Exact and Numerical Solutions for Large Deflection of Elastic

Non-Prismatic Beams

by

Farid A. Chouery1

Abstract:

The solution for large deflections of beams that has not been solved in general in 260 years is

now presented in this paper for point loads and moments in any directions along the beam with

various end conditions. Also, the solution can be used for non-prismatic beams with various end

conditions and numerical solution is presented to obtain exact solutions. Curvilinear beams and

extensibility along the beam are also addressed.

Introduction:

The large deflection of beams has been investigated by Bisshopp and Drucker [1] for a point load

on a cantilever beam. Timoshenko and Gere [2] developed the solution for axial load. Virginia

Rohde [3] developed the solution for uniform load on cantilever beam. John H. Law [4] solved it

for a point load at the tip of the beam and a uniform load combined. In this paper the general

solution developed for a prismatic beam and in some cases for non-prismatic. However,

numerical integration maybe needed along with solving compatibilities equations for the

constants of integrations. A more general and preferable numerical solutions for a non-prismatic

beam is also given using only many point loads acting with an angle on the beam with a moment

on the node representing the approximate load. This point load can take any direction on the

1
Structural, Electrical and Foundation Engineer, US Army Corps of Engineers Seattle District

1
beam bending in the x-x direction or y-y direction of the moment of inertia. Thus the load is to be

resolved to x-x direction and y-y direction of the moment of inertia in each orthogonal deflection

given two non-linear differential equations. By solving each non-linear differential equation the

orthogonal deflection components can be obtained.

An approximation attempt has been investigated by Scott and Caver [5] for all problems in

which the moment can be expressed as a function of the independent variable. However, the

solution presented here is not an approximation and is of a closed form. As a consequence one

can take a load function and divide it into line segments for a non-prismatic beam with no axial

loads and by using the proposed solution the large deflection can be calculated with reasonable

accuracy. Thus, by taking smaller and smaller increments the accuracy is improved.

The solution is based on solving the non-linear differential equation of Bernoulli-Euler beam

theory;

y M ( x)
 …………………………………………………………………. (1)
 1  ( y ) 
2
3
EI ( x)

Where M(x) is the moment in the direction that corresponds to the moment of inertia I(x), E is

the modulus of elasticity, y is one of the orthogonal deflection say for Ix-x. Thus, if the other

deflection is desired, then another equation is needed where M(x) is the moment in the direction

that corresponds to the moment of inertia Iy-y(x). It is assumed the modulus of elasticity is

constant and the bending does not alter the length of the beam. Three different closed form

solutions are investigated for three different cases of the non-linear differential equation. In many

ways M(x) is not known until the final deflection is known so will assume M(x) is known in the

equations.

2
Case I:

M ( x)
In this case assume  f ( x) , a function of x only. Thus,
EI ( x)

y
 f ( x) ………………………………………………….…………… (2)
 1  ( y ) 
2
3

1 
Let y  tan   1  ( y ) 2  and y 
cos cos 2 

Where is a new variable, then substitute inq. 2 yields,

 cos  f ( x)   cos d   f ( x) dx  C1 or sin    f ( x) dx  C1

cos   1  sin    1 
2

 f ( x) dx  C 1
2
2


Thus,

y 
sin 

 f ( x) dx  C 1
………………………………………………...……. (3)
1   f ( x) dx  C 
cos 2
1

y ( x)   
 f ( x) dx  C 1
dx  C 2 ………………………………………………….. (4)
1   f ( x) dx  C 
2
1

Where, C1 and C2 are constants of integration. This off course the solution Scott and Carver

approximated as an infinite series not realizing it can be expressed in a closed form. Note: Eq. 4

gives an integral solution where if the moment is approximated by a curve it can give a better

approximation than small deflection equations approximations. It is seen that if the denominator

of Eq. 4 is approximated as a unity it would give the standard solution for small deflection

approximations. In the case of large deflection it is better to use point loads and moments for

approximating the general loading because it will be shown that it can be presented as elliptical

3
integrals of the first and second kind. Elliptical solutions enable us to have a closed form solution

as it has been successfully done through recursion [7, 8] instead of integration.

Case II:

M ( y)
In this case assume   g ( y ) , a function of y only. This happens in buckling problems
EI

and the moment of inertia is considered constant. Thus,

y
  g ( y ) ……………………………………………….……………… (5)
 1  ( y )  2
3

1 1 dx 1 dx dy x

Let y   y   2
 2  3
x x dx x dy dx x

By replacing yields,

x
 g ( y ) …………………………………………………..……………… (6)
 1  ( x)  2
3

By following case I analysis with interchanging x by y the solution becomes:

x  
 g ( y) dy  C 1
……………….…………………………………….……...……. (7)
1   g ( y ) dy  C 
2
1

x( y )   
 g ( y) dy  C 1
dy  C 2 ………………………………………...………….. (8)
1   g ( y ) dy  C 
2
1

Case III:

4
 M ( x, y )
In this case assume  h(ax  by ) , a function of x and y. This happens in combine
EI

bending and buckling problems with point loads and the moment of inertia is considered

constant. Thus,

y
 h(ax  by ) ……………………………………………………..……………… (9)
 1  ( y )  2
3

Consider the rotation of axis and let:

a b
u x y
a b 2 2
a  b2
2

……………………………………………………..………….. (10)
a b
v x y
a2  b2 a2  b2

Where u and v are the new variables then h(ax  by )  h a 2  b 2 u  

a2  b2 a2  b2
x (u  v)  A(u  v) where A 
2a 2a
……………………………..………. (11)
a2  b2 a2  b2
y (u  v)  B (u  v) where B 
2b 2b

And,

dx
 A(1  v)
du
…………………………………………….………………………….………… (12)
dy
 B(1  v)
du

5
dy B 1  v  a 1  v 
 
dx A 1  v  b 1  v 
………….……………......……. (13)
d 2 y a d 1  v  du a d 1  v  1 a  2v
      
dx 2
b du 1  v  dx b du 1  v  A(1  v) Ab 1  v 3

Or

3
 2
 
y  2a v  1 2 2
   4a b v

 1  ( y ) 
2
3
Ab (1  v) 3  a 2  1  v  2 
1  2   
a2  b2  2
 b2  a2 
2
3

 b2  a2   2
2

 b  1  v    v  2   1   2  
 b  a 2  2 
 b  a  

…………………….……………. (14)

Let

2
b2  a2  b2  a2 
v  2  1   2  tan 
2 
a  b2 a b 
…………………………………………………..…….. (15)
2
 b2  a2  
v  1   2 
2 
 a  b  cos 
2

Where is a new variable, then substitute inq. 14 yields;


 cos  h a 2  b 2 u    
sin     h a 2  b 2 u du  C1
………………....………….. (16)

 
cos   1    h a 2  b 2 u du  C1  
2

6
 2
 b2  a2  b2  a2
v  1   2 
2 
tan  
a b  a2  b2

 a2  b2 
  1   2 
2
 h 
a 2  b 2 u du  C1

a2  b2
a b 
2 
  
1   h a 2  b 2 u du  C1 
2 a2  b2

………………………………..……………. (17)

 a2  b2 
2
 h 
a 2  b 2 u du  C1 a2  b2
v   1   2   du  u  C 2 ………………..... (18)
a b 
2 
 
1   h a 2  b 2 u du  C1  
2 a 2  b2

Thus, after integrating with respect to u, substitute x and y from Eq. 10 and an explicit equation

of deflection in x and y is obtained. Another way of calculating x and y is pick u find v from Eq.

