Brigham Young University

BYU ScholarsArchive

Theses and Dissertations

2009-03-16

Second-Order Structural Analysis with One Element per Member

Jesse William Lyon
Brigham Young University - Provo

Follow this and additional works at: https://scholarsarchive.byu.edu/etd

Part of the Civil and Environmental Engineering Commons

BYU ScholarsArchive Citation

Lyon, Jesse William, "Second-Order Structural Analysis with One Element per Member" (2009). Theses
and Dissertations. 1756.
https://scholarsarchive.byu.edu/etd/1756

This Thesis is brought to you for free and open access by BYU ScholarsArchive. It has been accepted for inclusion
in Theses and Dissertations by an authorized administrator of BYU ScholarsArchive. For more information, please
contact [email protected], [email protected].
SECOND-ORDER STRUCTURAL ANALYSIS WITH ONE

ELEMENT PER MEMBER

by

Jesse W. Lyon

A thesis submitted to the faculty of

Brigham Young University

in partial fulfillment of the requirements for the degree of

Master of Science

Department of Civil and Environmental Engineering

Brigham Young University

April 2009
 BRIGHAM YOUNG UNIVERSITY

GRADUATE COMMITTEE APPROVAL

of a thesis submitted by

Jesse W. Lyon

This thesis has been read by each member of the following graduate committee and by
majority vote has been found to be satisfactory.

Date Richard J. Balling, Chair

Date David W. Jensen, Member

Date Paul W. Richards, Member

 BRIGHAM YOUNG UNIVERSITY

As chair of the candidate’s graduate committee, I have read the thesis of Jesse W.
Lyon in its final form and have found that (1) its format, citations, and bibliographical
style are consistent and acceptable and fulfill university and department style
requirements; (2) its illustrative materials including figures, tables, and charts are in
place; and (3) the final manuscript is satisfactory to the graduate committee and is
ready for submission to the university library.

Date Richard J. Balling

Chair, Graduate Committee

Accepted for the Department

E. James Nelson
Graduate Coordinator

Accepted for the College

Alan R. Parkinson
Dean, Ira A. Fulton College of Engineering
and Technology
 ABSTRACT

SECOND-ORDER STRUCTURAL ANALYSIS WITH ONE

ELEMENT PER MEMBER

Jesse W. Lyon

Department of Civil and Environmental Engineering

Master of Science

In this thesis, formulas for the local tangent stiffness matrix of a plane frame

member are derived by differentiating the member resistance vector in the displaced

position. This approach facilitates an analysis using only one element per member. The

formulas are checked by finite difference. The derivation leads to the familiar elastic and

geometric stiffness matrices used by other authors plus an additional higher order

geometric stiffness matrix.

Contributions of each of the three sub-matrices to the tangent stiffness matrix are

studied on both the member and structure levels through two numerical examples. These

same examples are analyzed three different ways for comparison. First, the examples are

analyzed using the method presented in this thesis. Second, they are analyzed with the

finite element modeling software ABAQUS/CAE using only one element per member.

Third, they are analyzed with ABAQUS using 200 elements per member. Comparisons
are made assuming the ABAQUS analysis which uses 200 elements per member is the

most accurate. The element presented in this thesis performs much better than the

ABAQUS analysis which uses one element per member, with maximum errors of 1.0%

and 40.8% respectively, for a cantilever column example. The maximum error for the

two story frame example using the ABAQUS analysis with one element per member is

42.8%, while the results from the analysis using the element presented in this thesis are

within 1.5%. Using the element presented in this thesis with only one element per

member gives good and computationally efficient results for second-order analysis.
 ACKNOWLEDGMENTS

The work presented in this thesis has resulted from the efforts of several

individuals. I would like to acknowledge my advisor, Dr. Richard J. Balling, for his

previous work on this subject, his encouragement and suggestions. I would also like to

thank my committee members for their time and effort in assisting me in this work.

Finally, I would like to thank my wife, Kristi, for her constant love and support, which

enabled me to complete this thesis.

 TABLE OF CONTENTS

1
U U Introduction .............................................................................................................. 1
U U

2
U U Literature Review ..................................................................................................... 5
U U

3
U U Derivation of the Member Tangent Stiffness Matrix .......................................... 11
U U

3.1
U U The Member Resistance Vector ....................................................................... 11
U U

3.2
U U Derivation of the Member Tangent Stiffness Matrix ....................................... 16
U U

4
U U Numerical Examples .............................................................................................. 25
U U

4.1
U U Example: Cantilever Column ........................................................................... 25
U U

4.1.1
U U Comparison with ABAQUS/CAE ................................................................ 26
U U

4.1.2
U U Validation of the Derivation of the Local Member Tangent Stiffness
U

Matrix ........................................................................................................... 28
U

4.1.3
U U Contributions of Sub-Matrices to the Tangent Stiffness Matrix .................. 29
U U

4.2
U U Example: Two-Story Plane Frame ................................................................... 33
U U

4.2.1
U U Comparison with ABAQUS/CAE ................................................................ 34
U U

4.2.2
U U Contributions of the Sub-Matrix to the Tangent Stiffness Matrix ............... 36
U U

5
U U Conclusions ............................................................................................................. 39
U U

References ....................................................................................................................... 41
U U

Appendix A
U U Member End Moments and Axial Force .......................................... 45
U U

A.1
U U U Derivation of the Member End Moments ........................................................ 45
U

A.2
U U U Derivation of the Axial Force .......................................................................... 50
U

Appendix B
U U ABAQUS Element B21 ...................................................................... 55
U U

vii
viii
 LIST OF TABLES

Table 4-1 Results of Three Analyses ....................................................................................26

U U

Table 4-2 Comparison to the 200 Element ABAQUS Analysis ...........................................26

U U

Table 4-3 Displacements from the Analysis on the Two-Story Plane Frame .......................34
U U

Table 4-4 Results from the ABAQUS Analysis with a Single Element per Member ...........34
U U

Table 4-5 Results from the ABAQUS Analysis with 200 Elements per Member ................35
U U

