A Finite Element Formulation for a

Beam with Varying Cross-Section

Geometry along Its Length: FlexibilityBased Approach
Bulent N. Alemdar, Ph.D., P.E.
SR Software Product Research Engineer
Bentley Systems, Inc.
Structural Engineering Group

Introduction
This study addresses a finite element formulation for a beam element with changing cross-section geometry along
its length. The study utilizes a flexibility-based formulation. In this method, internal element forces (axial and
bending) are used to derive exact form of element stiffness matrix. To this end, equilibrium equations are fully
satisfied along the element, which makes it sufficient to use one element per member to capture accurate results.
This is in contrast with conventional displacement-based formulation in which displacements fields are used to
obtain element stiffness matrix and equailibrium equations are only satisfied in a weighted integral form. Therefore,
one needs to use more than one element per member to capture accurate results.
In particular, a web-tapered beam element is chosen as an example but the formulation steps given in the study can
be easily extended to beams with different cross-section variation. Due to varying cross-section geometric properties along the length, the centroid axis is not a straight line, but rather a curved line. Hence, the proposed element
considers changes in the centroid axis and calculate the stiffness matrix that includes effects of the curved centroid
axis.
An elastic material behavior is assumed and no geometric nonlinear effects are considered. In addition, it is
assumed that deformations are small and section remains section after the deformation. This study also provides a
method to include shear deformations within the proposed element. In addition, it is shown how one can include
distributed load effects on the element response. Finally, two numerical examples are provided to show the merits
of the proposed element.

8/17/11

Geometry Definition on Cross-section and Along Beams Length

The element cross-section definition is given in the following figure in which unequal flange lengths and thicknesses are used. Also, a linear variation of element depth ("h" in the figure) is assumed in this study.

Cross-section Area, A(x)

Referring to the figure above, the cross-section area at any point along the element can be represented as follows:
A HxL = tft bft + tfb bfb + h HxL tw

(1)

in which x is the web height of the member. A linear variation of "h" is assumed:
h HxL =

where h1 and h2 are shown in the following figure.

h2 - h1
x + h1
L

(2)

8/17/11

Combining Eq. (1) and (2), it is obtained that

A HxL = A1 + 1 x

in which

(3)

A1 = tft bft + tfb bfb + h1 tw

Hh2 - h1L
1 =
tw
L

(4)

and A1 is the area of the cross-section at left end of the element. It should be noted that A(x)=A1 if 10. In other
words, it is a constant cross-section (prismatic element) if 10.

Cross-section Torsion Constant, J(x)

Section torsional constant is expressed according to the following equation:
J HxL =

tft 3 bft +

tfb 3 bfb +

1
3

tw 3 h HxL

(5)

where it is assumed that the cross-section is composed of thin components. Thus,

J HxL = J1 + 1 x

where
1
J1 =

tft 3 bft +

3
1 =

tfb 3 bfb +

3
Hh2 - h1L 1
L

(6)

tw 3 h1

(7)

tw 3

Again, note that J1 is the section torsion constant at left end of the element.

Cross-section Moment of Inertia: Iyy(x)

In order to calculate Iyy , position of cross-section centroid is needed. Note that the centroid axis is a curved line due
to effects of varrying cross-section geometry along the element. This can be explicitly expressed as follows:
`
z HxL =

1
2

bfb t2fb + bft h HxL + tfb +

tft

tft + h HxL

2
tft + h HxL tw L

h HxL
2

+ tfb tw

(8)

Hbfb tfb + bft

`
where "z (x)" is measured from the bottom of the cross-section (see the figure). Finally, an explicit form of Iyy is
obtained as follows:

8/17/11

Iyy HxL =

bfb tfb

1
12

bft tft 3 + bft tft h HxL + tfb +

`
z HxL -

tfb

tw Hh HxLL + tw h HxL
3

`
- z HxL

tft

12

Cross-section Moment of Inertia: Izz (x)

h HxL
2

1
+

bfb tfb 3 +

12
`
+ tfb - z HxL

(9)

