The Finite Strip Method (FSM)

1. Introduction
This is the method of semi-numerical and semi-analytical nature. It is suitable for the analysis of
rectangular plates and plane-stress elements or structures being the combination of both.
Therefore, the following types of civil engineering structures can be dealt with: bridge slabs, box
girders, roofs consisting of plane elements, etc.
However, the method has one important drawback, which limits its versatility the analysed
elements must be rectangular and simply-supported at two opposite edges.

bridge slabs

roofs of industrial halls

2. Analysis of plates by FSM

The plate is considered as thin, i.e. it is assumed that the variation of displacements across the
plate thickness in negligible and the plate is replaced with a plane surface corresponding to the
mid-surface of the real 3D body. Also the in-plane deformations are neglected. Hence, at each
point of the mid-surface the deflection and its derivatives with respect to two in-plane co-ordinates
x and y are enough to define the deformation state.
2
22
y
q(x,y)
1

q(x,y)
I

2
x
11

q(x,y)

The co-ordinate system is introduced with the axis x along two opposite simply-supported edges.
The two edges parallel to the axis y may have any type of support.
In the FSM the plate is divided into the finite number of strips running along the axis y, thus each of
the strips is simply-supported. The strips are connected along the so called nodal lines. All the
strips may have any width, not necessarily the same one.
The starting point for the derivation of the method procedures is the approximation of displacement
function wI(x,y) for the points on a single strip I. Let us denote the nodal lines along this strip as i
and j.

i
j

L
wi

wj
wi

b
Nodal line

Finite strip

wj

Polynomial

Sine series

The deflection of the strip is approximated as a combination of the sine harmonic series in the
longitudinal direction y (analytical aspect) and the polynomial function fm in the transverse direction
x (numerical aspect)
w I (x, y ) =

fm (x ) sin

m =1

r
my
my
= (Am + Bm x + ...) sin
L
L
m =1

where: r is the adopted number of harmonic functions in the series and Am, Bm, etc. are the
coefficients of the polynomial function fm, which must be found from the boundary conditions
corresponding to the deflections and transverse slopes at the nodal lines i and j.
The continuity conditions along the nodal lines require the adjacent strips to have the same
deflections and slopes. For an arbitrary strip and the m-th harmonic function we have four
prescribed nodal values wim, im, wjm, jm, so the required polynomial must be of the third order

fm (x ) = Am + Bm x + Cm x 2 + Dm x 3
The nodal values of displacements and slopes are

wi =

i =

w im sin

m =1
r

im sin

m =1

my
L

wj =

my
L

j =

my
L

w jm sin

m =1
r

jm sin

m =1

my
L

The unknown coefficients in the polynomial are obtained from the boundary conditions
x=0

fm (x ) = w im

and

fm (x )
= im
x

x=b

fm (x ) = w jm

and

fm (x )
= jm
x

and after some reordering one gets:

fm (x ) = C0 i w im + C1i im + C0 j w jm + C1 j jm
where the coefficients C are

C0 i = 1 3

x2
x3
+
2
b2
b3

C1i = x 2

C0 j = 3

x2
x3

2
b2
b3

C1 j =

x2 x3
+
b b2

x3 x2

b
b2

Note, that these functions are identical with the shape functions N2, N3, N5 and N6 in the clampedclamped beam.

The deflection approximation can be given in the matrix form

w I (x, y ) =

Cb w bm
I

sin

m =1

my
L

with the following definitions of the vectors

C b = C0 i
I

C1i

C0 j

C1 j

w bm

w im

im
=
w jm

jm

The superscript I denotes the I-th strip and the subscript b the bending state. Note, that the nodal
parameters wim, im, wjm, jm, are actually the amplitudes of m-th harmonic functions describing the
deflections and transverse slopes along the nodal lines.
Having specified the deflection approximation in terms of nodal parameters wim, im, wjm, jm, we
can consider the energy and derive the equilibrium conditions for the strip.
The total energy for one strip I consists of the strain energy and the energy of loading

U I = Us + Uq
I

The strain energy can be expressed in terms of bending and torsional moments, Mx, My, Mxy and
the corresponding curvatures
L b

