Influence Line Analysis of Bridges Using MATLAB

Saleh I. Aldeghaither
*
* College of Engineering, King Saud University,
P.O. Box 800, Riyadh 11421, Saudi Arabia



ABSTRACT
This paper presents a computer procedure for computing influence lines for
plane frame bridge models. The procedure is based on the stiffness matrix
equation in conjunction with fixed end actions representing the effect of moving
load. The procedure does not need subdivision of the structure members or
specification of load position prior to stiffness matrix analysis. Specification of
the load and effect positions comes after the stiffness matrix analysis. The
procedure was implemented into a computer program using MATLAB and can
easily be implemented into any standard computer program for matrix structural
analysis.


..==\' ,= .'=.=\' .'. ==\\ ,.'.\' === ,'==\ -\'= .,.= -,= ,., .=,\' '.= .=.=.
-..'>= .',.'., ==.=\' ==\' ,.=. _= .'==\', _.'=.,' ,\=.\' ,'.=.=' _\= -,=\' .=
.'==.' '=' ..= ,=.\ -='=\' ,.= _ -,=\' .= -,\'= _=>. ' .'=' _\ '=.=\' .'==' ,
.'==\', ,\=.\' , -,\==\' ,.'.\' === ' \' ==. .== .,.=. ,.'..\' === _,==
-.==\' -,>,.'.=.' _=\', .'==\', _='=.' '=.=\' ,\=. .=, ',,\= ==\' ,., -,\==\'
,,=> g='., _' _ ',,,=. _>=, -\,= -,=\' -.==, ,. .=,\' '.= _ .'==\', ,\=.\\ .
g='., ,'.=.=', -,=\' .MATLAB

INTRODUCTION
Influence lines have important applications for the design of structures that are
subjected to live or moving loads, e.g. bridges. Influence lines show the
influence of positioning of a unit load on a function [1,2]. The function (or load
effect) could be anything that varies as the load traverse the structure, such as
moment or shear at a point in a structure, axial force in a member or support
reaction. Once the influence line for an effect at a point in a structure has been
established, the maximum effect caused by the live load can be determined.
Although determination of influence lines for determinate structures is
straightforward, it is rather involved for indeterminate structures. Obviously,
determination of influence lines by positioning unit loads at several locations on
the structure is impractical. Belegundu [3] presented a procedure for computing
influence lines for determinate and indeterminate structures based on what he
called the adjoint variable method. The method, however, requires subdivision
of the members in order to calculate the influence lines for functions within
members. Furthermore, the effects for which influence lines are desired need to
be specified prior to finite element analysis. The Muller-Breslau principle, on
the other hand, provides a more effective way of obtaining qualitative influence
lines. For quantitative evaluation of influence lines, the Muller-Breslau
procedure is somewhat involved. Cifuentes and Paz [4], however, utilized the
Muller-Breslau principle in conjunction with finite elements to determine the
influence lines. They used the NASTRAN commercial code to demonstrate the
procedure. The procedure, however, requires subdivision of the structure
members in order to calculate the influence lines at points within members.
Moreover, for each influence line, a new input data and hence a new round of
finite element analysis is needed. Bridge oriented software like QconBridge [5]
are based on subdivision of bridge spans into segments, normally ten segments
per span, prior to finite element analysis. This results in a higher degree of
freedom model and hence expensive computation cost. The method presented
in this paper is based on the stiffness method and has the advantage of not
requiring subdivision of members nor the specification of the desired load
effect, its location or the load movement increment prior to the stiffness matrix
analysis. After a standard stiffness matrix analysis has been completed, the
method relies on simple statics to calculate the influence lines for any effect at
any section in the structure using any desired load movement increment.

COMPUTATIONAL PROCEDURE
Consider the frame shown in Fig. 1. The stiffness matrix equation for the frame
can be written as [6]
F
F F K = ] [ (1)

where [K] is n n stiffness matrix of the structure, is an (n 1) vector of nodal
displacements and F and F
F
are, respectively, the vectors of nodal and fixed-end
forces associated with member loadings. Since there are no applied nodal
forces, the vector F contains zeros for all degrees of freedom except those
associated with support reactions.











(a)










(b)
Figure 1: (a) Plane frame with a unit load on member i. (b) Fixed-end forces of
member i.

Fixed-End Forces: The Fixed-end forces due to a unit transverse load
positioned at distance xl
i
(0 1 x ) from the left node (I-node) of span i can be
written as

f
i
F
I
F
I
F
I
F
J
F
J
F
J
F
i
i i i
i i
x
N
V
M
N
V
M
l l l
l l
x
x
x
( )

(
(
(
(
(
(

0
1
0
0
0
3
0
2
0 2
0
0
0
0
0
3
0
2
0 0
1
2
3

(2a)
or symbolically

f [h] g
i
F
i
x x ( ) ( ) =

(2b)
xl
i

Unit load
l
i






V
J
F

V
I
F

I
J
xl
i

l
i

Unit load

F
I
M

F
J
M

where
{ } g( ) x x x x
T
= 1
2 3
(3)

Thus, the vector function of fixed-end forces can be regarded as a nonlinear
combination of four constant vectors (the vectors of [h]) that are independent
of load position, x. The combination coefficients are the terms of g(x).