Table 4-6 Error between the Present Analysis and the 200 Element per Member
U

ABAQUS Analysis ....................................................................................................35

U

Table 4-7 Error between the Two ABAQUS Analyses .........................................................36

U U

ix
x
 LIST OF FIGURES

Figure 3-1 A Plane Frame Member .......................................................................................11

U U

Figure 3-2 Geometric Properties of a Plane Frame Member ...............................................13

U U

Figure 4-1 Cantilever Column ...............................................................................................25

U U

Figure 4-2 Convergence Study of ABAQUS Analysis .........................................................27

U U

Figure 4-3 Contribution Percentages for Variable Horizontal Displacement .......................30

U U

Figure 4-4 Contribution Percentages for Variable Vertical Displacement ...........................30

U U

Figure 4-5 Contribution Percentages for Variable Rotation .................................................31

U U

Figure 4-6 Contribution Percentages for Variable Displacements and Rotations .................31
U U

Figure 4-7 Contribution Percentages to the Structure Tangent Stiffness Matrix ..................32
U U

Figure 4-8 Two-Story Plane Frame .......................................................................................33

U U

Figure 4-9 Contributions to the Structure Tangent Stiffness Matrix .....................................37

U U

Figure A-1 An Infinitesimal Section of a Plane Frame Member ..........................................45

U U

Figure A-2 A Plane Frame Member .....................................................................................47

U U

Figure A-3 A Plane Frame Member .....................................................................................51

U U

xi
xii
1 Introduction
0B

The first objective of this thesis is to present a second-order nonlinear plane frame

analysis method which uses only one element per member. Many of the procedures

currently available employ finite element techniques which involve dividing the members

of the frame into multiple elements for a more accurate analysis. Analyzing large frames

with such an approach can be computationally intensive. The analysis presented herein

uses a single element per member.

The finite element modeling software ABAQUS/CAE is used to illustrate the

importance of this objective in Section 4 of this thesis. The results from ABAQUS

models using only one element per member are compared to the results of ABAQUS

models which use 200 elements per member. The resulting joint displacements are very

different, with errors as large as 41% for a cantilever column example and 43% for a two-

story plane frame example.

The second objective is to derive a tangent stiffness matrix for the plane frame

element that is the direct derivative of the resistance vector. This relationship stems from

the Newton-Raphson method used in the nonlinear matrix stiffness method. To illustrate

the importance of this direct derivation, a brief summary of the Newton-Raphson method

follows.

1
 For a second-order analysis, the governing equation of the matrix stiffness method

is shown in Equation 1-1.

f = z (u ) (1-1)

where f is the force vector of the element, and z is known as the resistance vector, a

nonlinear function of the displacement vector u. To solve this system of nonlinear

equations, an iterative solution method is used, namely the Newton-Raphson method

(Cheney and Kincaid 2008). The procedure begins with an initial guess for u,

u ( 0) = (u1( 0) , u 2( 0) , u 3( 0) ,..., u i( 0) ) T (1-2)

which is an approximate solution to Equation 1-1. Let Δu be a correction vector for the

initial displacement vector, u(0), such that

u ( 0) + Δu = (u1( 0) + Δu1 , u 2( 0) + Δu 2 , u 3( 0)
(1-3)
+ Δu 3 ,..., u i( 0) + Δu i ) T

is a more accurate solution. The Taylor expansion leads to the following equation

f ≈ z (u ( 0) + Δu ) ≈ z (u ( 0) ) + kT Δu (1-4)

where kT is the Jacobian matrix, which in structural analysis is the tangent stiffness

matrix. It is defined by

⎡ ∂z1 ∂z1 ∂z1 ∂z1 ⎤

⎢ K ⎥
⎢ ∂u1 ∂u 2 ∂u3 ∂u ⎥
i
⎢ ∂z ∂z 2 ∂z 2 ⎥
⎢ 2 ⎥
⎢ ∂u1 ∂u 2 ∂u3 ⎥
k T = z ′(u ) = ⎢ ∂z 3
(0)
∂z3 ∂z3 ⎥ (1-5)
⎢ ⎥
⎢ ∂u ∂u 2 ∂u3 ⎥
⎢ 1 ⎥
⎢ M O ⎥
⎢ ∂zi ∂z ⎥
i⎥
⎢
⎢⎣ ∂u1 ∂u ⎥
i⎦

2
Note that the tangent stiffness matrix is the derivative of the resistance vector with

respect to each of the displacements. If kT is nonsingular, then the corrected

displacements can be found by solving Equation 1-4 for Δu.

The derivation of the tangent stiffness matrix presented in section 3 of this thesis

is the work of Dr. Richard J. Balling. The remainder of this thesis, including an extensive

literature review in Section 2, the analysis of two examples in Section 4 and their

comparisons to ABAQUS analyses, the validation of the derivation of the tangent

stiffness matrix by finite difference and the study of the contributions of the three sub-

matrices to the tangent stiffness matrix is the work of Jesse W. Lyon.

3
4
2 Literature Review
1B

In the literature, a variety of methods exist for the derivation of the element

tangent stiffness matrix. However, many of these methods are not sufficiently accurate

for the analysis of long slender members without the use of multiple elements. Also, a

majority of these tangent stiffness matrices are not derived from the relationship between

the resistance vector and the tangent stiffness matrix.

Typically, the tangent stiffness matrix is the sum of two matrices, the elastic and

geometric stiffness matrices. The elastic stiffness matrix accounts for the effects of the

elastic deformations of the element, and is generally the same throughout the literature.

The geometric stiffness matrix deals with the effects of large displacements on the

element, and varies depending on the derivation.

The conventional method of deriving the tangent stiffness matrix uses the

principle of virtual work or virtual displacements, and the general or simplified beam

theory equations are presented in Yang and Kuo (1994). While analyzing curved beams,

Yang et al. (2007) summarized the conventional method. Their derivation of the stiffness

matrix for a two-dimensional beam element is applicable to this thesis. The incremental

equilibrium for the response of an element to externally applied loads is derived from the

principle of virtual work. Then using the finite element procedure, an incremental

stiffness equation for the element is obtained, which includes both the elastic and

5
geometric stiffness matrices. This approach has also been employed by Yang and Chiou

(1987), Zhao and Wong (2006), and Leu et al. (2008).