Us

1
=
20

2w I
2w I
2w I
Mx

M
+
2
M
y
xy

xy
x 2
y 2
0

dxdy

while the loading energy involves the deflections

L b

Uq =
I

q (x, y )w dxdy
I

0 0

In the matrix form the strain energy can be written as

L b

Us =
I

T
1
MI I dxdy
200

where the vectors of moments and curvatures

2w I

2
2x I
w
I =
y 2
2 I
2 w
xy

Mx

M = My
M xy

were introduced. After the substitution of deflection wI the curvatures vector takes the form
I =

m =1

where the matrix BbmI is

I
bm

w bm

B bm

C0i ' ' sin k m y

2
= k m C0 i sin k m y
2k mC0i ' cos k m y

C1i ' ' sin k m y

C0 j ' ' sin k m y

k m C1i sin k m y

k m C0 j sin k m y

2k mC1i ' cos k m y

2k mC0 j ' cos k m y

C1 j ' ' sin k m y

2
k m C1 j sin k m y
2k mC1 j ' cos k m y

m
L
Let us assume that the plate material is isotropic. Then from the theory of isotropic plates the
moments can be found

and k m =

2w I
2w I
+
M x = D
D
1
2
y 2
x
2w I
2w I
+
M y = D
D
1
2
x 2
y
M xy = 2Dxy

2w I
xy

where the plate stiffness, the coupled stiffness and the torsional stiffness

Eh 3
12 1 2
D1 = D
D=

1
D
2

Dxy =

for the plate of the thickness h were introduced. The vector of moments can be now expressed in
the matrix form
MI = D b I
I

with the matrix of bending stiffness coefficients for the isotropic plate

Db

D D1 0

= D1 D
0
0 0 Dxy

The moment vector can also be expressed in terms of the nodal displacement parameters
MI =

I
b

B bm w bm

m =1

and its transpose is

T

MI =

IT
bm

IT

B bm D b

m =1

Hence, the strain energy takes the form

Us

L b

1 r r
IT
IT
I
I
I
=
w bn B bn D b B bm dxdy w bm
2 m =1 n =1
0 0

The double integral will involve the following expressions

sin k m y sin k n ydy = L 2

for

cos k

my

mn

for

m=n

0 for m n
cos k n ydy =
L 2 for m = n

Hence, the single sum is sufficient in this expression and we get

Us =
I

L b

1 r
IT
IT
I
I
I
w bm B bm D b B bm dxdy w bm
2 m =1
0 0

The loading energy takes the form

Uq =
I

w bm

IT

m =1

L b

IT
bm

q (x, y ) sin k m ydxdy

0 0

The energy of the entire plate Up is the sum of energies for all the N-strips
Up =

I =1

In the presented energy approach the theorem is used saying, that the system is in the equilibrium
when the entire potential energy has the minimal value. This leads to the conditions of vanishing
partial derivatives of Up calculated with respect to the nodal displacement parameters in wbmI for
every m-th harmonic function
U p
w 0m

= 0,

U p
0 m

U p

= 0,

w 1m

=0,

U p
1m

=0,

U p
w Nm

= 0,

After this differentiation the set of equations is obtained for every m

N

K bm w bm =
I

I =1

bm

I =1

where the stiffness matrix for a single strip I and m-th harmonic function
L b

K bm =
I

IT
bm

D b B bm dxdy

0 0

and the loading vector for this strip and this harmonic function
L b

Pbm =
I

IT
bm

q (x, y ) sin k m ydxdy

0 0

were introduced.
The explicit form of the strip stiffness matrix is

K bm

k b1 k b 3

k b2
=

sym.

kb4
k b5
k b1

k b5
k b 6
k b3

k b2

with:
k b1 = 6

L
13
12 L 2
6L 2
4
D+
Lbk m D +
k m Dxy +
k m D1
3
70
5 b
5b
b

U p
Nm

=0

k b2 = 2

L
1
4
2
4
2
2
D+
Lb 3 k m D +
Lbk m Dxy +
Lbk m D1
b
210
15
15

k b3 = 3

L
11
1
3
4
2
2
D+
Lb 2 k m D + Lk m Dxy + Lk m D1
2
420
5
5
b

k b 4 = 6

L
9
12 L 2
6L 2
4
D+
Lbk m D
k m Dxy
k m D1
3
140
5 b
5b
b

k b 5 = 3

L
13
1
1
4
2
2
D+
Lb 2 k m D Lk m Dxy
Lk m D1
2
840
5
10
b

k b6 =

L
1
1
1
4
2
2
D
Lb 3 k m D Lbk m Dxy
Lbk m D1
b
280
5
30

The loading vectors depend on the type of loading. For instance:

- the point load

y
P0
L
x0

Pbm

y0
b

2
3

x0
x0
1 3 2 + 2 3
b
b

2
3

x0
x0
x
2

0
b
b 2 P sin k y
=
0
m
2
3
3 x0 2 x0

b2
b3

2
3
x0 + x0

b
b 2

- the patch load

y
q0

Pbm

x1
x2

y1 y
2

x3
x4
x 2 + 3
b
2b
2

3
4
x 2 x + 1 x
2

= 2 33 b 4 4 b q0 c m
x
x

3
2

b
2b

1 x3 1 x4
+

3 b 4 b2

x
cm =

1
(cos k m y 1 cos k m y 2 )
km

x n = x2 x1
n

for n = 1, 2, 3 and 4

The assembly of the stiffness matrix Kbmp and the loading vector Pbmp for the entire plate divided
into N-strips is carried out according to the following scheme
w0m, 0m
0
1
w1m, 1m
1
2
w2m, 2m
2
3
w3m, 3m
3

wNm, Nm

w0m
w1m

0m

w1m

Kbm2

1m

w2m

1m

Kbm3

2m

Kbmp =

w0m Pbm1

Kbm1

0m

Pbm2

w2m

2m

w3m

w3m

Pbmp =

3m

3m Pbm3

KbmN

PbmN

wNm

wNm

Nm

Nm

Note, that the directions of axes of local co-ordinates in the strips coincide with global co-ordinates
and no transformation is necessary.
As for the boundary conditions, the simple supported opposite edges are already inherent in the
system in the form of the sine series functions, which fulfil the conditions of vanishing
displacements w and bending moments (second derivatives of w with respect to y).
On the other hand, the support conditions on two remaining edges must be introduced. If any of
the edge parameters is vanishing, then the corresponding rows and columns in Kbmp and the
corresponding element of Pbmp can be removed or replaced with zeros. For instance, if the edge
along the 0-nodal line is simply supported, then the displacement w0m vanishes and the first row
and the first column in Kbmp must be modified as well as the first element of Pbmp.
In this way the global set of equilibrium equations for the entire plate for the m-th harmonic function
is obtained
K bm w bm = Pbm
p

The solution of these equations provides the vector of amplitudes of sine functions for deflections
and transverse slopes along all the nodal lines for the m-th harmonic function. The value of the
displacement at an arbitrary point of the plate is obtained by a summation of results for all the
assumed r-harmonic functions according to the formula
w I (x, y ) =

C b w bm
I

sin

m =1

my
L

The method ensures the continuity of deflections and slopes between the strips along the nodal
lines. However, due to the approximate form of the displacement function in the strips, the bending
and torsional moments calculated using the second derivatives of displacements are not
continuous. The approximate values of moments along the nodal lines can be obtained as mean
values computed from the moments yielding from two adjacent strips.
For instance, for the nodal line i lying between the strips I1 and I we get

0
1

I1
i

Mright

I 1

I 1
b

B bm

I 1

(x = b )w bm I 1

m =1

I
Mleft =
I

1
I 1
I
Mright + Mleft
2
where:
Mi =

m =1

I
b

B bm (x = 0 )w bm
I

and the appropriate matrices Bbm have the form

6
b 2 sin k m y

I 1
B bm (x = b ) =
0

6
b 2 sin k m y
2
I
B bm (x = 0 ) = k m sin k m y

2
sin k m y
b
0
0

4
sin k m y
b
0
2k m cos k m y

6
sin k m y
b2
2
k m sin k m y
0

sin k m y
b

2k m cos k m y

6
sin k m y
b2
0
0

sin k m y
b

The differences between the moments obtained from the adjacent strips decrease with the
increasing number of strips.
Note, that generally the accuracy of the results obtained using the FSM depends on two
parameters: the number of strips N and the number of the harmonic functions r.
3. Analysis of plane stress elements (plate-like elements loaded in their plane) by FSM
We consider thin plane elements, i.e. it is assumed that the variation of displacements across the
element thickness in negligible and the element is replaced with a plane surface corresponding to
the mid-surface of the real 3D body. Due to the existence of only in-plane loading the
displacements are also only in-plane. Thus, we have two displacement functions u and v. Likewise
in the plate analysis, the element is divided into a finite number of strips, which span the entire
length of the element between two simply-supported opposite edges.
y
u = 0, y = 0
p( x )

2 3

p( y )
L

v(x,y)
u(x,y)
u = 0, y = 0
x

Let us introduce the strain and stress vectors

u

x x
v

= y =
y
xy

u v
+

y x

and write down the physical law for the plane stress element
= Dp

x

= y
xy

where the plane stress stiffness matrix has the general form
Ex xEx

Dp =
(1 x y ) y0E y E0y

(1 x y )E xy
0
0

For the isotropic material

E x = E y = E , x = y = , E xy =

E
=G
2(1 + )