18 then find x and y from Eq. 11.

Hence, the solution for the nonlinear differential equation has been obtained for three cases and

applications will follow.

Application for Case I - Numerical Solution for Any Load Function Non-Prismatic Beam:

This example is to demonstrate the solution for a cantilever beam. Other boundary conditions for

beams are similar. First divide the beam into segmental beams of each length Li and on each

node of the segment insert the equivalent load Pi and moment Qi to approximate the real load,

see FIG.1. The moment on the segment beam at xi is:

7
M 0  P0 ( x  x0 )  Q0 for x 0  x  x1
M 1  P0 ( x  x0 )  Q0  P1 ( x  x1 )  Q1 for x1  x  x 2


i
M i   Pj ( x  x j )  Q j for xi  x  xi 1 ………….........……………… (19)
j 0



n 1
M n 1   Pj ( x  x j )  Q j for x n 1  x  x n
j 0

Where, all xi are unknown.

Pi
P1 Pi+1
P0 P2
L0 L1 Li
Qi Qi+1
x
Q0 Q1 Q2 Qi
Qi+1
Q2 Pi+1
Q1
P2 Pi
Q0
P1
P0 Segment
x0 Beam i
Length = Li
x1
x2
xi
x
xi+1
xn = L

FIG.1 – Cantilever Beam Analysis – Non-Prismatic Beam

8
 Point Loads are in the y direction

Now the moment of inertia:

i
I i  I ( LTi ) where LTi   Li ………………………………………………………….. (20)
j 0

Where, the moment of inertia is approximated at each beam segment to be constant2 and the

moment of inertia function is assumed continuous. Thus:

Mi 1  i 
f i ( x)    Pj ( x  x j )  Q j  ……………………………….....…………………. (21)
EI i EI i  j 0 

Substitute Eq. 21 in Eq. 3 and find the slope on the segmental beam i yield:

1  i 
 0.5Pj ( x  x j )  Q j ( x  x j )  C1i
2

EI i  j 0 
y i ( x)  for xi  x  xi 1
2
 1  i  
1   0.5 Pj ( x  x j )  Q j ( x  x j )  C1i 
2

 EI i  j 0  

………………………..…....…………… (22)

Apply compatibilities yield:

y i 1 ( xi )  y i ( xi ) at x  xi
……………………......……………… (23)
Seg. Bm. i  1 to Seg. Bm. i

At x = L, where L is the length of the beam at x = xn = L,

y n ( xn )  y n ( L)  0 ……………………………………………………………....…………. (24)

Where n is the number of beam segments and n+1 is the total numbers of beam segments; apply

Eq. 24 in Eq. 22 yields:

2
Note: this approximation does not mean there are stress singularities due to sharp corners at the discontinuities
where each segment meets. All it means is the actual deflection of that segment can be approximated with the
deflection of a beam with constant moment of inertia.

9
 1  n 1 
y n ( L)  0   0.5Pj ( L  x j )  Q j ( L  x j )  C1n 1  0
2

EI n 1  j 0 
or …....….……. (25)
1  n 1 
C1n 1  C1    0.5 Pj ( L  x j )  Q j ( L  x j ) since Pn  Qn  0
2

EI n 1  j 0 

When applying Eq. 23 for all i yield:

C10  C11  C12        C1n 1  C1 and

1  i 
 0.5 Pj ( x  x j )  Q j ( x  x j )  C1
2

EI i  j 0 
y i ( x)  for xi  x  xi 1
2
 1  i  
1   0.5Pj ( x  x j )  Q j ( x  x j )  C1
2

 EI i  j 0  

…………………………….…………… (26)

assuming I i 1  I i at the joints (See Appendix B). Now impose the length of the beam segment

to be un-extendible, yields,

1   y i ( x)  dx
xi 1
Li  
2
or
xi

xi 1 dx
Li   ……………….....…….………. (27)
xi 2
 1  i  
1   0.5 Pj ( x  x j )  Q j ( x  x j )  C1
2

 EI i  j 0  

Eq. 27 does not lend itself to a simple solution (see appendix D for setting up Elliptic functions)

and numerically complex. To simplify the equation assume the increments are small enough such

that the slope throughout the interval of xi ≤ x ≤ xi+1 is the same (see Appendix A for the general

solution), so:

y i ( xi )  y i ( xi 1 ) ………………………………………………………...………………….. (28)

Thus,

10
Li  xi2  y i2  xi 1   y i ( xi )  ……………………………......……………………… (29)
2

Or

xi 1  xi
Li  …………………......………….. (30)
2
 1  i  
1   0.5Pj ( xi  x j )  Q j ( xi  x j )  C1
2

 EI i  j 0  

Thus, at i = 0 yields:

x1  x0 x1  x0
L0  
1  C1
2 2
 1 
1  
0.5P0 ( x0  x0 ) 2  Q0 ( x0  x 0 )  C1 ……………...………. (31)
 EI 0 

At i = 1 yields

x 2  x1
L1 
2
 1 
1  
0.5 P0 ( x0  x1 ) 2  Q0 ( x0  x1 )  1

0.5 P1 ( x1  x1 ) 2  Q1 ( x1  x1 )  C1 
 EI 1 EI 1 

x 2  x1
 ……………………......……………. (32)
2
 1 
1   
0.5P1 ( x 0  x1 ) 2  Q1 ( x0  x1 )  C1
 EI 1 

Substitute x1 – x0 from Eq. 31 in Eq. 32, for a given C1 and x2 – x1 is found.

At i = 2 yields

x3  x 2
L2 
2
 1 
1  
0.5 P0 ( x 2  x0 ) 2  Q0 ( x1  x0 ) 
1
 
0.5 P2 ( x 2  x1 ) 2  Q2 ( x 2  x1 )  C1 
 EI 2 EI 2 

………………......…………………………… (33)

11
Substitute x1 – x0 from Eq. 31 and x2 – x1 from Eq. 32 in Eq. 33, for a given C1 and x3 – x2 is

found, where x2 – x0 = (x2 – x1) + (x1 – x0).

Thus, if guessing C1 then find xi+1 – xi , can be found since the denominator of Eq. 30 is always

n
known from previous equations. And since  (x
i 0
i 1  xi )  x n  x 0  L  x0 , x0 can be found.

Therefore for a given C1 x0 , x1 , x2 , ……. , xn-1 can be solved, then proceed by checking the

end slope of Eq. 24 or Eq. 25. If it is not satisfied update C1 with numerical analysis until all the

variables are found. For the deflection from Eq. 4 yields:

1  i 
 0.5Pj ( x  x j )  Q j ( x  x j )  C1
2

EI i  j 0 
yi ( x)   dx  C 2 i for xi  x  xi 1
2
 1  i  
1   0.5Pj ( x  x j )  Q j ( x  x j )  C1
2

 EI i  j 0  

………………………....…………………. (34)

To find C2i assume compatibility and enforce:

y n ( x n )  y n ( L)  0 and find C 2 n 1
y n 1 ( x n 1 )  y n ( x n 1 ) and find C 2 n  2
etc.....

