1
w
1
1
1
4
b
a a b b c a 2 d b 2
j +3
j+2
k +4
k +3
k +2
x j + 4
y
y
x y
x
= p jk 2 + 2 +
b
b
a b
a
j =0 k =0
a
r
a 2 x 1 2 b 2 y 1 2
+ 1
c a 2 d b 2
j +3
j+2
k +4
k +3
k +2
r
t
x j + 4
y
y
x y
x
= p jk 2 + 2 +
b
b
a b
a
j =0 k =0
a
a 2 x 2 x 1 b 2 y 2 y 1
+ + +
a 4 d b
b 4
c a
(76)
One can see Eq. 76 can be expanded to give another expression on Taylor
polynomial and the solution can be obtained.
D. Clamped Triangular Plate:
Similar to what we have done before by inspection at x = 0 and y = 0 and
at the edge of triangle Eq. 60 must have a root at x = 0 and y = 0 and at
the edge of triangle (see Fig 2) to have zero deflections. Also for a
clamped plate wx = 0 at x = 0 and at the edge of triangle wy = 0 at y = 0
and at the edge of triangle then Eq. 60 must have another root at x = 0
and y = 0 and at the edge of triangle to have zero slopes and Eq. 60 can
be written as:
2
D
x y x y
w = + 1
4
b
a b a b
2 r
x y
p jk . (77)
a b
j =0 k = 0
t
18
x y
p jk
a b
j =0 k =0
r
(0,b)
x y
+ 1 = 0
a b
(a,0)
D
x y x y
w = + 1
4
b
a b a b
x y
p jk . (78)
a b
j =0 k =0
r
19
y
(0,b)
x y
+ 1 = 0
a b
(a,0)
r
t
D
x y
x y
2
w
R
x
y
p jk .... (79)
[
(
,
)
]
=
4
b
a b
a b
j =0 k =0
R( x, y ) = p jk x j y k
.. (80)
j = 0 k = 0
20
y
(0,b)
R ( x, y ) = 0
(a,0)
21
not an issue then the initial value of Eq. 59 is sufficient for solution
which requires only one matrix inversion.
We give the equations for loosing supports for a symmetric loading on a
rectangular plate with any edge condition as follows:
x
dx
1 a
a
2 0 1 + [w ( x, b / 2)]2
dy
1 b
y
b
2 0 1 + w (a / 2, y ) 2
y
(81)
y
y
(a,b)
D
x y
x x y y r t
w
=
1 1 p jk
4
b
a b
a a b b j =0 k =0
x j + 2 x j +1 y k + 2 y k +1
= p jk
b
a b
j =0 k =0
a
r
.. (82)
22
aD
y
x y
x
wx = p jk ( j + 2) ( j + 1)
b4
b
a b
a
j = 0 k =0
. (83)
k +1
k
j+2
j +1
r
t
D
x
y x
y
w y = p jk (k + 2) (k + 1)
b3
a
b a
b
j =0 k = 0
.... (84)
a2D
y
x y
x
wxx = 2 p0 k + p jk ( j + 2)( j + 1) ( j + 1) j
b4
b
a b
a
k =0
j =1
. (85)
j +2
k 1
j+2
j +1
r
t
x
y x
y
w yy = 2 pk 0 + p jk (k + 2)(k + 1) (k + 1)k
b2
a
b a
b
j =0
k =1
.... (86)
Using the equation for finding the moments at the edges we write:
Mx
x =0
x =a
= D
wxx
[1 + (w ) ]
x
My
x =0
x=a
= D
3
2 2
w yy
[1 + (w ) ]
yy
3
2 2
w yy
[1 + (w ) ]
3
2 2
=0
wxx
[1 + (w ) ]
xx
3
2 2
(87)
=0
When solving the two equation for 1/rx = 0 and 1/ry = 0 at x = 0 and x = a,
we see the only way Eq. 87 is satisfied is when
wxx
x =0
x =a
= 0 and w yy
x =0
x =a
= 0 (88)
y =0
y =b
= 0 and w yy
y =0
y =b
= 0 (89)
23
We can see that is true already by substitution in Eq. 85 and Eq. 86 that:
wxx
y =0
y =b
= 0 and w yy
x =0
x =a
=0
at
x = 0 (90)
Thus set p0k = p1k in the polynomial of Eq. 82 and the condition is
satisfied.