Yang and McGuire (1986) presented a similar derivation. They modified the

incremental stiffness equation by including a nonuniform torsional force in the element

force vector. This leads to a more complex tangent stiffness matrix.

Yang et al. (2002) developed the elastic and geometric stiffness matrices from

force-displacement relationships derived using incremental theory. The deformation of a

beam element due to externally applied loads is described as being a series of small

incremental translations and rotations. These displacements are calculated for each step

with reference to the previous step; however, rigid body motion is excluded. After

deriving the element forces and moments, the elastic and geometric stiffness matrices are

calculated. This procedure produces a tangent stiffness matrix that is identical to the

conventional stiffness matrix presented by Yang and Kuo (1994).

Izzudin (2006) used a rotational spring analogy to determine the geometric

stiffness of both a two-dimensional and three-dimensional frame element. The stiffness

of a beam element is related to the stiffness of a rotational spring attached to the element

and the element chord. The stiffness of the spring is calculated using simple linear

kinematics, which leads to a linear tangent stiffness matrix. This method is conceptually

easy to understand and can be modified such that only a single member is required for a

simplified linear analysis. This approach does not produce a tangent stiffness matrix that

is the direct derivative of the element resistance vector.

Oran (1973a) proposed a derivation of the tangent stiffness matrix using a

differential relationship between the stiffness matrix and the element force vector similar

6
to the one used in the Newton-Raphson method. The member force vector presented by

Oran (1973a) differs from the member resistance vector used in this thesis in that it

contains only three elements instead of six elements. As a result, Oran’s member tangent

stiffness matrix is a 3x3 elastic stiffness matrix. Oran (1973b) extended the previous

derivation to three dimensions. Chandra et al. (1990) used the same geometric stiffness

matrix as Oran (1973b) in their analysis method.

Gu and Chan (2000) derived the stiffness matrix of an element which included the

effects of initial imperfections. The equations of equilibrium which incorporate the

imperfections are constructed for the element in its deformed configuration. Force-

deflection equations are then derived from the equilibrium equations, and the tangent

stiffness matrix of the element is computed using a derivative approach similar to Oran

(1973a). The analysis they presented requires only one element per member. Gu and

Chan (2005) extended this derivation to space frames.

The inclusion of initial imperfections in the derivation was also explored by Chan

and Zhou (1995). The deformed shape of the member is assumed to be a fifth-order

polynomial. Boundary conditions are applied that incorporate the initial imperfection in

order to obtain the shape function for the element, which is used to derive the energy

function. The tangent stiffness matrix is formed by a differential equation involving the

total potential energy function. Once again, this approach does not take advantage of the

relationship between the resistance vector and the tangent stiffness matrix.

Albermani and Kitipornchai (1990) presented a tangent stiffness matrix that was

comprised of a deformation stiffness matrix in addition to the conventional elastic and

geometric stiffness matrices. An equation for the strain energy in terms of the axial strain

7
is derived using the principle of virtual displacements. Using a Lagrange interpolation

polynomial to approximate the displacement field, the strain energy equation is integrated

over the cross sectional area of the element. The resulting equation includes the

deformation stiffness matrix, which is a function of the element deformations.

Zhou and Chan (2004a) derived the secant and tangent stiffness matrices which

incorporate the formation of a plastic hinge along the length of the element. A

displacement function is derived as the superposition of an elastic deformation function

and a triangular plastic deformation function. Then, using the total potential energy

function modified to account for the plastic hinges, the member forces are developed.

The tangent stiffness matrix is derived by differentiating the force vector with respect to

the member displacements, similar to Oran (1973a). Zhou and Chan (2004b) extended

this derivation to incorporate three plastic hinges per member. The formation of plastic

hinges allows an analysis that uses only a single element per member.

Haktanir (1994) developed a tangent stiffness matrix for planar bars using force-

displacement relationships and the method of complementary functions. Equations for

the stiffness of the element, along with the constitutive and compatibility equations are

used to derive a set of governing differential equations for the behavior of the element.

These equations are used to calculate the member end forces. From the equation relating

the member displacements to the member end forces, the element tangent stiffness matrix

is derived by the method of complementary functions.

Moran et al. (1998) proposed a derivation of the secant and tangent stiffness

matrices using strain energy. First, the total strain energy is calculated by integrating the

function for the strain energy density within the element volume. An equation for the

8
total potential energy is obtained by using the principle of virtual work. Then, by

employing the finite element procedure, symmetric expressions for the generalized and

simplified secant and tangent stiffness matrices are derived.

So and Chan (1991) presented a tangent stiffness matrix for a modified Hermite

cubic beam element. First, a shape function for the element is derived using the

assumption that the deflected shape of the element is a fourth-order polynomial. Then,

from the total potential energy function of the element, the tangent stiffness matrix is

calculated.

The element derived in this thesis allows for an analysis using only a single

element per member. The tangent stiffness matrix for this element is the direct derivative

of the resistance vector, and as such can be incorporated directly into the Newton-

Raphson method. In order to achieve the objectives of this thesis, certain simplifications

were made in the derivation of the tangent stiffness matrix. For example, the governing

constitutive equation for the second-order Euler beam theory used is

dθ
M = EI
dx

where the derivative of the slope is with respect to the local coordinate axis, x which runs

longitudinally along the undeformed member. Shoup and McLarnan (1971) presented a

more accurate equation with reference to the arc length of the element in the deformed

configuration, as follows

dθ
M = EI
ds

Though this equation leads to a more complicated solution to the differential equations

within the analysis, it deserves future investigation.