And the solution is found numerically.

Application for Case II - Numerical Solution for Any Load Function Non-Prismatic Beam:

This solution is very similar to the application of Case I or the previous application but instead of

using x and xi substitute for y and yi and can find C1, y0 , y1 , …… , yn-1 and C2i for the

deflection of Eq. 8 see FIG.2.

12
 L0 L1 Q i Li
P0 P1 P2 Pi Pi+1 Qi+1 x
Q0 Q1 Q2 y
Qi
Pi Qi+1 Pi+1
y0 y1 Q2
P2
Q1
P1
Q0

P0
Segment
x0 Beam i
Length = Li
x1
x2
xi
x
xi+1
xn = L

FIG.2 – Cantilever Beam Analysis – Non-Prismatic Beam

Point Loads are in the x direction

Application for Case III - Numerical Solution for Any Load Function Non-Prismatic Beam:

Again implement the solution using similar procedures as in the application of Case I, see FIG.3.

If assume at xi the resultant moment in the x and y direction is Mxi and Myi and if assume at xi the

resultant force in the x and y direction is Rxi and Ryi , then the moment for Eq. 9 becomes:

R xi R yi
h(ax  bx)  ( x  xi )  ( y  yi ) where
EI i EI i
………………………….…………. (34.1)
M M yi
xi  xi  xi and yi  yi 
R xi R yi

Thus, if translating temporarily the axis to a local axis with xi and y i Eq. 9 becomes:

13
 y R xi R yi
 h(ai x  bi y )  ai x  bi y , where ai   and bi   …………. (34.2)
 1  ( y )  2
3
EI i EI i

Note: for segment i-1 and segment i at x = xi the resultants are the same so

ai  ai 1 and bi  bi 1 ………………………………………………………….……. (34.3)

Thus if impose

y i 1 ( xi )  y i ( xi ) at x  xi
Seg. Bm. i  1 to Seg. Bm. i

Which means from Eq. 13 and 34.3 yields:

vi 1 (u i )  vi (u i ) at u  u i ……………………..…………………………. (35)

This implies C10  C11  C12        C1n 1  C1 , assuming I i 1  I i at the joints (Se Appendix

B). u0 , u1 , ……. , un can be found by using the approximation:

y i ( xi )  y i ( xi 1 ) or vi (u i )  vi (u i 1 ) ……..……………………......………………….. (36)

Thus,

2 2
xi2 y i2  a2  b2   a2  b2 
Li  xi2  y i2  u i   u  i i
(1  v
 )   i i
(1  v
 ) 
u i2 u i2
i
 2a i i
  2bi i

   

………………………..………………………… (37)

Or

2 2
ai2  bi2 1  vi (u i )  1  vi (u i ) 
Li  (u i 1  u i )     ……..…………………..………….. (38)
2  ai   bi 

14
Which is again for a given C1 ui‘s can be found from Eq. 38. When using Eq.10:

n 1
a n 1
 (u i 1  ui )  u n  u0  L  u 0 . Thus, all ui‘s can be obtained by back substitutions
i 0 a n21  bn21

with checking if C1 satisfies:

y n ( x n )  y n ( L)  0
or …………………….………………… (39)
  a n 1 
 1  v  L 
 n
 a2  b2 
a n 1  1  vi (u n )  a n 1   n 1 n 1   0
  
bn 1  1  vi (u n )  bn 1   a n 1 
 1  v  L 
 n
 a2  b2 
  n 1 n 1 

Now, find C2n-1 from the last segment to satisfy

 a n 1  a n 1
v n (u n )  v n  L  L …………………………...………………….. (40)
 a2  b2  a n21  bn21
 n 1 n 1 

and find all of C2i from the compatibility equation of deflection yi and using Eq. 11:

u n 1  v n 1 (u n 1 )  u n 1  v n (u n 1 ) and find C 2 n  2
u n 2  v n  2 (u n  2 )  u n  2  v n 1 (u n  2 ) and find C 2 n 3
etc.....

When done find all of v0 , v1 , ……… , vn-1 then substitute in Eq. 11 to find xi and yi .

15
 P1 Pi+1
L0 L1 P2 Q i Li
P0 Qi+1 x
Q0 Q1 Q2 Pi y
Qi Qi+1
y0 y1 Q2
P1 P2 Pi Pi+1
Q1
Q0
P0
Segment
x0 Beam i
Length = Li
x1
x2
xi
x
xi+1
xn = L

FIG.3 – Cantilever Beam Analysis – Non-Prismatic Beam

Point Loads are in any direction

Application Examples of Case III Numerical Solution:

(a) Fishing Pole: an application for Case III is shown on FIG. 4, where the beam has an angle

α with the horizontal and a load P0 and P1 hanging from the beam. One common

application is a fishing pole that has a fish that has the load P0 from the vertical and P1 =

0. In this case when varying the angle α with the horizontal the moment changes and the

deflection curve change giving a smaller or bigger moment with various elastic curves.

This affects the ability to pull the fishing with the real and having various controls on

catching the fish by puling or letting go the line. An experienced fisherman does this

16
 procedure naturally and gets the credit for not loosing the fish. A good fishing rod would

be designed to have a moderate elastic curve configuration when varying P0 and the angle

α. The equations for a non prismatic fishing pole can be:

Px  P1 cos  and Py  P1 sin  Thus the function hi in Eq.9 becomes :

P0
h0 (a 0 x, b0 y )  ( x  x0 ) cos   ( y  y 0 ) sin  
EI 0
or
Pj
( x  x ) cos   ( y  y 
i
hi (ai x, bi y )   j j ) sin   ai ( x  xi )  bi ( y  y i )
j 0 EI i

where,
i Pj x j i Pj y j
 EI i
 EI i i Pj i Pj
ai   bi  
j 0 j 0
xi  yi  cos  and sin 
i Pj i Pj EI i EI i
 EI
j 0
 EI
j 0
j 0 j 0

i i

In this situation each beam segment can be translated in a new local axis by xi and y i to

have Eq. 10 ready for rotation of axis to satisfy Eq. 9 and then translate to the global axis.

For example translate the local axis by xi and y i , and substitute in Eq. 17 and 18 the new

coordinates (u  u i ) and (v  vi ) for u and v in the solution of Eq. 17 and Eq. 18, where

u i and vi are obtained from substituting xi and y i in equation 10 with the replacement of

a by ai and b by bi .

(b) Curved Beam: Another application example where the beam is a non-prismatic curved

beam. Thus, if subdivide the curved beam to a smaller segments cantilever straight

17
 beams3 with constant moment of inertia as in using FIG. 3. If αi is the angle the

segmental beams make with the horizontal, then the problem can be solved by taking

each deflection curve derived from the local axis of the segmental beam and rotated by

the angle αi then translate by xi and y i to the global axis and the problem can be solved

numerically.