t
r
y k + 2 y k +1
0 = 2 p0 k + p jk 2( j + 1)
b
k =0
j =1
b
at
x=a
. (91)
k +2
k +1
y
y
0 = 6 p0 k + p jk 2( j + 1)
b
k =0
j =2
b
Or set
1 r
p0 k = p1k = p jk 2( j + 1) .. (92)
3 j=2
t
1 t
p jk 2(k + 1) . (93)
3 k =2
j +1
y 2 1
4
b
24
k +1
.. (94)
a
b
and y = satisfied the boundary conditions. When
2
2
j +1
y 2 y
b b
k +1
.. (95)
y 2 y
b b
k +1
. (96)
k
2
r
y 3 y 2 y t
y y y
D
y
w y = p j 0 2 1 + 2 p j1 2 3 + + (k + 1) p jk 2 1
b3
b b k =2
b
j =0
b b b
b
x 2 x
a a
j +1
. (97)
And wxx and wyy becomes:
j 1
2
t
x 2 x r
x x x
a2D
wxx = 2 p0 k + 2 p1k 6 6 + 1 + ( j + 1) jp jk 2 1
b4
a j = 2
k =0
a a a
a
y 2 y
b b
k +1
(98)
k 1
2
r
y 2 y 1 t
y y y
D
w yy = 2 p j 0 + 2 p j1 6 6 + 1 + (k + 1)kp jk 2 1
b2
b
j =0
b b b
b
k =2
x 2 x
a a
j +1
.. (99)
25
y =0
y =b
x =a
must have:
pj0 = - pj1 and p0k = - p1k . (100)
Thus when selecting these coefficients and substituting in Eq. 94 the
boundary condition are satisfied.
Now we seek the moments at x = a/2 and y = b/2. First we find wxx and wyy
from Eq. 98 and Eq. 99 and using Eq. 100 yields:
6b 4 t
1
wxx = 2 p0 k
a D k =0
4
6b 2
w yy =
D
1
p j 0
4
j =0
r
k +1
j +1
. (101)
1
p0 k
4
k =0
t
1
M y = 6b 2 p j 0
4
j =0
r
k +1
j +1
1
6b p j 0
4
j =0
r
j +1
6b 4
2
a
1
p0 k
4
k =0
t
k +1
.. (102)
For a square plate a = b then w is the same for the coordinate (x,y) and (y,x)
then j=k and we can write pjk = pkj = pjj = pj and substituting in Eq. 95 yields:
r
r
x 2 x y 2 y
D
w = p j
b4
j =0 j = 0
a a a a
2 j+2
. (103)
Many other boundary conditions can be satisfied with finding the proper
Taylor series expansion. One more conditions will be addressed for the case
of a rectangular plate with one free edge and the others are clamped edge.
26
Section 3:
A. Three Sided fixed rectangular Plate and one free edge:
Consider Fig 6.
(a,b)
x = z(y)
L
y
z ( y ) = Ai (104)
b
i =0
D
x x y
w = 1
4
b
a a b
2 r
x y
p jk .. (105)
a b
j =0 k =0
t
27
Mx
x= z ( y )
= D
wxx
[1 + (w ) ]
x
My
x= z ( y )
= D
3
2 2
w yy
[1 + (w ) ]
yy
3
2 2
w yy
=0
[1 + (w ) ]
3
2 2
(106)
wxx
[1 + (w ) ]
xx
3
2 2
=0
z( y)
0
dx
1 + (w x ( x , y ) )
= a . (107)
Which makes the solution complicated, but the solution can be enforced a
certain coordinates by taking a set of known yi say (for example dividing b
to increments b/T and yi = ib/T , i = 0, 1,2, ..) then each Eq 106 and 107 is a
set of i equations to add more equations to Eq. 58 and the algorithm of
finding pjk also finds Ai. Once Ai is found Eq. 106 and 107 can be used to
verify the accuracy of other points beside yi.