The boundary conditions for the simply supported opposite edges y = 0 and y = L are
u = 0 , y = 0
The approximation of displacements in a single strip has the form
my
my

r
r
sin L
sin L
u
(E + Fm x + ...)

fm (x ) =

=
my m
v m =1 cos my
m = 1 cos

L
L

where Em, Fm, etc. are the unknown coefficients. The assumed form of displacement functions
fulfils the boundary conditions at the simply supported edges. The presence of the function
sin(my/L) in the displacement u ensures the fulfilment of the displacement boundary conditions.
As for the stress boundary conditions we have

y = x + y =

u
v
+
x
y

The differentiation with respect to x in the first term does not change the function sin(my/L), while
the differentiation with respect to y in the second term transforms the function cos(my/L) into
sin(my/L). Hence, both the terms contain the desired function sin(my/L) and the conditions for
the vanishing y are also automatically fulfilled.
The calculation of displacements for the strip I limited by the nodal lines i and j involves the
following nodal parameters: uim and ujm for u as well as vim and vjm for v. Thus, each of the two
approximations of displacements will require a linear polynomial function with two coefficients Em
and Fm. They are found from the boundary conditions, which for instance for u read:
x=0

fm (x ) = u im

x=b

fm (x ) = u jm

This leads to the following matrix relation

r
u
I
I
=
C pm w pm

v
m =1

where the matrix of coefficients is

C pm

x
my
1 b sin L

0
x
my

1 cos
b
L

x
my
sin
b
L
0

x
my
cos
b
L

and the nodal parameters for the single strip I are assembled into the vector

w pm

u im

v im
=
u jm

jm

Again, similarly as in the plate analysis, to derive the strip stiffness matrix and the loading vector
the energy approach is used. The strain energy for a strip I in the plane stress element can be
expressed as
Us =
I

h
2

L b

h
2

I I dxdy =

0 0

L b

+ y y + xy xy )dxdy

x x

0 0

The potential energy of the loading is

L b

Uq =
I

(p(x )u + p(y )v )dxdy

0 0

The vector of strains can be expressed as

=

I
pm

w pm

m =1

with the matrix of derivatives

B pm

1
sin k m y

=
0

k 1 x cos k y
m
m
b

1
sin k m y
b

0
x

k m 1 sin k m y
b

1
cos k m y
b

0
km

x
cos k m y
b

x
k m sin k m y
b

1
cos k m y

b
0

m
.
L
With this in hand the physical law can be expressed in the matrix form
and k m =

I
p

B pm w pm

m =1

Thus, the energy parts can be put as:

Us =
I

Lb

h r
IT
IT
I
I
I
w
B pm D p B pm dxdy w pm

pm

2 m =1

0 0

Uq =
I

L b

IT
I p (x )
w pm C pm
dxdy
p(y )
m =1
0 0

Considering the fact, that the energy for the entire plane stress element consists of the energies for
all the strips
Ue =

I =1

and using the similar conditions of equilibrium as in the plate case

U e
=0
w pm

we get the set of equilibrium equations for the entire element for each m-th harmonic function
N

I
pm

w pm =
I

I =1

pm

I =1

The stiffness matrix for the single strip I is obtained from

L b

K pm = h
I

IT

B pm D p B pm dxdy

0 0

and its explicit form is

K pm

k p1 k p 3

k p2
=

sym.

k p4
k p5
k p1

k p5

k p6
k p3

k p 2

where:
2

k p1 =

Lk
Lbk m
L
L
Lbk m
+
, k p2 =
+
, k p 3 = m ( )
4
2b
6
6
2b
2

kp4 =

Lk
L
Lbk m
Lbk m
L
+
, k p 5 = m ( + ) , k p 6 =

2b
6
4
12
2b

and

hE x
,
1 x y

hE y
1 x y

, = h x = h y , = hE xy

The following steps of the solution are analogous to the plate analysis.
4. Analysis of compound structures
Let us consider a box girder structure and its discretisation into the finite strips. The cross-section
looks like this

Such structures can be considered as compound of rectangular elements, which are

simultaneously subjected to bending and plane stress action. The stiffness matrix of such an
element (strip) can be obtained as an appropriate assembly of stiffness matrices for strips in the
bending state and in the plane stress state.
The equilibrium for a single strip in its local co-ordinates can be expressed by the matrix equation
~ I~ I ~ I
Km w
m = Pm
Let us represent the strip matrices in bending and plane stress using the (22)-submatrices