9
10
3 Derivation of the Member Tangent Stiffness Matrix
2B

The member tangent stiffness matrix is the direct derivative of the member

resistance vector with respect to the member displacements. In this chapter a formula

for the member resistance vector as a function of the member displacements will first be

derived. The derivation of the member tangent stiffness matrix, involving the

differentiation of the resistance vector will follow.

3.1 The Member Resistance Vector

8B

The member resistance vector is the set of forces required to hold the member in a

displaced position. Consider the single plane frame member shown in Figure 3-1 . The
X X

dashed line represents the deformed configuration of the member.

M2
θ1
M1 F
θ2

x
L0

Figure 3-1 A Plane Frame Member

11
Expressions for the member end moments, M1 and M2 are obtained using the governing

differential equation of second-order Euler beam theory,

d 4Δ d 2Δ
EI − F =0 (3-1)
dx 4 dx 2

where E is the modulus of elasticity of the member, I is the moment of inertia of the cross

section, F is the axial force in the member and Δ is the equation for the deformed

configuration of the member in terms of x. The solution to Equation 3-1 can be written in

terms of the nodal rotations, θ1 and θ2,

⎛ 4EI ⎞ ⎛ 2EI ⎞
M1 = ⎜⎜ + FL 0 A G ⎟⎟θ1 + ⎜⎜ + FL 0 B G ⎟⎟θ 2
⎝ L0 ⎠ ⎝ L0 ⎠
(3-2)
⎛ 2EI ⎞ ⎛ 4EI ⎞
M 2 = ⎜⎜ + FL 0 B G ⎟⎟θ1 + ⎜⎜ + FL 0 C G ⎟⎟θ 2
⎝ L0 ⎠ ⎝ L0 ⎠

where L0 is the undeformed length of the member. For a more thorough derivation of

Equation 3-2 refer to Appendix A. AG, BG and CG are the geometric coefficients and are

dependent on the sign of the axial force. For F < 0

sin λ − λ cos λ 4
AG = CG = − + 2
λ (2 − 2 cos λ − λ sin λ ) λ
(3-3)
λ − sin λ 2
BG = − + 2
λ (2 − 2 cos λ − λ sin λ ) λ

where

−F
λ = L0
EI

12
If F > 0, the geometric coefficients are

λ cosh λ − sinh λ 4
AG = CG = − 2
λ (2 − 2 cosh λ − λ sinh λ ) λ
(3-4)
sin λ − λ 2
BG = − − 2
λ (2 − 2 cosh λ − λ sinh λ ) λ

where

F
λ = L0
EI

As λ goes to zero, both Equations 3-3 and 3-4 degenerate to

2 1
AG = CG = BG = − (3-5)
15 30

The rotation angles θ1 and θ2 in Equation 3-2 can also be written in terms of the

member displacements with the use of Figure 3-2 , which shows the geometric properties
X X

of a plane frame member for both the original and deformed positions.

θ2
u6
u5
α
u3 B
θ1 u4
α0
u2 x2

A u
1

x1

Figure 3-2 Geometric Properties of a Plane Frame Member

13
The member in its deformed position is represented by the dashed line. The initial

position of the member is represented by the solid straight line between points A and B.

The coordinates of A and B in the global coordinate system are (x1A, x2A) and (x1B, x2B),

respectively. Note that translational displacements u1 and u4 are parallel to the global x1

axis, and translational displacements u2 and u5 are parallel to the global x2 axis. Note that

rotational displacements u3 and u6 are measured from lines parallel to the initial line

between A and B, while rotations θ1 and θ2 are measured from a straight line between the

displaced ends of the member. The original length of the member is given by the

following equation

L0 = ( x1B − x1A ) 2 + ( x 2B − x 2A ) 2 (3-6)

The deformed length of the member due to the nodal translations is defined as follows

L = ( x1B + u 4 − x1A − u1 ) 2 + ( x 2B + u 5 − x 2A − u 2 ) 2 (3-7)

The nodal rotations measured from the deformed reference line are

θ1 = u 3 − (α − α 0 ) and θ 2 = u 6 − (α − α 0 ) (3-8)

where

⎛ x 2B + u 5 − x 2A − u 2 ⎞ ⎛ x B − x 2A ⎞
α = sin ⎜⎜
−1
⎟ and α 0 = sin −1 ⎜⎜ 2
⎟ ⎟⎟ (3-9)
⎝ L ⎠ ⎝ L 0 ⎠

14
The axial force, F is given by the following equation

EA
F= (L − L0 ) (3-10)
L0

Using simple statics, the member resistance vector can then be written as

⎡ −F ⎤
⎢ (M 1 + M 2 ) ⎥
⎢ L ⎥
⎢ M1 ⎥
zˆ = ⎢ ⎥ (3-11)
⎢ F ⎥
⎢ (M 1 + M 2 ) ⎥
⎢− L ⎥
⎢ M2 ⎥
⎣ ⎦

The ^ above z signifies that the end forces of magnitude F and (M1+M2)/L in Equation 3-

11 are parallel to the axes of the local member coordinate system. The local x̂ 1 axis runs

from the displaced position of A to the displaced position of B, and the local x̂ 2 axis is

normal to the local x̂ 1 axis. A transformation matrix is used to convert the member

stiffness matrices and resistance vectors from the local coordinate system of each

member to the global coordinate system, so that they are all compatible with one another.