Non-Prismatic Plates
If we have a set of square plates with different thickness welded or attached
together to make one big plate (where the perimeter can have triangular
plates if needed) then we need to match the slope and deflection around each
plate. This can be done by matching a set of points xi and yi for the function
x = z1(y), y = z2(x) at the perimeter similar to the last example. Even though
this starts to look like finite element it is a much better representation to
include large deflection and more accurate.
28
q(x)
x
L
q ( x) = an sin
n =1
n x
L
. (108)
Now we pick Taylor series y(x) that satisfies the boundary condition. Thus
the series must have a root at x = 0 and x = L and
M ( x) x =0 = &y& x =0 = 0 and M ( x) x = L = &y& x = L = 0
We will show the behavior of the solution for a 5th , 6th , 7th , 8th and 9th order
polynomial and compare the results. We find the following equation satisfies
the boundary condition:
5th order:
y ( x) =
L4
EI
4
3
x 5 5
k3 x
k x
x 2
+
+
k
k
k
+ k1 3 (109)
1
1
3
2 L
2 L
L 3
L 3
6th order:
6
5
4
3
k 4 x
k 4 x
5
2
L4 x
x 3
x 5
y ( x) =
k1 + k 2 k1 + k 2 + + k 4 + k1 + k 2
3
2 L
3
2 L
EI L
L 2
L 2
(110)
29
7th order:
y ( x) =
L4
EI
6
5
4
x 7
k5 x
5
5
x
x 7
k
k
k
k
k
k
+
+
+
+
+
1
1
2
3
2
3
2
3
2 L
L
L 2
L
k x
3
2
x 5
+ k5 + k1 + k 2 + k3 5
2
3
2 L
L 2
3
. (111)
8th order:
8
7
6
5
4
k 6 x
L4 x
7
5
5
x
x
x 14
y ( x) =
k1 + k 2 + k3 + k 4 k1 + k 2 + k 3 + k 4 +
EI L
2
2
3
2 L
L
L
L 3
3
k x
5
3
2
x 11
+ k6 + k1 + k 2 + k3 + k 4 6
2
2
3
2 L
L 3
. (112)
9th order:
9
8
7
6
5
4
k 7 x
L4 x
14
7
5
5
x
x
x
x
y ( x) =
k1 + k 2 + k3 + k 4 + k5 6k1 + k 2 + k3 + k 4 + k5 +
EI L
3
2
2
3
2 L
L
L
L
L
3
k x
11
5
3
2
x
+ k7 + 5k1 + k 2 + k3 + k 4 + k5 7
3
2
2
3
2 L
L
. (113)
Where k1 , k 2 , k 3 , k 4 , k5 , k 6 and k 7 are constants to be found. Now following the
proposed solution we express the equation for the slope in Fourier series as:
y& = bn cos
n=0
n x
L
. (114)
where
bn =
1
2 L
n x
dx = 2 y ( Lu ) cos nu du
y ( x) cos
0
L 0
L
When integrating Eq. 109, 110, 111, 112 and 113 using Eq. 114 then
differentiating three time to get the pressure and equating to Eq. 108 we
obtain a set linear system of equations in ki. When inverting the matrix for
each polynomial equation we obtain the following solutions:
30
5th order:
k1
-0.052359878
a1/2
-0.130899694
-0.087266463
a2/2
=
k3
.. (115)
6th order
k1
k2
k4
-0.024223654
0.072670961
0.072670961
-0.052359878
-0.104755807
-0.087266463
-5.0307E-17
0.06562175
-0.024223654
-0.229676124
0.072670961
0.247129416
-0.104755807
-0.047469887
0.004447133
a1/2
a2/2
-0.078431662
a3/2
(116)
-0.218012883
7th order
k1
k2
k3
k5
-5.55112E-17
-0.131243499
0.072670961
0.459352247
-0.180108893
1.249E-16
0.277945823
-0.017788533
0.06562175
0.720435572
-0.131243499
-1.11178329
0.003812983
-0.229676124
-1.062812804
0.459352247
1.752289761
0.050820917
0.247129416
0.66691391
-0.598978587
-1.365627767
-0.106136585
-0.047469887
-0.022510129
-0.079593151
-0.086298663
a1/2
a2/2
-0.218012883 -0.598978587
a3/2
-0.078431662 -0.079593151
a4/2
(117)
8th order
k1
k2
k3
k4
k6
a1/2
a2/2
a3/2
a4/2
a5/2