Since the cross-section has a symmetry along z-z axis, the position of the centraid is always at y = 0. Thus, the
following immediately applies:
Izz HxL =

bft 3 tft +

12

bfb 3 tfb +

12

12

tw 3 h HxL ;

(10)

Element Formulation: Flexibility Based Approach

The principle of virtual work can be expressed as follows:

Lo

FT d x - QT q = 0

(11)

where F is the weigted function (which is chosen in such a way that it satisfies equilibrium equations), d is the
generalized cross-section strains, Q is the force vector applied at element ends and finally, q is the displacement
vector. It should be noted that the above integration is carried out at initial position (i.e., Lo is the initial length of
the element).
A consistent linearization of Eq. (11) leads to element flexibility matrix. Details of the linearization steps are not
given in this study but the reader is referred to the following publication: Alemdar, B.N. and White, D.W., (2005),
Displacement, Flexibility and Mixed Beam-Column Finite Element Formulation for Distributed Plasticity
Analysis, Journal of Structural Engineering, Vol. 131, No. 12, pg. 1811-1815.
In general form, the flexibility matrix for axial, torsional and bending terms are respectively expressed as follows:
FA =

Lo

FJ =

Lo

FByy =

Lo

FBzz =

Lo

E A HxL
1

G J HxL
1

HDA HxLLT DA HxL x

HDJ HxLLT DJ HxL x

E Iyy HxL
1

E Izz HxL

IDyy HxLM Dyy HxL x

(12)

(13)

HDzz HxLLT Dzz HxL x

(14)

(15)

in which E and G are modulus of elasticity and shear, respectively, and A(x), J(x), Iyy HxLand Izz HxL are cross-

section geometry functions which are derived in the previous sections for a web-tapered beam element. Finally,
matrices D(x) are referred to as interpolated stress-resultant force fields and they satisfy equilibrium equations. This
is further explained in coming sections.
It should be noted that Eqs. (12) - (15) are exact flexibility matrices of the proposed element. Exact stiffness
matrices are simply the inverse of the flexibility matrices. The inverse of the flxibility matrices always exists since
D(x) contains the section forces corresponds to deformations only (i.e. D(x) does not contain any rigid body
modes).

8/17/11

Flexibility Matrix for Axial Deformations

In the absence of element loading, the axial force inside the element is constant. Hence, interpolated internal axial
force can be expressed as follows:
D HxL = DA T Q = H1L HQ1 L

in which Q1 is the axial force in the element. Then, Eq. (12) becomes
FA =

Lo

It is interesting to note that

LogA1 +

E A HxL

x =

(16)

Lo 1
E
A1

(17)

E 1

LogA1 +
Limit10

Lo 1
E
A1

Lo

E 1

(18)

E A1

In other words, the above term reduces to conventional flexibility term for a prismatic element.

Flexibility Matrix for Torsinal Deformations

Torsinal stiffness term is obtained in a similar way that the axial term is obatined. Again, under no torsional
loading, one can write that
D HxL = DJ T Q = H1L HQ2 L

(19)

in which Q2 is the constant torsional force in the element. Then, Eq. (13) becomes
FJ =

Lo

It is again interesting to note that

LogA1 +

GJ HxL

x =

Lo 1
E
J1

(20)

G 1

LogA1 +
Limit10

Lo 1
E
J1

G 1

Lo

(21)

G J1

Element Formulation for Bending in Major Axis

If the element is subjected to end moments only, the moment field inside the element is linear.

Thus,
D HxL = Dyy T Q = I 1 -

x
L

- xL M J

M1
N
M2

(22)

One can substitute Dyy into Eq. (14) to obtain the flexibility terms for major axis bending. Due to higly nonlinear
nature of Iyy , it is not practical to obtain a closed form solution. Instead, numerical evaluate of Eq. (14) is preferred.