The orthogonal transformation matrix for plane frames is defined as follows

⎛ cos α sin α 0 0 0 0⎞
⎜ ⎟
⎜ − sin α cos α 0 0 0 0⎟
⎜ 0 0 1 0 0 0⎟
T =⎜ ⎟ (3-12)
⎜ 0 0 0 cos α sin α 0⎟
⎜ 0
⎜ 0 0 − sin α cos α 0 ⎟⎟
⎜ 0
⎝ 0 0 0 0 1 ⎟⎠

15
where

x1B + u 4 − x1A − u1
cos α =
L
(3-13)
x 2B + u5 − x 2A − u2
sin α =
L

The global member resistance vector, written as a transformation of the local member

resistance vector is

z = Tzˆ (3-14)

3.2 9BDerivation of the Member Tangent Stiffness Matrix

The global member tangent stiffness matrix is the derivative of the global member

resistance vector. However, for computation within the matrix stiffness method, the local

tangent stiffness matrix is used. The local tangent stiffness matrix is written as

kˆT = Tk T T T (3-15)

Knowing the relationship between the global resistance vector and the global tangent

stiffness matrix, and then substituting from Equation 3-14, Equation 3-15 becomes

⎡ ∂z ∂z ∂z ∂z ∂z ∂z ⎤ T
kˆT = T ⎢ ⎥T
⎣ ∂u1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6 ⎦
(3-16)
⎡ ∂ (Tzˆ ) ∂(Tzˆ ) ∂ (Tzˆ ) ∂ (Tzˆ ) ∂ (Tzˆ ) ∂(Tzˆ )⎤ T
=T⎢ ⎥T
⎣ ∂u1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6 ⎦

16
Using the product rule, Equation 3-16 becomes

k̂ T = T ⎢
( )
⎡∂ TT
ẑ
( )
∂ TT
ẑ
( )
∂ TT
ẑ
( )
∂ TT
ẑ
( )
∂ TT
ẑ
( )
∂ TT ⎤ T
ẑ ⎥ T
⎢⎣ ∂u 1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6 ⎥⎦
(3-17)
⎡ ∂ẑ ∂ẑ ∂ẑ ∂ẑ ∂ẑ ∂ẑ ⎤ T
+⎢ ⎥T
⎣ ∂u1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6 ⎦

In order to continue with the derivation, a few preliminary steps are performed.

First, the equation for the deformed length given in Equation 3-7 is differentiated with

respect to the member displacements. The following 6 equations constitute Equation

3-18.

∂L
= − cos α
∂u1
∂L
= − sin α
∂u 2
∂L
=0 (3-18)
∂u 3
∂L
= cos α
∂u 4
∂L
= sin α
∂u 5
∂L
=0
∂u 6

Second, the end rotations are differentiated with respect to each member displacement.

To do this, α and α0 are differentiated. Remember that the length, L, is a function of the

member displacements, so the derivatives of α and α0 involve the application of the chain

rule.

17
The following 7 equations constitute Equations 3-19 and 3-20.

∂α sin α
=
∂u1 L
∂α cos α
=−
∂u 2 L
∂α
=0 (3-19)
∂u 3
∂α sin α
=−
∂u 4 L
∂α cos α
=
∂u 5 L
∂α
=0
∂u 6

∂α 0 ∂α 0 ∂α 0 ∂α 0 ∂α 0 ∂α 0
= = = = = =0 (3-20)
∂u1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6

The following derivatives of θ1 and θ2 constitute Equation 3-21.

∂θ1 ∂θ 2 ∂α sin α
= =− =−
∂u1 ∂u1 ∂u1 L
∂θ1 ∂θ ∂α cos α
= 2 =− =
∂u 2 ∂u 2 ∂u 2 L
∂θ1
=1
∂u 3
∂θ 2
=0 (3-21)
∂u 3
∂θ 1 ∂θ 2 ∂α sin α
= =− =
∂u 4 ∂u 4 ∂u 4 L
∂θ1 ∂θ ∂α cos α
= 2 =− =−
∂u 5 ∂u 5 ∂u 5 L
∂θ1
=0
∂u 6
∂θ 2
=1
∂u 6

18
Third, the cosine and sine of α must be differentiated with respect to the different

displacements. The following equations constitute Equation 3-22.

∂ (cos α ) sin 2 α
=−
∂u1 L
∂ (sin α ) cos α sin α
=
∂u1 L
∂ (cos α ) cos α sin α
=
∂u 2 L
∂ (sin α ) cos 2 α
=−
∂u 2 L
∂ (cos α ) ∂ (sin α )
= =0 (3-22)
∂u 3 ∂u 3
∂ (cos α ) sin 2 α
=
∂u 4 L
∂ (sin α ) cos α sin α
=−
∂u 4 L
∂ (cos α ) cos α sin α
=−
∂u 5 L
∂ (sin α ) cos 2 α
=
∂u 5 L
∂ (cos α ) ∂ (sin α )
= =0
∂u 6 ∂u 6

The derivative of the transformation matrix with respect to u1 can now be determined.

⎛ sin α − cos α 0 0 0 0⎞
⎜ ⎟
⎜ cos α sin α 0 0 0 0⎟
∂T sin α ⎜⎜ 0 0 0 0 0 0⎟
⎟ = − S sin α
=−
∂u1 L ⎜ 0 0 0 sin α − cos α 0⎟ L
⎜ 0 0 0 cos α sin α 0⎟ ⎟
⎜
⎜ 0 0 0 0 0 0 ⎟⎠
⎝

19
where

⎛ sin α − cos α 0 0 0 0⎞
⎜ ⎟
⎜ cos α sin α 0 0 0 0⎟
⎜ 0 0 0 0 0 0⎟
S =⎜ ⎟
⎜ 0 0 0 sin α − cos α 0⎟
⎜ 0 0 0 cos α sin α 0 ⎟⎟
⎜
⎜ 0
⎝ 0 0 0 0 0 ⎟⎠

The remaining derivatives of the transformation matrix are similarly derived. The

preceding equation and the following 5 equations constitute Equation 3-23.