. (118)
9th order
k1
k2
k3
k4
K5
k7
2.7452E-16
0.004447133
-0.017788533
0.003812983
0.050820917
-0.106136585
-0.040478794
0.182154574
-0.269531601
0.093305925
0.074523349
-0.053491716
-5.35683E-15
-0.180108893
0.720435572
-1.062812804
0.66691391
-0.022510129
0.518128567
-2.331578551
4.158719394
-3.674817977
1.610379068
-0.002513738
9.13158E-15
0.277945823
-1.11178329
1.752289761
-1.365627767
-0.086298663
-0.655756467
2.950904103
-5.429484287
5.23230919
-2.796218283
-0.097553632
(119)
31
a1/2
a2/2
a3/2
a4/2
a5/2
a6/2
x
cL
L
q ( x ) = 2q
n c
L sin n x .. (120)
n
L
1 cos
n =1
Thus:
an
=
2
1 cos
n
n c
L ..... (121)
Substituting in Eq. 115, 116 117, 118 and Eq. 119, and graphing the results
for c = .7 yields Fig 11A and Fig 11B; Table 1 shows the results in
comparing with the exact solution.
2- Expressing the loading of Fig. 9 in Fourier Series to obtain the pressure
q ( x ) = 4 q
i =1
sin
n c
n
sin
L
L sin n x (122)
n
L
32
x
(c )/L
(c + )/L
cL
L
n
L sin n x .... (123)
n
L
L
sin
Let go to zero and we get the pressure representing a point load as follows:
q ( x ) = 2
P
n c
n x
.. (124)
sin
sin
L i =1
L
L
Thus:
an
P
n c
= sin
(125)
2
L
L
Substituting in Eq. 115, 116 117, 118 and Eq. 119, and graphing the results
for c = .3 yields Fig 12A and Fig 12B; Table 2 shows the results in
comparing with the exact solution.
33
1 cos
n c
n x
L
(126)
q ( x) = 4q cos
sin
L
n
L
i =1
x
(c )/L
(c + )/L
cL
-q
L
1 cos
M
n c
n x
L
q ( x) = 4 2 n cos
sin
.... (127)
2
L i =1
L
L
n
34
M
q ( x) = 2 2
L
n cos
i =1
n c
n x
... (128)
sin
L
L
Thus:
an
M
n c
= 2 n cos
.... (129)
2
L
L
Substituting in Eq. 115, 116 117, 118 and Eq. 119, and graphing the results
for c = .5 yields Fig 13A and Fig 13B; Table 3 shows the results in
comparing with the exact solution.
All of the results shows Taylor approximation follows Fourier
approximation and the error improves with adding higher term to Taylor
polynomial.
For the point load we can see that under the pressure
dV
dx
so the
x =cL
pressure approaches infinity under the load. This has happen because we
approximated the actual shear, which is a discontinuous function, by a
continuous function. As in the actual shear curve at that point the shear is
never defined under the load2 since it has two values and the question
becomes which value can we use. In practice the shear is always taken as
max(V (cL ) , V (cL + ) ) for an appropriate increment where can be found
by testing for ultimate values around cL. Thus the solution for the shear is
correct for all values except under the load and can be taken as
max(V (cL ) , V (cL + ) ) of the approximate Taylor polynomial. If we have a
load that has been approximated with many point loads then it is best to
determine what shear value to use under the load based on design practice
and represent Taylor polynomial approximation as a discontinuous function
with point values under the load. If we differentiate the shear to obtain the
pressure then the pressure under the load can not be determined using Taylor
polynomial as a continuous function and should be taken
P = V (cL ) V (cL + ) .