L0, I0
α
P0

L1, I1
x0 P0 sinα
α
P0 cosα P1
P0
α
P1
x1
x

x
y

xn = L

FIG.4 – Fishing Pole Example (a)

3
Note: this approximation does not mean there are stress singularities due to sharp corners at the discontinuities
where each segment meets. All it means is the actual deflection of the slightly curved segment can be approximated
with the deflection of a beam with straight segment with a constant moment of inertia.

18
(c) Bow and Arrow: The ancient structural problem in archeries or shooting bow and arrow

can finally be solved. Amazingly, can design a curved non-prismatic beam to give the

proper deflection curve for a human precise measurement giving the best comfortable

result for a more accurate bull’s-eye. Possibly, a unique design for each athlete, so

putting tension by changing the string size can be less desirable. To simplify the

equations and show the example, a non-prismatic curved beam will not be used but use a

non-prismatic straight beam, see FIG. 5. The function hi in Eq. 9 becomes:

y0

x y
x0
0.5P

L
β
L - x0

P
L  ( L  x0 ) 2
2

0.5P

FIG.5 – Bow and Arrow Example (c)


1 P  P 
h0 (a 0 x, b0 y )  ( x  x0 ) 2   ( y  y 0 ) cot   2 
EI 0
    
1  P ( L  x0 )  P 
 ( x  x0 )   ( y  y 0 )  
EI 0   2 L2
 ( L  x ) 2
 2 
 0

19
 1  P  P 
hi (ai x, bi y )  ( x  x 0 ) 2   ( y  y 0 ) cot   2 
EI i
    
1  P ( L  x0 )  P 
 ( x  x 0 )   ( y  y 0 )  
EI i   2 L2
 ( L  x ) 2
 2 
 0

 ai ( x  x0 )  bi ( y  y 0 )
where
P P ( L  x0 )
ai  and bi  
2 EI i 2 EI i L2  ( L  x 0 ) 2

In this situation each beam segment can be translated by x0 and y 0 in its local axis to

satisfy Eq. 9 and then to the global axis. For example translate by x0 and y0, and substitute

in Eq. 17 and 18 by the new coordinates (u-ui) and (v-vi) for u and v of Eq. 17 and Eq. 18,

where u i and vi are obtained from substituting x0 and y 0 in equation 10 with the

replacement of a by ai and b by bi .

Column with Load through Fixed Point: The problem of column with load through fixed

point was presented by Timoshenko and Gere [2]. Jong-Dar Yau [6] presented a solution

for Closed-Form Solution of Large Deflection for a Guyed Cantilever Column Pulled by

an Inclination Cable. A more general problem is to allow the fixed point D to have a

coordinate point (xd, yd) instead of the coordinate point (xd, 0) as in FIG. 6. As if the tip of

the column is attached by a cable with a shackle to point D and the shackle is being

tightened. The column is assumed a non-prismatic. This situation can also happen in a

vertical fishing pole, where the fish pulls with an angle β. Thus, moment becomes:

20
M  ( x  x0 ) Py  ( y  y 0 ) Px
 ( x  x0 ) P sin   ( y  y 0 ) P cos 

P  ( yd  y0 ) ( x d  x0 ) 
h(ai x, bi y )  ( x  x 0 )  ( y  y0 ) 
EI i  ( y d  y 0 ) 2  ( x d  x0 ) 2 ( y d  y 0 ) 2  ( x d  x0 ) 2 


 ai ( x  x0 )  bi ( y  y 0 )
where,
P ( yd  y0 ) P ( x d  x0 )
ai  and bi  
EI i ( y d  y 0 ) 2  ( x d  x0 ) 2 EI i ( y d  y 0 ) 2  ( x d  x0 ) 2

y
y0 P
x0

L,I β

xd

yd

D
x

FIG.6 - Column with Load through Fixed Point (d)

In this situation each column segment can be translated by x0 and y 0 in its local axis to

satisfy Eq. 9 and then to the global axis. For example translate by x0 and y0, and substitute

in Eq. 17 and 18 by the new coordinates (u-ui) and (v-vi) for u and v of Eq. 17 and Eq. 18,

21
 where u i and vi are obtained from substituting x0 and y 0 in equation 10 with the

replacement of a by ai and b by bi .

The ± sign in Eq. 3, 7 and 17:

The ± sign in the solution of Eq. 3, 7 and 17 can be used interchangeably when the slope of the

deflection curve goes to infinity at a point and the slope change sign. Another word the

deflection curve is not a function anymore and becomes circular. In Eq. 17 v   when

y  a / b . At that point the correct sign of Eq. 17 must be used.

Other End Condition: See Appendix E

Curvilinear Beams:

For a curvilinear beam with a function y = R(x) the new radius of curvature must satisfy the

following equation:

1
rnew  …………………………………………………………………. (41)
M ( x) 1

EI rold

So that if M(x) = 0 the radius of curvature does not change and remain of the function y = R(x).

Thus Eq. 1 becomes:

y M ( x) ( x)
R
   t ( x) ………………………………………… (42)
 1  ( y ) 
2
3
EI ( x) 
  2 
 1  R ( x) 
3

 

And solution is as Eq. 3 and Eq. 4 by replacing f(x) by t(x).

22
Extensibility:

In order to account for extensibility of the beam will analyze a beam segment Li. Let αi be the

directional angle of the load Pi and θi the angle at Pi representing the slope of the beam at that

point as in Fig. 7.

Li
θi

90 - αi
Pi
αi

180 – (90 - αi) – θi = 90 – (θi - αi)

Fig. 7 Extensibility

For extensibility of a small arc length ds in Li expressing the change in length (shortening) εi due

to the axial load in Li as:

xi 1 Pi sin(   i )
 i   ds ……………………………………………….……………… (43)
xi Ai E

Expressed as (ΣPL/AE) where Ai is the area of the segment Li and the load Pi is the resultant at

every ds. In a shortening condition, the shape of the deflection curve did not change only the

curve has shrunk. The resultant moment is affect by extensibility due to the change of

23
xi , y i or u i for a new xi , y i or u i . Then for a given deflection curve derived without including

extensibility that gives xi , y i or u i yields:

i i x j 1 Pj sin(   j )
xi  xi   ˆ j iˆ  xi    cos  ds
j 0 j 0
xj Aj E
j Pj sin  cos  j  cos  sin  j  dx
 xi   
x j 1
ds
j 0
xj Aj E ds
 dy dx 
j
Pj  cos  j  sin  j 
 xi   
x j 1  ds ds  dx
j 0
xj Aj E
 dy 
j
Pj  cos  j  sin  j 
 xi   
x j 1  dx  dx dx
j 0
xj Aj E ds
 dy 
j
Pj  cos  j  sin  j 
 dx  1
 xi   
x j 1
dx
Aj E 1  dy / dx 
xj 2
j 0

j
x j 1 Pj  y cos  j  sin  j  1
 xi    dx 
Aj E 1   y 
xj 2
j 0

i i Pj sin(   j )
y i  y i   ˆ j ˆj  y i   
x j 1
sin  ds
j 0 j 0
xj Aj E …………………. (44)
j
x j 1 Pj  y cos  j  sin  j  y
 yi    dx 
Aj E 1   y 
xj 2
j 0

Alternatively, if would like to include the effect extensibility on the moments then Eq. 27

becomes4:

4
Note the extensibility Eq. 45 is slightly conservative since  ds  L
i when integrate from xi to xi 1 .

24
 1   y i ( x) dx
xi 1
Li   i  
2
xi

or
Pi sin(   i )
1   y i ( x) dx  
xi  1 xi  1
Li  
2
ds
xi xi Ai E
Pi sin  cos  i  cos  sin  i 
1   y i ( x) dx   1   y i ( x) dx
xi 1 xi  1

2 2
xi xi Ai E
 dy dx 
Pi  cos  i  sin  i 

xi 1
1   y i ( x) dx  
2 xi  1  ds dy  1   y ( x)2 dx
i
xi xi Ai E
 y i ( x) 1 
Pi  cos  i  sin  i 
 
1   y i ( x) 1   y i ( x)
2 2


xi 1
1   y i ( x) dx  
2 xi  1   1   y ( x)2 dx
i
xi xi Ai E
xi 1 P  y  i ( x) cos  i  sin  i 
1   y i ( x) dx  
xi 1

2 i
dx
xi xi Ai E
 yi 1  yi Pi cos  i xi 1  xi Pi sin  i
1   y i ( x) dx 
xi 1

2

xi Ai E Ai E

………………………………….. (45)

This equation can be adjusted with the rotation of the axis for αi of Case III and implemented per

Eq. 13 and instead of updating xi in Eq. 27 update u i and Eq. 17 when substituting Eq. 17 in Eq.

13 to obtain y i (x) of Eq. 45 to find xi. This assumes the total deflection curve is shortened or

elongated by  x ,  y and the solution in Eq. 4, 8, 18 remains the same and only is effected by xi

and yi as in Eq. 19, where:

n 1
Pi sin(   i )
 x   
xi 1
cos  ds
i 0
xi Ai E
………………………………………………….. (46)
n 1 xi  1 Pi sin(   i )
 y    sin  ds
i 0
xi Ai E

25
Finale note on extensibility and large deflections: extensibility may have a minor effect on the

moments however it can affect the buckling deflection criterion. In general when loading a beam

the moment and axial load reduces with time however the deflection increases with time until the

final xi , y i or u i occurring at tfinal . In this case the safety factor on the stresses must account for

the dynamic problem of loading and reloading and care must be taken when using large

deflections in design.

Comparison with current methods for large deflections: It would be very difficult to draw

conclusions from one or two examples when comparing the exact solutions with any

approximate method including finite elements. Thus, comparison is left out to a more in-depth

study in a different article. The finding in this paper stands alone on its own two feet, is complete

and it is a bench mark.

Conclusion:

The closed form and general solution of non-linear differential equation of Bernoulli-Euler beam

theory is solved numerically for general loading function for a non-prismatic beam and can be

approximated for a non-prismatic curved beam when the presented solution of curvilinear beam

is not used. In some cases it is solved in closed form for prismatic and non-prismatic beam. In

general the Elastica, as called by Timoshenko and Gere [2], is solved.

26
References:

1. Bisshopp, K. E., and Drucker, D. C., “Large Deflections of Cantilever Beams,” Quarterly

of Applied Mathematics, Vol. 3, 1945, pp. 272-275

2. Timoshenko, S. P. and Gere, J. M., “Theory of Elastic Stability” 1961, McGraw-Hill

Book Company, New York, pp. 76-82 and pp. 55-57

3. Rohde, F. V., “Large Deflection of Cantilever Beam with a Uniformly Distributed Load,”

Quarterly of Applied Mathematics, Vol. 11, 1953, pp. 337-338

4. Lau, John H. “Large Deflection of Cantilever Beams,” Journal of Engineering Mechanics

Division, Proceedings of the American Society of Civil Engineers, Vol. 107, NO. EM1,

February, 1981. pp. 259-264

5. Scott, E. J. and Carver, D. R. “On the Nonlinear Differential Equation for Beam

Deflection,” ASME Applied Mechanics Division June, 1955, pp. 245-248.

6. Jong-Dar Yau "Closed-Form Solution of Large Deflection for a Guyed Cantilever

Column Pulled by an Inclination Cable" Journal of Marine and Technology, Vol. 18, No.

1, 2010, pp 120-136

7. Nellis, J. William "Tables of Elliptic Integrals" NASA Contractor Report NASA CR-289

Prepared under Grant No. NsG-293 by IOWA State University, Ames, Iowa for

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION . WASHINGTON,

D. C. AUGUST 1965

8. B. C. Carlson "Three Improvements in Reduction and Computation of Elliptic Integrals"

Journal of Research of National Institute of Standards and Technology, Volume 107,

Number 5, September-October 2002, pp 413-418

27
 APPENDIX A

General algorithm solution of solving for xi for application example Case I:

From Eq. 27 let the function:

dx
 i ( x)    c ……………..…..………. (47)
2
 1  i  
1   0.5 Pj ( x  x j )  Q j ( x  x j )  C1
2

 EI i  j 0  

 i
When setting  0 there is no solution and the function  i (x) is completely odd but
x

translated, increasing and crossing the x axis once. Thus there is only one root xi*. To proceed to

 2 i
find the inflection points for  0 rewrite Eq. 47 to be:
x 2

dx
 i ( x)    c …………………………………….........………..….. (48)

1  C i ( x  Ai ) 2  Bi 
2

Where:

 (P x
j 0
j j  Qj )
Ai  i

P
j 0
j

i
1
Bi   Ai2 C i 
EI i
 ( 0 .5 P x
j 0
j
2
j  Q j x j )  C1 ……………………………......……………….. (49)

i
1
Ci 
EI i
 0 .5 P
j 0
j

And from Eq. 25:

28
 1  n 1 
C1    0.5 Pj ( L  x j )  Q j ( L  x j ) ………………………......………………… (50)
2

EI n 1  j 0 

Thus the inflection points are only three and they are:

xi1  Ai
Bi …………………………………………………………………………… (51)
xi 2,3  Ai  
Ci

The function becomes:

 1  3  5    (2k  1)
  

dx
 c   1  
2k
 i ( x)   C i ( x  Ai ) 2  Bi  dx  c

1  C i ( x  Ai ) 2  Bi 
2
 k 1 2  4  6    2 k 

………….………………………………. (52)

The slopes and the deflections become:

C i ( x  Ai ) 2  Bi
y i ( x)  for xi  x  xi 1 …………………..………….. (53)

1  C i ( x  Ai ) 2  Bi  2

y ( x)  
C ( x  A )  B  dx
i i
2
i
 C 2i for xi  x  xi 1
1  C ( x  A )  B 
i 2
2
i i i

1   1  3  5    (2k  1)
 Ci ( x  Ai ) 3  Bi ( x  Ai )    Ci ( x  Ai ) 2  Bi  2 k 1 
 dx  C 2i
3  k 1 2  4  6    2k 


C ( z  A )  B  dz
x
i i
2
i
 C 2i
1  C ( z  A )  B 
Ai 2 2
i i i

…………………………………………… (54)

And find

L
C 2 n 1    y n ( x) dx
An 1
xi 1 xi 1
…………..……………………………….……… (55)
C 2 i 1   y i ( x) dx   y i 1 ( x) dx  C 2 i
Ai Ai 1