Element Formulation for Bending in Minor Axis

The stiffness terms for bending in minor axis follows the similar steps given in the previos section. Referring to the
following figure, one can write

8/17/11

D HxL = Dzz T Q = I

x
L

-1

x
L

And then, Eq. (15) is again numerically evaluated.

MJ

M1
N
M2

(23)

Shear Deformations Due to Bending

Shear deformations due to bending are integrated into the formulation in such a way that an equivalent shear area
concept is utilized. The complementary internal virtual work due to shear deformations is
W*int =
Vo

where = =

V
As

1
V

(24)

, V: shear at cross-section and As is equivalent shear area of the cross-section. It should be noted

that the above integration is carried out over the initial volume of the element. Then, substituting dV = As dx, Eq.
(24) becomes
W*int =

As

V x

Vo

(25)

Shear stresses due to applied end moments are

V=I

1
L

1
L

MJ

M1
N = I
M2

1
L

1
L

MM

(26)

and equivalenet shear area along the member is assumed to be changing linearly:
As HxL = As1 K1 -

x
O + As2 K O
L
L
x

(27)

where As1 and As2 are the equivalent shear areas at element left and right end, respectively. Thus, Eq. (26) becomes
W*int = M

1 1
G L2

L
1
1 1
N
x M
1 1
0 As HxL

(28)

or, flexibility matrix due to shear deformations under major axis bending is
1 1
Fsmaj =

G L2

L
1
1 1
N
x
1 1
0 As HxL

(29)

Note that this matrix should be added to Eq. (14) before taking the inverse of the flexibility matrix. Again, numerical integration can be used to evaluate the integral.
A similar approach is also followed for shear deformations due to bending in minor axis. Thus,
1 1
Fsmin =

G L2

L
1
1 1
N
x
1 1
0 As HxL

(30)

where As (x) represents equivalent shear areas under minor axis bending.

Distributed Load (Major Axis Loading)

Distributed loads effect on the element response can be also included within the flexibility-based formulation. In
this section, a trapezoidal load is studied as an example to show how to calculate fixed end moment effects. Other
types of loads can be also handled in a similar way.
The element is subjected to a trapezoidal load as shown below and the load definition F(x) is
6

8/17/11
Distributed loads effect on the element response can be also included within the flexibility-based formulation.
In
this section, a trapezoidal load is studied as an example to show how to calculate fixed end moment effects. Other
types of loads can be also handled in a similar way.
The element is subjected to a trapezoidal load as shown below and the load definition F(x) is

F HxL =

q1 - q2

a1 q2 - a2 q1
x +

= 1 x + 2

a1 - a2

(31)

a1 - a2

And one can calculate corresponding deflections due to the applied load as follows:
=J

Lo Dzz T M HxL
1
N=
x
2
E Iyy HxL
0

(32)

in which Dzz and Iyy HxL are explicity derived in previous sections. Note that M(x) is the moment field due to
applied load F(x) and it is given below:
M HxL =

where
1
V1 =
L

V1 x - q1

V1 x - q1 Ha2 - a1L J

q1 Ha2 - a1L

a2-a1
2

a2 - a1
2

Hx-a1L2
2

V1 x

- HF HxL - q1L

0 x < a1

Hx-a1L 2

a2-a1

+ x - a2N - Hq2 - q1L J

6
a2-a1

+ L - a2 + Hq2 - q1L

a1 b x b a2

NJ

+ x - a2N

a2 - a1

a2 - a1

(33)

a2 b x b L

+ L - a2

(34)

And finally, one can convert calculated end rotations to equivalent "Fixed End Moments" as follows
Ff = J

Mf1
N = KByy
Mf2

(35)

in which KByy is the element stiffness matrix, which is the inverse of the flexibility matrix given in Eq. (14).