~ I
K bm

k b1 k b 3

k b2
=

sym.

kb4
k b5

~ I
K pm

k p1 k p 3

k p2
=

sym.

k p4
k p5

k b1

k p1

k b5
k b 6 K bmii I
=
k b 3 K bmji I

k b2

I
K bmij
I
K bmjj

k p5

k p 6 K pmii I
=
k p 3 K pmji I

k p 2

I
K pmij
I
K pmjj

Now the stiffness matrix for a strip in a compound structure can be given in the following form

~ I
Km

K pmii I

sym.

K pmij

K bmii

0
K pmjj

0
I
K bmij
0

I
K bmjj

This matrix corresponds to the following vectors of nodal displacement parameters (amplitudes)
and nodal forces for the I-th strip and the m-th harmonic function

~ I
w
m

u~im
~
v im
~
w
im
~
im
= ~
u
~jm
v jm
w
~
im
~jm

~ I
Pm

~
X
im
~
Yim
Z~
~im
M
= ~ im
X jm
Y~jm
~
w im
~
M
jm

Each nodal line has four nodal parameters three for displacements and one for transverse slope.
y, y~

Displacements and forces

in local co-ordinates
i

z~ z

~jm ,
~
M jm

~
u~jm , X jm
x~
~ , Z~
w
jm
jm

z~ z

jm ,
M jm

x~
z

v jm ,Y jm

y, y~

z~

i
j

~
v~jm ,Y jm

y, y~

Displacements and forces

in global co-ordinates

u jm , X jm
w jm , Z jm

x~

The relations between the forces are

~
X jm = X jm cos + Z jm sin
~
Y jm = Y jm
~
Z jm = X jm sin + Z jm cos
~
M jm = M jm
what leads to the following transformation rule

cos

0
sin

~
0 X jm X jm

0 Y jm Y jm
= ~
0 Z jm Z jm
~

1 M jm M

jm

sin

1
0
0 cos
0

The direction cosine matrix can be denoted as

cos

0
C=
sin

0 sin
1
0
0 cos
0

0
0

and the transformation matrix is

C 0
T=

0 C

Now the transformation of vectors of the displacement and force nodal parameters can be given as
~ I
I
~ I = Tw I
w
Pm = TPm
m
m
I
~ I
w m = TT w
m

~ I
I
Pm = T T Pm

and for the strip stiffness matrix we have

~ I
I
K m = TT K m T
The following steps of the solution are analogous to the plate analysis.
5. Analysis of multi-span and column supported plates by the FSM
The plates with inter-span supports, both point type or knife-edge type, can be solved using a
combination of the FSM and the flexibility method. Reactions in the additional inter-span supports
are considered as redundant forces and the single span plate for which the solution can be found
using the pure FSM is taken as a modified (determinate) system. Thus the scheme of the flexibility
method is set. The continuous reaction in the knife-edge support can be approximated as a set of
point loads, for instance at the points where the support line intersects the nodal lines of the FSM
discretisation.

P
X1
X3
XN
X2

The identity of the modified system with removed supports and the original system with supports
present is ensured by the kinematical conditions requiring the deflections at the removed supports
to be zero

1 = 0
2 = 0
...

N = 0
After consideration of the basic states: X1 = 1, X2 = 1, XN = 1 and P and the application of the
superposition rule the canonical equations of the flexibility method are obtained

11 X 1 + 12 X 2 + ... + 1N X N + 1P = 0
21 X 1 + 22 X 2 + ... + 2N X N + 2P = 0
...

N 1 X 1 + N 2 X 2 + ... + NN X N + NP = 0
The flexibility coefficients ik, which are the appropriate negative deflections (the redundant forces
vectors have upwards orientation, opposite to the axis z) at the points where forces Xi are applied
are obtained from the FSM analysis of the modified one-span plate loaded by an appropriate
loading state Xk = 1. For instance, the set of displacements 1i, 2i, Ni follows from the state
Xi = 1.

1i
3i
ii

Ni

2i
Xi = 1
Thus, N basic states and the state P must be solved to formulate the set of canonical equations.
These equations are then solved and the values of redundant forces are found.
Then the last stage of the solution follows, where the modified one-span plate is loaded
simultaneously with the external loading and all the redundant forces. Alternatively, the
superposition rule can be used but this requires the computer storage of the complete results for all
the basic states, including deflections and moments.