And let

29
  1  3  5    (2k  1)
  

xi 1 dx xi  1
1   2  4  6    2k C i ( x  Ai )  Bi
2k
 i ( xi )    Li   2
 dx  Li
xi

1  C i ( x  Ai ) 2  Bi 
2 xi
 k 1 

…………………………………………. (56)

For the numerical procedure, use Newton-Raphson method for systems of nonlinear algebraic

equations. The following is the result for the Jacobian matrix. For a given function  i ( xi ) from

Eq. 56 the derivative with respect to xm for m = 1 to n is:

x  xi 1

 i ( xi )  x B  1 Bmi
  Ami  ( x  Ai ) mi   ( xi 1  xi )
x m  x m 2 Bi  1  C ( x  A ) 2  B
i i i  
2
x  xi
2 Bi

 C ( x  A ) 

1  3  5    (2k  1)  1  Bmi

xi 1 2k
2 k  2
 Bi dx
2  4  6    2k  2  Bi
i i
xi
k 1

x  xi 1

 x B  1 B 
xi  1 1  C i ( x  Ai ) 2  Bi  2

  Ami  ( x  Ai ) mi   mi  dx
 x m   1  C ( x  A ) B  
3
2 Bi  1  C ( x  A ) 2  B 2 2 Bi xi
2 2 2
i i i x  xi i i i

………………………………………………………. (57)

Where:

Pm
Ami  i
for m  i and
P
j 0
j

Ami  0 m  i
……….………….. (58)
Pm x m  Qm 1
Bmi  2 Ai Ami C i   Pm ( L  xm )  Qm  for m  i and
EI i EI n
1
Bmi  Pm ( L  xm )  Qm  mi
EI n

30
 i
Thus, choose xi   Li for the initial condition and the Newton-Raphson method requires the
j 0

updated x  x  J    . Where, x  is the updated vector of xi , xis the old vector of xi ,
1

J  is the Jacobian matrix evaluated xi and   is the vector of function of Eq. 56 evaluated xi .

Example:

In some cases this solution is the exact solution when the loads are actually point loads and

moment for a beam. Setting up the solution for two point loads and two moments on a beam that

is of two moment of inertia, see Fig. 8, and substituting yield,

C1  
1
EI 1

0.5 P0 ( L  x0 ) 2  0.5 P1 ( L  x1 ) 2  Q0 ( L  x0 )  Q1 ( L  x1 ) 
Q0
A0  x0 
P0
0.5 P0
C0 
EI 0

B0   A02 C 0 
1
EI 0
 
0.5 P0 x02  Q0 x 0  C1

A00  1 A10  0
1 1
B00  P0 ( L  x0 )  Q0  B10  P1 ( L  x1 )  Q1 
EI 1 EI 1

x0 P0  x1 P1  (Q0  Q1 )
A1 
P0  P1
0.5( P0  P1 )
C1 
EI 0

B1   A12 C1 
1
EI 1
 
0.5 P0 x 02  0.5 P1 x12  (Q0 x0  Q1 x1 )  C1

P0 P1
A01  A11 
P0  P1 P0  P1
P0 P1
B01  L  A1  B10  L  A1 
EI 1 EI 1

31
Let H consider the variables in Eq. 57

 0
 H m 0 ( x0 , x1 , A0 , B0 , C 0 , Am 0 , Bm 0 )
x m
 1
 H m1 ( x0 , x1 , A1 , B1 , C1 , Am1 , Bm1 )
x m

 H ( x , x , A , B , C , A , B ) H 10 ( x0 , x1 , A0 , B0 , C 0 , A10 , B10 )
J   00 0 1 0 0 0 00 00 
 H 11 ( x0 , x1 , A1 , B1 , C1 , A01 , B01 ) H m1 ( x0 , x1 , A1 , B1 , C1 , A11 , B11 ) 

And an exact numerical solution is obtained.

P1
P0
L0, I0 Q1 L 1, I 1
x
Q0 Q1

P1
Q0

P0
x0

x1

xn = L

FIG.8 - Example

If for example the beam has to be divided to small increments due to the load function or the

moment of inertia function, a less computational analysis may be selected as in Eq. 28.

32
 APPENDIX B

Finding coefficients C1i with no discontinuity in the moment of inertia:

In Application for Case I, II, and III – (Numerical Solution for Any Load Function Non-

Prismatic Beam including the examples), the moment of inertia at the joints were imposed equal.

In the following equations this assumption will be shown valid and in Appendix C the derivation

for discontinues beam for abrupt changes of the moment of inertia will be derived.

To start with the closed form solution of Eq.1 through Eq 18 will be used and the moment of

inertia is taken as:

1 r

  an x n for Case I defined in the interval 0  x  L 
EI ( x) n 0 

r 
1   n  1, 2,........, r and r  
  an y n for Case II defined in the interval 0  y  y 0
EI ( y ) n 0 


1 r 
  an u n for Case III defined in the interval 0  u  u 0 
EI (u ) n 0 
………………………………. (59)

The following proof is for Case I. All other cases can be done with a similar proof.

From Eq. 3 the integral term for each segment becomes

 r  i 
Z i ( x)  C1i   f ( x)dx  C1i     a n x n   Pj ( x  x j )  Q j  dx  C1i for xi  x  xi 1
 n 0  j 0 

……………………………...……………..….. (60)

33
Using integration by parts starting with

r i
u   an x n and dv   Pj ( x  x j )  Q j yeilds,
n 0 j 0

 r n 
 i Pj ( x  x j ) 2   r n 1 
 i Pj ( x  x j ) 2 
Z i ( x)    a n x    Q j ( x  x j )     na n x  
  Q j ( x  x j )  dx
 n 0   j 0 2 
  n 1  j 0 2 


……………………………………………………. (61)

Continuing the integration by parts on each integral leads to

r r n!  i Pj ( x  x j ) k  2 Q j ( x  x j ) k 1  
nk  
Z i ( x)      an x     for x i  x  x i 1
k  0  n  k ( n  k )!   j 0 ( k  2 )! ( k  1)! 
  

……..………………………… (62)

At xi Eq. 3 for to consecutive segments at the joint becomes:

Z i ( xi )  C1i
y i ( xi )  and
1  Z i ( xi )  C1i 
2

………………………………………………………….. (63)
Z i 1 ( xi )  C1i 1
y i 1 ( xi ) 
1  Z i 1 ( xi )  C1i 1 
2

Apply compatibilities yields:

y i 1 ( xi )  y i ( xi ) at x  xi
………………………………………………… (64)
Seg. Bm. i  1 to Seg. Bm. i

Thus from Eq. 63

Z i 1 ( xi )  C1i 1  Z i ( xi )  C1i ……………………………………………………………… (65)

Substituting in Eq. 62 yields;

r  r  i 1 Pj ( x i  x j ) k  2 Q j ( x i  x j ) k 1  
 n! nk 
Z i 1 ( x i )  Z i ( x i )     a n x i       ……..…... (66)
k  0  n  k ( n  k )! 
  j 0 ( k  2 )! ( k  1)! 
 