Closing Remarks
The stiffness matrix developed in the previous sections must be augmented with rigid body modes to obtain a
complete form of element stiffness matrix. The stiffness matrix given in the previous section is deu to the deformations modes, in which rigid body modes are excluded (see the first figure below). Then, one needs to add rigid body
modes (hence, size of the stiffness matrix is chaged from (6,6) to (12,12)) and this can be accomplished with a
transformation matrix T.
K = TT Ke T

(36)

8/17/11

Numerical Examples:
Two numerical examples are provided in this study, which are also used as a benchmark test examples in AISC
Steel Design Guide 25: Frame Design Using Web-Tapered Members. In both examples, the element stiffness
matrix is calculated based on the proposed element formulation and compared to those given in the reference. Note
that only one element per member is sufficient because the exact form of stiffness terms are calculated.
Doubly Symmetric Web-Tapered Beam-Column:
The problem definition is provided in the figure below. The following numerical values are used:
bft = 6 in. tft = 0.25 in. bfb = 6 in. tfb = 0.25 in. tw = 0.125 in.
h2 = 25 - 2 * 0.25 = 24.5 in. h1 = 10 - 2 * 0.25 = 9.5 in.
E = 29 000 ksi G = 11 153.85 ksi
L = 16.36* 12 = 196.32 in.

8/17/11

K=

11.511

-760.578

-11.511

-1499.26

0.414058

40.6421

-0.414058

40.6458

748.534

-748.534

40.6421

5319.12

-40.6421

2659.74

-760.578

82 213.7

760.578

67 103.1

4.17358

-4.17358

-11.511

760.578

11.511

1499.26

-0.414058

-40.6421

0.414058

-40.6458

-748.534

748.534

40.6458

2659.74

-40.6458

5319.84

-1499.26

67 103.1

1499.26

227 232.

-4.17358

4.17358

Doubly Symmetric Web-Tapered Beam-Column:

The problem definition is provided in the figure below. The following numerical values are used:
bft = 6 in. tft = 0.5 in. bfb = 6 in. tfb = 0.375 in. tw = 0.21875 in.
h2 = 40.75 - 2 * 0.4375 = 39.875 in. h1 = 10 - 2 * 0.4375 = 9.125 in.
E = 29 000 ksi G = 11 153.85 ksi
L = 15.10* 12 = 181.2 in.
The second example is given below. Member cross section is defined with unequal flange lengths. The following
numerical values are used for the example:

8/17/11

K=

51.4257

-2374.56

-51.4257

-6943.79

0.922516

83.5563

-0.922516

83.6037

1639.48

-1639.48

83.5563

10 092.2

-83.5563

5048.23

-2374.56

207 819.

2374.56

222 450.

27.0087

27.0087

-51.4257

2374.56

51.4257

6943.79

-0.922516

-83.5563

0.922516

-83.6037

-1639.48

1639.48

83.6037

5048.23

-83.6037

10 100.8

-6943.79

222 450.

6943.79

1.03576 106

27.0087

27.0087

Conculusion
In this study, a flexibility-based element formulation is adapted to derive exact stiffness matrix for a beam with
varying geometric cross-section properties along its length. The proposed element considers curved centroid axis
position within its stiffness formulation. Calculated stiffness terms are exact (the only approximation introduced
into the formulation comes from numerical integration of corresponding stiffness terms.). To this end, one element
per member is sufficient to capture accurate results for members with varying cross-sections. Shear deformations
and effects of distributed loads on the element are also studied in this study.
The proposed formulation particularly targets beams with linearly tapered web, but the concept presented in this
paper can be easily extended to other type of varying cross-sections. Currently, the proposed element is used in a
structural software poduct , RAM Elements (www.bentley.com/en-us/products/ram%20elements/).

Author
Bulent N. Alemdar is a Senior Software Product Research Engineer at Bentley Systems, Inc. He received his Ph.D.
from Georgia Institute of Technology in Structural Engineering. His research interests includes nonlinear finite
element analysis, structural stability, structural dynamics and software developement. He can be reach at [email protected].

10