The stiffness matrix equation for member i can then be expressed as


) ( ) ( ) ( ] [ x x x
i i
g [h] f k = (4)

where (x) and f(x) are, respectively, vector functions of the displacements and
forces at the member ends.
Equation (4) can be transformed to global coordinates to yield

) ( ) ( ) ( ] [ x x x
i i
g [H] F K = (5)

where

i
T
i
T
T
i
T
i
x x
x x
[h] [C] H
f [C] F
[C]
[C] [k] [C] K
=
=
=
] [
, ) ( = ) (
, ) ( ) (
, ] [

(6)

and [C] is the displacement transformation matrix for a plane frame member [6].
Equations similar to eq. (5) can be obtained for all spans that would be
subjected to the moving load. For other members the fixed-end force terms in
eq.(5) vanish.

Now let n be the total degrees of freedom of the structure and m the number of
spans that would be loaded, then the stiffness matrix equation of the structure
can be assembled from the member stiffness equations (eq. 6) to yield the
following

[ ]
) 1 4 (
2
1
) 4 ( 2 1
) (
) (
) (
) ( ) ( ] [

=
m
m
m n m
x
x
x
x x
g
g
g
] H [ ] H [ ] H [ F K
M
L
(7)

where
i
] [H is an n 4 matrix whose rows contain zeros and the rows of the six-
row matrix
i
] [H , arranged according to the global coordinate numbers of
member i and


g
g
0
i
x
x i
( )
( )
=

when the unit load is on member ,

(
(

)
`

=
)
`

) (
) (
) (
) (
) (
2
1
2 1
x
x
x
x
x
m
m
p
f
p
f
p
f
p p
f
pp pf
fp ff
g
g
g
H
H
H
H
H
H
F
0
M
L


(8)


where
f p
x ( ) and are the vectors of unknown and prescribed nodal
displacements, respectively. F
p
x ( ) is the vector of unknown forces (support
reactions). From eq. (8), the unknown displacements are given by

[ ] [ ] [ ] [ ]
[ ]
[ [ [ [ [ [ [ [
f ff f f f
m
m
ff fp p
x
x
x
x
( )
( )
( )
( )
=

1
1 2
1
2
1
H H H
g
g
g
] ] L
M
(9a)
or

=
(x) g
(x) g
(x) g
m
2
1
M
f f
x ] [ ) (
(9b)

where

[ ] [ ]
m f 2 f 1 f
1
ff
] H [ ] H [ ] H [ ] [K L L

=
m f f f f
] [ ] [ ] [ ] [
2 1
(10a)
and
[ [ [ [ [ [ [ [ =

ff fp p
] ]
1
(10b)

Similarly, the support reactions can be written as


F [F
g
g
g
F
p
m
x
x
x
x
( )
( )
( )
( )
=

+ ] ]] ]
1
2
M
(11)

where
[ ]
m p p p f pf
] [ ] [ ] [ ] ][ [ ] [
2 1
H H H K F L + =
(12a)
and
F ] ] = [ [ [ [ [ [ [ [
pp p pf
(12b)


It should be noticed that and F would vanish if the prescribed displacements

p
are zeros.
Equations (10) and (12) give solutions for nodal displacements and support
reactions, which are independent of load position within loaded spans.
Expressions for the nodal displacements and support reactions as functions of x,
the load position, can be obtained from eqs. (9) and (11), respectively. For
example, as the unit load traverses member j, the structure nodal displacements
are given by


[ ] [ ]
[ [ [ [ [ [ [ [
f ff f
j
j ff fp p
x x ( ) ( ) =

1
1
H g ] ]
(13)

where x changes from 0 to 1 as the load moves from I-node to J-node of
member j. The change in the values of x can be selected at any rate to reflect
the desired load movement increment. Support reactions can be expressed in
similar manner using eq. (11).