34
Or,

C1i 1  C1i ……………………….………………………………………………………. (67)

Thus:

C10  C11  C12        C1n 1  C1 ….…………………………………………………… (68)

Thus, the coefficients C1i with no discontinuity in the moment of inertia is correct and the

moment of inertia of the segment for Eq. 60 can be approximated as:

l i 1

EI i 1 

li
I ( x) i i 1
where li   L j and li 1   L j ….…….……………………………… (69)
Li 1 j 0 j 0

35
 APPENDIX C

Finding coefficients C1i with discontinuity in the moment of inertia:

Application for Case I, II, and III are all similar and only Case I will be addressed. By using Eq.

64 and E. 65 in Eq. 22 the Zi can be written at xi as:

Z i 1 
1 i 1

 0.5Pj ( xi  x j ) 2  Q j ( xi  x j )
EI i 1 j 0

and …………………………………………… (70)

 0.5P ( x 
i 1
1
Zi  j i  x j ) 2  Q j ( xi  x j )
EI i j 0

Thus from Eq. 65 yeilds;

 1 1  i 1
C1i 1    
  0.5 Pj ( xi  x j )  Q j ( xi  x j )  C1i
2

 EI i EI i 1  j 0
or …………………………….. (71)
 1 1  i 1
C1i    
  0.5 Pj ( xi  x j )  Q j ( xi  x j )  C1i 1
2

 EI i 1 EI i  j 0

Starting with Eq. 25 yields;

1  n 1 
C1n 1    0.5 Pj ( L  x j )  Q j ( L  x j ) . ………………………………………… (72)
2

EI n 1  j 0 

Thus, all of the C1i can be found consecutively from Eq. 71 and Eq. 30 becomes:

xi 1  xi
Li  …………………………………. (73)
2
 1  i  
1      
2
0.5 Pj ( x i x j ) Q j ( x i x )
j  C1i
 EI i  j 0  

So Eq. 31 becomes:

x1  x0
L0  …………………………………………………………………………… (74).
1  C10 
2

36
So the approximate numerical method with Eq. 28 starts by guessing C10 then find (x1 – x0) from

Eq. 74 and Eq. 71 with i = 1yields;

 1 1 
C11     
 0.5 P0 ( x1  x0 )  Q0 ( x1  x0 )  C10 …………………………………….. (75)
2

 EI 0 EI 1 

So C11 can be found. By using Eq. 73 for i = 1yields;

x 2  x1
L1  …………………………………….. (76)
2
 1 
1   
0.5 P0 ( x1  x0 ) 2  Q0 ( x1  x0 )  C11 
 EI 1 

Now find (x2 – x1) using Eq. 74, Eq. 75 and Eq. 76. , then find (x2 – x0) = (x2 – x1) + (x1 – x0) and

substitute in Eq. 71 yields;

 1 1 
C12  
EI

EI
 
 0.5 P1 ( x 2  x1 )  Q1 ( x 2  x1 )  0.5 P0 ( x 2  x 0 )  Q0 ( x 2  x0 )  C11
2 2

 1 0 

…………………………………………………………………………… (77).

So all of the C1i then find (xi+1 – xi) can be found using Eq. 71 and Eq. 73. Now find (xn – x0)

from

( x n  x0 )  ( x n  x n 1 )  ( x n 1  x n  2 )  ...........  ( x1  x0 ) …………………………………. (78)

If (xn – x0) = (L – x0) then x0 = L - (xn – x0), and all of x1 , x2 , ……. , xn-1 can be found. By

checking the end condition of Eq. 25 or Eq. 72 with Eq. 71

37
 1  n 1 
y n ( L)  0   0.5 Pj ( L  x j )  Q j ( L  x j )  C1n 1  0
2

EI n 1  j 0 
or
1  n 1 
C1n 1    0.5 Pj ( L  x j )  Q j ( L  x j ) since Pn  Qn  0
2

EI n 1  j 0 
………………… (79)
check if

 1 1  n 1
C1n 1  
EI

EI
 2

  0.5Pj ( x n 1  x j )  Q j ( x n 1  x j )  C1n  2
 n2 n 1  j  0

If it is not satisfied update C10 with numerical analysis until all the variables are found. For the

deflection from Eq. 4 yields:

1  i 
 0.5 Pj ( x  x j )  Q j ( x  x j )  C1i
2

EI i  j 0 
y i ( x)   dx  C 2 i for xi  x  xi 1
2
 1  i  
1   0.5 Pj ( x  x j )  Q j ( x  x j )  C1i 
2

 EI i  j 0  

………………………....…………………. (80)

To find C2i assume compatibility and enforce:

y n ( x n )  y n ( L)  0 and find C 2 n 1
y n 1 ( x n 1 )  y n ( x n 1 ) and find C 2 n  2
etc.....

And the solution is found numerically.

For using Newton-Raphson method in Appendix A, replace Bi in Eq. 43 by

38
 i

 (P x
j 0
j j  Qj )
Ai  i

P
j 0
j

i
1
Bi   Ai2 C i 
EI i
 ( 0 .5 P x
j 0
j
2
j  Q j x j )  C1i ………………………..……………………(81)

i
1
Ci 
EI i
 0 .5 P
j 0
j

Where C1i is from Eq. 71 and from Eq. 79

1  n 1 
C1n 1    0.5 Pj ( L  x j )  Q j ( L  x j ) since Pn  Qn  0 …………….….. (82)
2

EI n 1  j 0 

So all of the C1i can be calculated from the vector {x} and update x   x  J   .
1

Where, x  is the updated vector of xi , xis the old vector of xi , J  is the Jacobian matrix

evaluated xi and   is the vector of function of Eq. 50 evaluated xi .

39
 APPENDIX D

Converting the Equations for Case I application to Elliptic functions:

Eq. 26 and Eq. 27 in Case I applications can be re-written as in Eq. 48 yields:

xi 1 dx
Li  
xi

1  C i ( x  Ai ) 2  Bi 
2

……………………………..…………………………… (83)
C i ( x  Ai ) 2  Bi
y i 

1  C i ( x  Ai ) 2  Bi 
2

Where Ai , C i and Bi are define in Eq. 49

Re-writing Eq. 77 to become:

xi 1 dx
Li   …………………………………………………. . (84)
xi

1  C i ( x  Ai )  (1  Bi )  1
2

2

Now let

C i ( x  Ai )  1  Bi cos  with C i dx   1  Bi sin 

and ………………………… (85)
C i ( x  Ai ) 2  (1  Bi ) cos 2 

1  Bi  2 i 1 sin 
Li    d ………………………………………….. (86)
Ci 1i

1   (1  Bi ) sin   1 2

2

Now multiply the square in the denominator in Eq. 86 and rearrange yield:

1  2 i 1 d
Li  
2C i
 1i 2
………………………………………………… (87)
 1  Bi 
1   sin 2 
 2 
 

Where:

40
  Ci   C /2 
1i  cos 1  xi  Ai   cos 1  i xi  Ai 
 1  Bi   pi 
and ……………………………… (88)
 Ci   C /2 
 2 i 1  cos 1  xi 1  Ai   cos 1  i xi 1  Ai 
 1  Bi   pi 

Where

1  Bi
pi  ……………………………………..………………………………………….. (89)
2

And Eq. 87 becomes:

1  2 i 1 d
Li  
2C i
 1i
1  p i2 sin 2 
………………………………………………………….. (90)

Which can be expressed in Elliptic Integral as follows:

1
Li  F ( pi ,  2 i 1 )  F ( pi , 1i ) ……………………………………...………………. (91)
2C i

Where the function F is the elliptic integral of the first kind.