∂T cos α
=S
∂u 2 L
∂T
=0 (3-23)
∂u 3
∂T sin α
=S
∂u 4 L
∂T cos α
= −S
∂u 5 L
∂T
=0
∂u 6

Substituting from Equation 3-23, Equation 3-17 then becomes

L
1
[( ) (
kˆT = T − S T sin α zˆ S T cos α )zˆ 0 (S T
) ( ) ]
sin α zˆ − S T cos α zˆ 0 T T

⎡ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ⎤ T

+⎢ ⎥T
⎣ ∂u1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6 ⎦

1
kˆT = TS T zˆ[− sin α cos α 0 sin α − cos α 0]T T
L
(3-24)
⎡ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ⎤ T
+⎢ ⎥T
⎣ ∂u1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6 ⎦

20
After some matrix multiplication, Equation 3-24 becomes

⎡ zˆ 2 ⎤
⎢ − zˆ ⎥
⎢ 1⎥
1⎢ 0 ⎥ ⎡ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ⎤ T
kˆT = ⎢ ⎥[0 1 0 0 − 1 0] + ⎢ ⎥T
L ⎢ zˆ5 ⎥ ⎣ ∂u1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6 ⎦
⎢− zˆ 4 ⎥
⎢ ⎥
⎢⎣ 0 ⎥⎦

⎡0 zˆ 2 0 0 − zˆ 2 0⎤
⎢0 − zˆ 0 0 zˆ1 0⎥⎥
⎢ 1

⎢
1 0 0 0 0 0 0⎥
kˆT = ⎢ ⎥
L ⎢0 zˆ 5 0 0 − zˆ 5 0⎥
(3-25)
⎢0 − zˆ 4 0 0 zˆ 4 0⎥
⎢ ⎥
⎢⎣0 0 0 0 0 0⎥⎦
⎡ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ∂zˆ ⎤ T
+⎢ ⎥T
⎣ ∂u1 ∂u 2 ∂u 3 ∂u 4 ∂u 5 ∂u 6 ⎦

The local member resistance vector give in Equation 3-11 must be differentiated

with respect to each of the displacements, u1 through u6. Equation 3-11 is repeated

below.

⎡ −F ⎤
⎢ (M 1 + M 2 ) ⎥
⎢ L ⎥
⎢ M1 ⎥
zˆ = ⎢ ⎥ (3-11)
⎢ F ⎥
⎢ (M 1 + M 2 ) ⎥
⎢− L ⎥
⎢ M2 ⎥
⎣ ⎦

where M1 and M2 are given by Equation 3-2, and F is given by Equation 3-10. Each term

in the resistance vector is dependent on some or all of the member displacements. The

21
equations for the end moments and axial force are substituted into Equation 3-11 and the

geometric coefficients are assumed to be the constants given by Equation 3-5 in order to

find the derivatives of the resistance vector. Equation 3-11 and its derivatives, along with

Equation 3-12 are then substituted into Equation 3-25 and the local member tangent

stiffness matrix is determined to be

kˆT = kˆE + kˆG1 + kˆG 2 (3-26)

where

⎡ A A ⎤
⎢ I 0 0 − 0 0 ⎥
I
⎢ 12 6 12 6 ⎥
⎢ 0 0 − ⎥
⎢ L2 L L2 L ⎥
⎢ 6 6 ⎥
EI ⎢ 0 4 0 − 2 ⎥
kˆE = L L
L0 ⎢⎢− A A ⎥ (3-27)
0 0 0 0 ⎥
⎢ I I ⎥
⎢ 0 12 6 12 6⎥
− − 0 −
⎢ L2 L L2 L⎥
⎢ 6 6 ⎥
⎢ 0 2 0 − 4 ⎥
⎣ L L ⎦

⎡ ⎤
⎢0 0 0 0 0 0 ⎥
⎢ ⎥
⎢0 1 1 1 1 1 1 ⎥
+ 2 0 − − 2
⎢ LL0 5L 15L LL0 5L 15L ⎥
⎢ 1 2 1 1 ⎥
⎢0 0 − − ⎥
15L 15 15L 30 ⎥
kˆG1 = FL0 ⎢ (3-28)
⎢ ⎥
⎢0 0 0 0 0 0 ⎥
⎢ ⎥
⎢0 1 1 1 1 1 1 ⎥
− − 2 − 0 + 2 −
⎢ LL0 5 L 15L LL0 5 L 15 L ⎥
⎢ 1 1 1 2 ⎥
⎢0 − 0 − ⎥
⎢⎣ 15L 30 15L 15 ⎥⎦

22
 ⎡ ⎤
⎢ 0 D3 0 0 − D3 0⎥
⎢ ( D + D2 ) ( D1 + D2 ) ⎥
⎢ D3 − 1 0 0 − D3 + 0 0⎥
⎢ L L ⎥
⎢ ⎥
⎢ − D1 0 0 − D1 0 0⎥
kˆG 2 = EA⎢ ⎥ (3-29)
⎢ 0 − D3 0 0 D3 0⎥
⎢ ⎥
⎢− D + ( D1 + D2 ) 0 0 0 0 0⎥
⎢ 3
L ⎥
⎢ ⎥
⎢ D2 0 0 − D2 0 0⎥
⎣ ⎦

where

2θ1 θ 2 2θ 2 θ1 M 1 +M 2
D1 = − D2 = − D3 = (3-30)
15 30 15 30 EAL2

If L is set equal to L0 in Equations 3-27 and 3-28 then k̂ E and kˆG1 become the

elastic and geometric stiffness matrices commonly found in the literature (ex. Yang et. al.

2002, Zhao et. al. 2006). The third matrix, kˆG 2 , is a higher-order local member geometric

stiffness matrix. Note that kˆG 2 is not symmetric. This is because the axial force does not

take into account the lengthening of the curved shape in Figures 3-1 and 3-2 due to the

end rotations, θ1 and θ2. A cubic polynomial is used to approximate the deformed shape

of the member due to the end rotations in order to calculate the following expression for