Local response
The member nodal displacements in local coordinates, (x), can be obtained
from the structure nodal displacements (eq. 9) by transformation. Equation (4)
can then be used to give the nodal forces of member j as functions of the load
position x

f [k f
j
I
I
I
J
J
J
j
j j
F
x
N x
V x
M x
N x
V x
M x
x x ( )
( )
( )
( )
( )
( )
( )
] ( ) ( )

= +

(14)
where the fixed-end force terms f
j
F
x ( ) vanish when the load is not on member j.
The member axial force is simply given by

N x N x N x
j J I
( ) ( ) ( ) =
(15a)

and the shear and moment at a distance sl
j
from the member I-node (0 1 s )
are given by (see Fig. 2)

V s x V x H s x
M s x M x s V x H s x
V s x V x
M s x M x sV x
I
I I
I
I I
( , ) ( ) ( )
( , ) ( ) ( ( ) ( ))
( , ) ( )
( , ) ( ) ( )
=
= +
=
= +

when the unit load is on the member
and


otherwise
(15b)

where H s x ( ) is the step function:

H s x
s x
( ) =
>

1
0
when
otherwise









Figure 2: Shear and moment at distance sl from I node in member j
The above procedure was effortlessly implemented into a computer program
using MATLAB. The steps involved for determining the influence lines are
summarized as follows:

) , ( x s V
I
V
I
xl
i

sl
i

1

I
M

) , ( x s M

step 1: Construct the structure stiffness matrix [K]
step 2: Compute the matrix [H]
j
(eq. 6) and assemble into the load
matrix m j
j
, , 1 = , L ] H [
step 3: Partition the system equation as shown in eq. (8) and solve for
[ ] [F] and (eqs. 10 and 12)
*
.
step 4: Obtain reactions influence coefficients (ordinates of influence lines),
if required, by substituting values for x in eq. (11).
step 5: Compute the influence coefficients of a member axial force, shear or
moment at any location within a member by substituting values for s
and x into eqs. (14-15).

Note that [F] and ] ]] ] [ [[ [ are the results of a standard static stiffness matrix
analysis. The desired influence lines are subsequently computed from these
results using member stiffness equations and simple statics. It may be worth
mentioning that in the computer program, the abscissas of the influence lines
can be arranged to reflect the actual positions of the load with respect to a
reference point. The reference point could be the point at which the load would
start traversing the structure.

COMPARISON WITH OTHER METHODS
The determination of influence lines can be based on positioning unit loads at
several locations on the structure, the adjoint method [2] or the Muller-Breslau
principle [3]. All of the aforementioned methods as well as the present method
are implemented using the standard finite element formulation, namely the
displacement method. The effectiveness of a method against the others is a
function of the CPU time, preparation of data and the flexibility of the method.
The CPU time that a method needs to obtain a set of influence lines for a
structure depends mainly on the required number of rounds of the FE analysis
and the degrees of freedom of the structure. Obviously the degrees of freedom
of the structure would increase if the method requires member subdivisions. The
effort needed for preparation of data and the robust and flexibility of the method
are also of great importance to the user. Therefore, the aforementioned criteria
can be used to assess the effectiveness of the present method.
Table 1 shows a comparison of the present method against the Adjoint and
Muller-Breslau methods. It is evident from the Table that the present method is
superior in all comparison criteria which proves the efficiency of the method.


*
Note that inversion of the stiffness matrix as implied by eq. (10) is not required. Instead, direct reduction of the
system of equations using standard elimination procedure is more efficient.
NUMERICAL EXAMPLES
Two example problems are considered to demonstrate the efficiency of the
present procedure. The first is a four-span continuous beam and the second is a
rigid frame that resembles a typical bridge structure. The problems are solved
using a computer program that was developed based on the procedure described
in this paper.


Beam Problem
The continuous beam shown in Fig. 3a has been considered by some authors
[3,4] to demonstrate different methods for determining influence lines.
Belegundu [3] only presented the influence line for the moment at support B
using the adjoint method. With the present method, the influence lines for any
action (reaction, shear, moment, axial force) at any point in the structure for
any load movement can be obtained in one cycle stiffness matrix analysis in
conjunction with simple statics. The input data for this problem are those
required for a standard stiffness matrix analysis plus the specification of the
members that would be traversed by the moving load (members 1,2,3 and 4 in
this problem).
The output of the program includes tabulated values of influence ordinates and
plots of influence lines. Due to space limitation, only plots of influence lines are
presented. The influence lines for support reactions, shears and moments at
selected points in the beam are shown in Figs. 4b-d. Obviously other influence
lines, if needed, can easily be obtained without repeating the stiffness matrix
analysis.

Frame Problem
The rigid frame shown in Fig. 4a is typical in bridge structures [7]. The
dimensions shown in the figure and member properties are chosen to reflect
those of real bridges of this type. Once again, the input data for this problem are
the data required for a standard stiffness matrix analysis plus the specification
of the members that would be traversed by the moving load (members 1,2 and 3
in this problem).
The influence lines for moments and shears in spans AB and BC are presented
in Figs. 4b-e. The curves shown describes moments and shears at (3 m, 6 m,...,l)
in the spans. Figure 4f shows the influence lines for moment, shear and axial
force in member BE.