Similarly Eq. 83 becomes:

C i ( x  Ai ) 2  (1  Bi )  1
y i ( x)  ………………………………….………………… (92)

1  C i ( x  Ai )  (1  Bi )  1
2

2

Now substitute Eq. 89 in 92 yields:

 2 pi2 sin 2   1
y i ( x) 
2 pi sin  1  pi2 sin 2 
where …………………………………………………………… (93)
pi
x cos   Ai
Ci / 2

The deflection between the joints becomes:

41
 y i 1 xi  1 C i ( x  Ai ) 2  (1  Bi )  1
 y dx  y i 1  y i   dx ………………………………. (94)
yi xi

1  C i ( x  Ai ) 2  (1  Bi )  1  2

Eq. 94 can be re-written and using Eq.84 yields:

xi 1 C i ( x  Ai ) 2  (1  Bi ) xi 1 1
y i 1  y i   dx   dx
xi

1  C i ( x  Ai ) 2  (1  Bi )  1 2 xi
 
1  C i ( x  Ai ) 2  (1  Bi )  1
2

xi 1 C i ( x  Ai ) 2  (1  Bi )
 dx  Li
xi

1  C i ( x  Ai )  (1  Bi )  1 2
2

…………………………..……………. (95)

Now substitute Eq. 89 in 95 yields:

1  Bi  2 i 1 p i2 sin 2 
y i 1  y i  Li 
Ci
 1i
p i 1  p i2 sin 2 
d

1  2 i 1 pi2 sin 2 
 Li 
Ci / 2
 1i
1  p i2 sin 2 
d
……… (96)
  2 i 1
2 1  2 i 1 
 Li   d   1  p i2 sin 2  d 
2C i   1i 1  pi2 sin 2   1i

2  2 i 1 2
  Li 
2C i
 1i
1  pi2 sin 2  d   Li 
Ci
E ( pi ,  2 i 1 )  E ( pi , 1i )

Where the function E is the elliptic integral of the second kind.

42
 APPENDIX E

Other End Conditions:

End Condition #1: This condition has a pin at bottom end and rotational fixed at the top end but

free to translate at the top as in Fig. 9.

y0

y
x0
Q0

RiAi
Ri

Fig. 9 Pin – Rotational Fixed Column

dy
Thus the moment Q0 at the tip of the column makes 0 Now at y = yn = 0 and x = L
dx @ y  x 0

fn(L) = 0 since the moment at the bottom is zero or Eq. 21 becomes:

43
 M n 1 1  n 1 
f n ( L)    Pj ( L  x j )  Q j   0
EI n 1 EI n 1  j 0 
or
n 1
Q0  P0 ( L  x0 )   Pj ( L  x j )  Q j  0 ……………….……..……….. (97)
j 1

or
n 1
Q0   P0 ( L  x0 )   Pj ( L  x j )  Q j 
j 1

Let

i
Ri   Pj …………………………………………………….…………………….. (98)
j 0

Then from Eq. 49

1  
 
n 1
Q0
A0  
R0 R0 
  P ( L  x 0 )   P ( L  x j )  Q j  …………………………….…….. (99)
j 1 

dy
Thus 0  from Eq. 83 C 0 ( A0 ) 2   B0
dx @ y  x 0

And Eq. 24 does not apply. So rewriting Eq. 23 using Eq. 83 to

C i 1 ( xi  Ai 1 ) 2  Bi 1  C i ( xi  Ai ) 2  Bi
or …………………………………………… (100)
Bi  C i ( xi  Ai ) 2  C i 1 ( xi  Ai 1 ) 2  Bi 1

Therefore by substituting Ai , Ai-1 , Bi-1 in Eq. 99 then Bi is found and  2 i 1 , 1i and pi are

found from Eq. 88 and Eq. 89 and the problem is solved using the elliptical integral.

Note: Eq. 56 and 57 can be written as:

1  2i 1 d
 i ( xi )  
2C i
 1i
1  pi2 sin 2 
 Li …………………………………………….. (101)

44
 x  xi 1

 i ( xi )  x B  1 Bmi
  Ami  ( x  Ai ) mi   ( xi 1  xi )
x m  x m i 
2 Bi  1  C ( x  A ) 2  B
i i 
2
x  xi
2 Bi

 C ( x  A ) 

1  3  5    (2k  1)  1 B

xi 1 2k
 2k   mi i i
2
 Bi dx
k 1 2  4  6    2k  2  Bi xi

x  xi 1

 x B  1 Bmi  2 i 1 
1  cos 2   Bi sin 2 
2

  Ami  ( x  Ai ) mi    d
 x m    
3
2 Bi  1  C ( x  A ) 2  B 2
4 Bi p i2 2C i 1i
i i i x  xi
sin  1  p sin
2 2
i
2 2

………………………………………………………… (102)

End Condition #2 Same as condition #1 except fixed at x = L as in Fig 10.

dy
Using Eq. 93 at i = n, xn = L  0 yields:
dx @ xL

Ci 1 ( L  Ai 1 ) 2  Bi 1  0 ……………………….…………………….……………… (103)

For fn(L) = 0 since the total moment at the bottom is zero implies

M n 1 1  n 1 
f n ( L)    Pj ( L  x j )  Q j  Qn   0
EI n 1 EI n 1  j 0 
or
n 1
Q0  P0 ( L  x 0 )   Pj ( L  x j )  Q j  Qn  0 ………………………….…………….(104)
j 1

or
n 1
Q0   P0 ( L  x0 )   Pj ( L  x j )  Q j   Qn
j 1

45
 y0

y
x0
Q0

RiAi
Ri

Qn
x

Fig. 10 Fixed – Rotational Fixed Column

Thus the procedure is to pick Qn and find Q0 from Eq. 97 and use procedure in condition #1

above to solve for yi for given Li and find all Bi then 1i 1 ,  2i and pi are from Eq. 82 and

Eq. 83 and check if Eq. 96 is satisfied if not update Qn and the problem is solved using the

elliptical integral.

End Condition #3 Pined both ends as in Fig 11.

Subdividing the column into two parts Part A and Part B at point E at line a – a where the slope

dy
 0 in which it is to be found. Separate the loads of Part A and Part B and solve for y0A for
dx

Part A using a straight cantilever column fixed at point E. Then solve Part B using end condition

#1 (pin at bottom end and rotational fixed at the top end but free to translate at the top.) and find

y0B . Check and see if y0A = y0B . If y0A not equal to y0B then move point E at line a – a up if y0A >

y0B or down if y0A < y0B until point E is found.

46
 y0A

RiAiA PART A

RiA

y0B
a a
E

RiAiB PART B
RiB

Fig 11 – Pined both ends.

End Condition #4 Fixed both ends.

This condition is similar to end condition #3 where Part A and B are as condition #2.

47