the axial force

F=
EA
(L − L0 ) + EA⎛⎜ 1 θ12 − 1 θ1θ 2 + 1 θ 2 2 ⎞⎟ (3-31)
L0 ⎝ 15 30 15 ⎠

23
The derivation of the Equation 3-31 is given in Appendix A. The new equation for the

axial force is used to obtain the following symmetric kˆG 2 matrix

⎡ D1 + D 2 D1 + D 2 ⎤
D3 − − D1 − D3 + − D2
⎢ 0
L
0
L
⎥
⎢ ⎥
⎢ D − D1 + D 2 (
L 0 D1 + D 2
2
) (
L 0 D1 D1 + D 2 ) D1 + D 2 (
L 0 D1 + D 2
2
) L 0 D 2 ( D1 + D 2 )
⎥
− D3 + −
⎢ 3 L L
2
L L L
2
L ⎥
⎢ (
L 0 D1 D1 + D 2 ) 2 (
L 0 D1 D1 + D 2 ) ⎥ (3-32)
⎢ − D1 L 0 D1 D1 − L 0 D1 D 2 ⎥
kˆG 2 = EA ⎢ L L
⎥
D1 + D 2 D1 + D 2
⎢ 0 − D3 + D1 0 D3 − D1 ⎥
⎢ L L ⎥
⎢ D +D (
L 0 D1 + D 2 )2 (
L 0 D1 D1 + D 2 ) D1 + D 2 (
L 0 D1 + D 2
2
) L 0 D 2 ( D1 + D 2 )⎥
⎢ − D3 + 1 2 −
2
− D3 −
2
− ⎥
⎢ L L L L L L ⎥
⎢ (
L 0 D 2 D1 + D 2 ) (
L 0 D 2 D1 + D 2 ) 2 ⎥
− D2 L 0 D1 D 2 D1 − L0 D 2
⎣⎢ L L ⎦⎥

24
4 Numerical Examples
3B

4.1 Example: Cantilever Column

10B

Consider the cantilever column shown in Figure 4-1 , where L = 240 inches, P =
X X

400 kips and Q = 50 kips. A large vertical load was chosen in order to assure that the

p-delta effects were significant. The column has a modulus of elasticity equal to 29,000

ksi, a cross-sectional area equal to 100 in2, and a moment of inertia of 833.3 in4.

Figure 4-1 Cantilever Column

The cantilever column was analyzed using the method presented in this thesis,

which incorporates the elastic, geometric and higher-order geometric stiffness matrices in

the tangent stiffness matrix. The analysis yielded end displacements at the top of the

column of 15.3 inches horizontally, 0.62 inches vertically downward and a rotation of

0.098 radians.

25
4.1.1 Comparison with ABAQUS/CAE
14B

The same cantilever column example was modeled using the finite element

software ABAQUS/CAE for comparison. The column was first analyzed using only one

element. It was then divided into 200 elements and a second analysis was performed.

The beam element B21 was used for both ABAQUS analyses, which is explained in

Appendix B. Table 4-1 shows the results of this comparison. The ABAQUS analysis
X X

which used 200 elements was considered to be the most accurate of the three methods

used to analyze this example. The errors between the results from the ABAQUS analysis

with 200 elements and the other two analyses are shown in Table 4-2.

Table 4-1 Results of Three Analyses

ABAQUS/CAE ABAQUS/CAE
Present
Displacement with 1 element with 200 elements
Method
per member per member
Horizontal (in) 15.344922 12.7932 15.3914
Vertical (in) -0.624892 -0.374151 -0.631485
Rotation (rad) -0.097584 -0.0848857 -0.0977828

Table 4-2 Comparison to the 200 Element ABAQUS Analysis

ABAQUS/CAE
Present
Displacement with 1 element
Method
per member
Horizontal (in) 0.3% 16.9%
Vertical (in) 1.0% 40.8%
Rotation (rad) 0.2% 13.2%

26
 The analysis method presented in this thesis which used a single element to model

the cantilever column and the ABAQUS model which used 200 elements yielded very

similar results, with a maximum error of 1.0%. The ABAQUS analysis which used only

one element to model the column had an error of 16.9% for the horizontal displacement

and a maximum error of 40.8%.

A convergence study was done to determine the number of elements required for

ABAQUS to obtain results with the same accuracy as the results from the analysis which

used the element presented in this thesis. Figure 4-2 plots the horizontal displacement

calculated by ABAQUS as a function of the number of elements used to analyze the

column. Ten elements were required for the ABAQUS analysis to produce a similar

horizontal displacement as the method presented in this thesis. The convergence study

also verified the assumption that the ABAQUS analysis with 200 elements was accurate.

16

15
Horizontal Displacement (in) .

14

13

12

11
0 25 50 75 100 125 150 175 200
No. of Elements Used

Figure 4-2 Convergence Study of ABAQUS Analysis

27
4.1.2 Validation of the Derivation of the Local Member Tangent Stiffness Matrix
15B

The cantilever column example was also used to validate the derivation of the

tangent stiffness matrix presented in section 3. An initial displacement vector was input

into the analysis and the corresponding structure resistance vector was calculated. The

following initial displacement vector was used

u = [1 1]
Τ
1 (4-1)

where displacements are measured in inches and rotations are measured in radians.

Using the finite difference method the derivative of the resistance vector was determined

and compared to the original tangent stiffness matrix. The structure tangent stiffness

matrix that corresponds to the displacement vector in Equation 4-1 is

⎡1180.3276 1161.9492 63641.1123⎤

K Τ = ⎢⎢1161.9492 12073.878 387609.247 ⎥⎥ (4-2)
⎢⎣63641.112 387609.25 19466044.3 ⎥⎦

The structure tangent stiffness matrix computed using the finite difference method is

⎡ 1180.33 1161.95 63646.99 ⎤

K Τ = ⎢⎢ 1161.95 12073.88 387628.56 ⎥⎥ (4-3)
⎢⎣63641.22 387609.25 19467906.1⎥⎦

which constitutes a maximum error of 0.0096%. Such a small error suggests that the

stiffness matrix calculated in the analysis is a reasonable approximation to the true

derivative of the resistance vector.

28
4.1.3 16B Contributions of Sub-Matrices to the Tangent Stiffness Matrix

Equation 3-26 shows that the local member tangent stiffness matrix is the sum of

three sub-matrices; k̂ E =the elastic, k̂ G1 =geometric and k̂ G 2 =higher-order geometric

stiffness matrices. Figures 4-3, 4-4, and 4-5 show the contributions of these three sub-

matrices to the local member tangent stiffness matrix in the first Newton-Raphson

iteration as a function of the horizontal displacement, vertical displacement, and rotation,

respectively. The contribution percentages were calculated element by element and then

averaged over all the elements in the matrix.