The reason the actual shear diagram has discontinuity because it is telling us in real life there no such
thing as a point load and in reality it is some kind of a pressure with some small around the load as
in Fig 9.
35
Similarly from the moment diagram we can see that under the shear
dM
dx
so the shear approaches infinity under the load. This has happen
x =cL
The reason the actual moment diagram has discontinuity because it is saying in real life there no such
thing as a moment at a point of application and in reality it is some kind of a pressure with some small
around the load as in Fig 10. For example if we try to put a moment using a pinion of a motor then in
reality the pinion of the motor could never have a zero radius and the radius can only be as small as and
transferring the load can only be possible by introducing some kind of a axis-symmetric pressure at the
point of application from - x cL .
36
c/L
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
c/L
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
c/L
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
c/L
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
37
0.12
1.4
0.1
1.2
0.08
1
q ( act ual)
0.8
0.06
Four ier
0.6
Moment Taylor
0.04
Taylor
0.4
0.02
0.2
0
- 0.2
0.2
0.4
0.6
0.8
1.2
0.2
0.4
0.6
0.8
1.2
- 0.02
x/ L
x/ L
1.5
0.12
0.1
1
0.08
0.5
q ( act ual)
0.06
Fourier
0.2
0.4
0.6
0.8
Moment Taylor
0.04
Taylor
0
1.2
0.02
- 0.5
0
0
-1
0.2
0.4
0.6
0.8
1.2
- 0.02
x/ L
x/ L
0.12
1.5
0.1
0.08
0.5
q ( act ual)
0.06
Four ier
0
0
0.2
0.4
0.6
0.8
1.2
Moment Taylor
0.04
Taylor
-0.5
0.02
-1
- 1.5
- 0.02
0.2
0.4
0.6
0.8
1.2
x/ L
x/ L
38
0.12
0.1
1.5
0.08
1
q ( act ual)
0.06
Four ier
Moment Taylor
0.04
Taylor
0.5
0.02
0
0
0.2
0.4
0.6
0.8
1.2
0
- 0.5
0.2
0.4
0.6
0.8
1.2
- 0.02
x/ L
x/ L
1.8
0.12
1.6
0.1
1.4
1.2
0.08
1
q ( act ual)
0.8
0.06
Four ier
0.6
Moment Taylor
0.04
Taylor
0.4
0.02
0.2
0
- 0.2 0
0.2
0.4
0.6
0.8
1.2
0.2
0.4
0.6
0.8
1.2
- 0.02
- 0.4
x/ L
x/ L
39
c/L
~1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
c/L
~1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
40
M+ max
error
-9.96%
-7.81%
-11.17%
-15.09%
-13.36%
-11.53%
-13.36%
-15.09%
-11.17%
-7.81%
1.4
1.2
0.2
1
0.8
0.15
0.6
V( act ual)
0.4
Four ier
0.2
Taylor
Moment Taylor
0.05
0
- 0.2
0.1
0.2
0.4
0.6
0.8
1.2
0
- 0.4
0.2
0.4
0.6
0.8
1.2
- 0.05
- 0.6
x/ L
x/ L
1.4
1.2
0.2
1
0.8
0.15
0.6
q ( act ual)
0.4
Fourier
0.2
Taylor
0
- 0.2
Moment Taylor
0.2
0.4
0.6
0.8
1.2
0
0
- 0.4
0.2
0.4
0.6
0.8
1.2
- 0.05
- 0.6
x/ L
x/ L
0.25
0.8
0.2
0.6
0.15
0.4
V ( act ual)
0.2
Fourier
0
- 0.2
0.2
0.4
0.6
0.8
1.2
Moment Taylor
Taylor
0.05
- 0.4
- 0.6
0.2
0.4
0.6
0.8
1.2
- 0.05
- 0.8
x/ L
x/ L
41
1.2
0.25
1
0.2
0.8
0.6
0.15
V ( act ual)
0.4
Fourier
0.2
0.1
- 0.2
Moment Taylor
Taylor