CONCLUSION
A new procedure for computing the influence lines for beams and plane frames
is presented. The procedure is based on the standard stiffness matrix analysis of
the structure in conjunction with simple statics. The efficiency of the procedure
comes from the fact that it does not require the subdivision of members nor the
specification of load movement increment prior to the stiffness matrix analysis.
It needs only the specification of members that would be traversed by the
moving load. After the stiffness matrix analysis is performed, the computation
of influence lines for any action at any point in the structure with any load
movement increment is accomplished by simple calculations. This gives the
user the ability to study and precisely locate the critical sections in the structure
when subjected to moving loads. The procedure was applied to some building
and bridges structures (two of them are shown in this paper) and proved to be
precise and robust. The extension of the procedure to compute forces and
moments envelopes is underway and will be presented in upcoming paper.


Table 1: Comparison of the present method with the Adjoint and
Muller-Breslau methods.
Adjoint method

[2]
Muller-Breslau
[3]

Present Method
Member
subdivision
Needed for I.Ls.
within a member
Needed for all
cases
Not needed. CPU

time No. of
rounds of
FE analysis
One for each I.L.
within a member
One for each I.L. One for all
I.Ls.
Data preparation Needs new data
for each I.L.
within a member
Needs new data
for each I.L.
Once for all
I.Ls. (original
structure data)
Robust and
flexibility
Desired I.L.
needs to be
determined prior
to FE analysis
(input data
depends on the
required I.Ls.)
Same as adjoint
method.
All desired
I.Ls. can be
obtained after
FE analysis
using simple
statics.















(a) Continuous beam


-0.2
0.2
0.6
1
0 6 12 18 24 30 36 42 48 54 60

(b) Influence lines for support reactions


-2
-1
0
1
2
3
0 6 12 18 24 30 36 42 48 54 60

(c) Influence lines for moments at supports



-1
0
1
2
3
4
0 6 12 18 24 30 36 42 48 54 60

(d) Influence lines for moments at specified locations within beam spans.

1 2 3 4
12 ft 18 ft 15 ft 15 ft
E B D 1.6I 0.9I I 1.2I C
A
R
E

R
D
R
C
R
B
R
A

M
A

M
D
M
B

M
C

M
52ft

M
34ft

M
21ft

M
9ft

M
5ft


-0.5
0
0.5
1
0 6 12 18 24 30 36 42 48 54 60

(e) Influence lines for shears just to the right of supports B, C and D.

Figure 3: Influence lines for a beam problem.














(a) Rigid frame structure




-4
-2
0
2
4
0 6 12 18 24 30 36 42 48 54 60 66


(b) Influence lines for moments in span AB.


6 m
ft
E
5
A
F
B C
D
30 m 18 m
18 m
1 2 3
4
15 m
M
BA

(V
D
)
R

(V
C
)
R

(V
B
)
R


-3
-1
1
3
5
0 6 12 18 24 30 36 42 48 54 60 66


(c) Influence lines for moments in spans BC.





-1
-0.5
0
0.5
1
0 6 12 18 24 30 36 42 48 54 60 66


(d) Influence lines for shear in spans AB.



-1
-0.5
0
0.5
1
0 6 12 18 24 30 36 42 48 54 60 66


(e) Influence lines for shear in spans BC.




-2
-1
0
1
2
0 6 12 18 24 30 36 42 48 54 60 66

(f) Influence lines for moment, shear and thrust in column BE

Figure 4: Influence Lines for a Frame Problem. Note: The influence lines
shown for spans AB and BC represent the moments and shears at 3m-intervals.
M
BC

M
CB

(V
C
)
L

(V
B
)
R

Thrust
Shear
Moment (M
BE)

(V
B
)
L

(V
A
)


REFERENCES
[1] Hsieh, Y., Elementary Theory of Structures, Prentice-Hall, Inc., Englewood Cliffs, N.
J., 1970.
[2] Hibbeler, R. C., Structural Analysis, Prentice-Hall, Inc., Englewood Cliffs, N. J., 1995.
[3] Belegundu, A. D., The Adjoint Method for Determining Influence Lines, Computers &
Structures, Vol. 29, No. 2, 345-350, 1988.
[4] Cifuentes, A. and Paz, M., A Note on the Determination of Influence Lines and Surfaces
Using Finite Elements, Finite Elements in Analysis and Design, Vol. 7, 299-305, 1991.
[5] Washington State Department of Transportation, QconBridge Software, Version 1.0c-
March, 1997.
[6] McGuire, W. and Gallagher, R. H., Matrix Structural Analysis, John Wiley & Sons,
Inc., New York, 1979.
[7] Heins, C. P. and Lawrie, R. A., Design of Modern Highway Bridges, John Wiley &
Sons, Inc., New York, 1984.