0.05
0
0.2
0.4
0.6
0.8
1.2
0
- 0.4
- 0.6
0.2
0.4
0.6
0.8
1.2
- 0.05
x/ L
x/ L
1.2
1
0.2
0.8
0.15
0.6
V ( act ual)
0.4
0.1
Fourier
0.2
Moment Taylor
Taylor
0.05
0
0
0.2
0.4
0.6
0.8
1.2
- 0.2
0
0
- 0.4
0.2
0.4
0.6
0.8
1.2
- 0.05
- 0.6
x/ L
x/ L
42
c/L
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
c/L
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
c/L
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
c/L
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
43
0.01
0.6
0.008
0.4
0.006
0.004
0.2
0.002
y( act ual)
Four ier
- 0.002 0
0.2
0.4
0.6
0.8
1.2
Taylor
0.2
0.4
0.6
0.8
1.2
Moment Taylor
- 0.2
- 0.004
- 0.006
- 0.4
- 0.008
- 0.6
- 0.01
x/ L
x/ L
0.01
0.6
0.008
0.4
0.006
0.004
0.2
0.002
q ( act ual)
Fourier
- 0.002 0
0.2
0.4
0.6
0.8
1.2
0
0
Taylor
0.2
0.4
0.6
0.8
1.2
Moment Taylor
- 0.2
- 0.004
- 0.006
- 0.4
- 0.008
- 0.01
- 0.6
x/ L
x/ L
0.01
0.008
0.4
0.006
0.004
0.2
0.002
y ( act ual)
Moment ( act ual)
Fourier
0
- 0.002 0
0.2
0.4
0.6
0.8
1.2
Taylor
0.2
0.4
0.6
0.8
1.2
Moment Taylor
- 0.2
- 0.004
- 0.006
- 0.4
- 0.008
- 0.01
- 0.6
x/ L
x/ L
44
0.01
0.6
0.008
0.4
0.006
0.004
0.2
0.002
y ( act ual)
0
- 0.002
Four ier
0
0.2
0.4
0.6
0.8
1.2
0
0
Taylor
0.2
0.4
0.6
0.8
1.2
Moment Taylor
- 0.2
- 0.004
- 0.006
- 0.4
- 0.008
-0.01
- 0.6
x/ L
x/ L
0.01
0.6
0.008
0.4
0.006
0.004
0.2
0.002
y ( act ual)
Fourier
0
- 0.002
0.2
0.4
0.6
0.8
1.2
0
0
Taylor
0.2
0.4
0.6
0.8
1.2
Moment Taylor
- 0.2
- 0.004
- 0.006
- 0.4
- 0.008
-0.01
- 0.6
x/ L
x/ L
45
From this exercise and examples we know what to expect from our plate
solution. By using Timoshenko pp 111 Eq. 133 the point load on a plate
can be introduced using the following pressure:
q ( x, y ) = a mn sin
m =1 n =1
m x
n y
sin
. (130)
a
b
Where
amn =
4P
m
n
sin
sin
. (131)
ab
a
b
Base on the analysis in one dimension we conclude that the shear at the
point under the load can be taken as
[max( Qx ( 1 , + 2 ) , Qx ( + 1 , + 2 ) , Qx ( 1 , 2 ) , Qx ( + 1 , 2 ) )]
and
max Q y ( + 1 , 2 ) , Q y ( + 1 , + 2 ) , Q y ( 1 , 2 ) , Q y ( 1 , + 2 )
.. (132)
for an appropriate increments (1 , 2 ) where (1 , 2 ) can be found by testing
for ultimate values around the point ( , ) . If we have a load that has been
approximated with many point loads then it is best to determine what
shear value to use under the load based on design and represent Taylor
polynomial approximation as a discontinuous function with point values
under the load. If we differentiate the shear to obtain the pressure then the
pressure under the load can not be determined using Taylor polynomial
as a continuous function and should be taken
P = Qx ( 1 , 2 ) Qx ( + 1 , + 2 )
or
P = Q y ( 1 , 2 ) Q y ( + 1 , + 2 )
. (133)
For the moment pressure equation in the x direction use Eq. 130 and Eq.
131 and let there be two point load in opposite direction at the coordinate
( , ) and ( + , ) then Eq. 131 becomes
46
amn =
n
4P
sin
ab
b
m ( + )
m
m
n
8P
m ( )
=
sin
sin
cos
sin
sin
a
a
ab
a
a
b
(134)
Now let the moment due to the point load be applied at the point ( , ) as
M 0 x = 2P . By substituting M0x in Eq. 134 we have:
amn =
4M 0 x
m
a 2b
m
a cos m sin n . (135)
m
a
b
a
sin
Let got to zero then we the pressure for a moment in the x direction
using Eq. 130 with the following:
amn =
4M 0 x
m
n
m cos
sin
. (136)
2
a b
a
b
Repeating the above analysis for the pressure for a moment in the y
direction using Eq. 130 with the following:
amn =
4M 0 y
ab
n sin
m
n
... (137)
cos
a
b
And the four moments surrounding the point under the point of
application can be taken as
M x ( 1 , + 2 ), M x ( 1 , 2 ), M x ( + 1 , + 2 ), M x ( + 1 , 2 )
and
M y ( + 1 , 2 ), M y ( 1 , 2 ), M y ( + 1 , + 2 ), M y ( 1 , + 2 )
.. (138)
for an appropriate increments (1 , 2 ) where (1 , 2 ) can be found by testing
for ultimate values around the point ( , ) . Thus, it is best to determine the
moment values to under the point of application based on design practice
and represent Taylor polynomial approximation as a discontinuous
47
function with point values under the load. If we differentiate the moment
to obtain the shear then the shear under the load can not be determined
using Taylor polynomial as a continuous function and should be taken
1
1
Qx = max M xy ( 1 , 2 ) M xy ( 1 , + 2 ) , M xy ( + 1 , 2 ) M xy ( + 1 , + 2 )
b
b
1
1
+ max [M x ( 1 , + 2 ) M x ( + 1 , + 2 )] , [M x ( 1 , 2 ) M x ( + 1 , 2 )]
a
and
1
1
Q y = max M xy ( 1 , + 2 ) M xy ( + 1 , + 2 ) , M xy ( 1 , 2 ) M xy ( + 1 , 2 )
a
a
1
+ max M y ( + 1 , 2 ) M y ( + 1 , + 2 ) , M y ( 1 , 2 ) M y ( 1 , + 2 )
b
b
] [
. (139)
Comment on selecting , 1 and 2 :
A practical selection of , 1 and 2 are recommended by Professor John
Stanton saying the point load in a concrete slab can be seen as a cone
propagating in the thickness of the slab. Thus an absolute smallest
increment of , 1 and 2 is the thickness of the plate t shaped in a circle.
This also becomes a restriction when subdividing a pressure function into
point loads for large deflection analysis and it becomes a condition of
using the solution in realm of elasticity. Loads that need finer increments
then twice the thickness of the slab cannot be approximated into point
loads and should be addressed differently when large deflection is the
issue. If large deflection is not of concern then Eq. 59 is the best
alternative and has been the methods used in standard practice for ages.
Final analysis:
As we can see any load can approximated by point loads for a more
conservative solution. However, the increment of divisions has a limiting
value as discussed above. The interesting part is the Cartesian solution is
it is sufficient solution and other coordinates transformations are not
necessary provided the load is contained in the boundary condition. For
example for a simply supported circular plate the boundary condition can
be satisfied as long as the load is contained in the circle.
48
dx
xi
0
i =
1 + [wx ( x, yi )]
yi
0
dy
1 + w y ( xi , y )
(140)
With these additional equations the solution can be found exactly for
point loads moments. The solution is similar to finding the coefficients
for large deflection of a beams. Thus we start with an initial value for the
Taylor polynomial coefficients plus the coefficients (xi, yi) and the free
boundaries and update numerically, and the solution becomes exact.
49