Figure 4-6 shows the contribution percentages of the three sub-matrices to the

local member tangent stiffness matrix as the horizontal and rotational displacements are

simultaneously increased. The vertical displacement was set equal to zero. The

following ratio was used to determine the rotation

θ 3
= (4-4)
Δ 2L

where Δ is the horizontal displacement and L is the height of the column. Equation 4-4 is

the same equation used in first-order linear analysis to find the rotation at the end of a

cantilever beam.

29
 100

90

80

70
Contribution (%) .

60

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100
Horizontal Displacement (in)

Elastic Matrix Geometric Matrix Higher-Order Geometric Matrix

Figure 4-3 Contribution Percentages for Variable Horizontal Displacement

100

90

80

70
Contribution (%) .

60

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100
Vertical Displacement (in)

Elastic Matrix Geometric Matrix Higher-Order Geometric Matrix

Figure 4-4 Contribution Percentages for Variable Vertical Displacement

30
 100

90

80

70
Contribution (%) .

60

50

40

30

20

10

0
0 0.1 0.2 0.3 0.4 0.5 0.6
Rotation (rad)

Elastic Matrix Geometric Matrix Higher-Order Geometric Matrix

Figure 4-5 Contribution Percentages for Variable Rotation

100

90

80

70
Contribution (%) .

60

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100
Horizontal Displacement (in)

Elastic Matrix Geometric Matrix Higher-Order Geometric Matrix

Figure 4-6 Contribution Percentages for Variable Displacements and Rotations

31
 Figures 4-3 through 4-6 show that the elastic stiffness matrix constitutes most of

the element’s response when the displacements are small. The contributions of the

geometric stiffness matrices increase as the deformations become larger. Figure 4-4

shows that the contribution of the higher-order geometric stiffness matrix is not affected

by the vertical displacement.

The same procedure was used to obtain Figure 4-6, which shows the contributions

of the sub-matrices to the structure tangent stiffness matrix. The horizontal and rotational

displacements were increased simultaneously using Equation 4-4. The vertical

displacement was set to zero. As Figure 4-7 demonstrates, the contributions from the
X X

geometric stiffness matrices to the structure tangent stiffness matrix increase as the

deformations become larger.

100

90
80

70
Contribution (%) .

60
50

40
30

20
10
0
0 10 20 30 40 50 60 70 80 90 100
Horizontal Displacement (in)

Elastic Matrix Geometric Matrix Higher-Order Geometric Matrix

Figure 4-7 Contribution Percentages to the Structure Tangent Stiffness Matrix

32
4.2 Example: Two-Story Plane Frame
11B

Consider the two-story plane frame shown in Figure 4-8 , where L1 = 120 inches
X X

and L2=144 inches. The forces, P = 400 kips and Q = 50 kips are applied to joints 3

through 6. The letter J with a subscript is used to label the joints.

P P

Q
J5 J6
Q
L2
P P
J3 J4
Q
Q
L2

J1 J2

L1

Figure 4-8 Two-Story Plane Frame

The columns have a cross-sectional area equal to 100 in2, and a moment of inertia of

833.3 in4. The girders have a cross-sectional area equal to 25 in2, and a moment of inertia

of 52.1 in4. All of the members have a modulus of elasticity equal to 29,000 ksi.

An analysis was performed which incorporated the elastic, geometric and higher-

order geometric stiffness matrices into the tangent stiffness matrix. The resulting joint

displacements of the analysis are shown in Table 4-3 .

X X

33
 Table 4-3 Displacements from the Analysis on the Two-Story Plane Frame

Horizontal Vertical
Rotation
Joint Displacement Displacement
(rad)
(IN) (IN)
J1 0.0 0.0 0.0
J2 0.0 0.0 0.0
J3 6.317 -0.199 -0.06800
J4 6.263 -0.213 -0.06787
J5 17.105 -0.619 -0.07054
J6 17.047 -0.642 -0.07057

4.2.1 Comparison with ABAQUS/CAE

17B

As was done with the cantilever column example, this same problem was modeled

using the finite element software ABAQUS/CAE for comparison. Table 4-4 shows the

results of the ABAQUS analysis which used only a single element per member. Table

4-5 shows the results of the ABAQUS analysis which used 200 elements per member.

Once again, both ABAQUS analyses were done using the B21 element explained in

Appendix B.

Table 4-4 Results from the ABAQUS Analysis with a Single Element per Member

Horizontal Vertical
Rotation
Joint Displacement Displacement
(rad)
(IN) (IN)
J1 0.0 0.0 0.0
J2 0.0 0.0 0.0
J3 4.927 -0.115 -0.05427
J4 4.926 -0.133 -0.05427
J5 13.315 -0.375 -0.05416
J6 13.315 -0.401 -0.05417

34
 Table 4-5 Results from the ABAQUS Analysis with 200 Elements per Member

Horizontal Vertical
Rotation
Joint Displacement Displacement
(rad)
(IN) (IN)
J1 0.0 0.0 0.0
J2 0.0 0.0 0.0
J3 6.362 -0.201 -0.06831
J4 6.308 -0.216 -0.06818
J5 17.215 -0.627 -0.07094
J6 17.156 -0.650 -0.07096

The comparison of the analysis using the method presented in this thesis and the

ABAQUS analysis which used 200 elements per member is given in Table 4-6 . The
X X

comparison between the ABAQUS analysis which used a single element per member and

the ABAQUS analysis which used 200 elements per member is given in Table 4-7.

Table 4-6 Error between the Present Analysis and the 200 Element per Member ABAQUS Analysis

Horizontal Vertical
Joint Rotation
Displacement Displacement
J3 0.7% 1.4% 0.5%
J4 0.7% 1.4% 0.5%
J5 0.6% 1.2% 0.6%
J6 0.6% 1.2% 0.5%

35
 Table 4-7 Error between the Two ABAQUS Analyses

Horizontal Vertical
Joint Rotation
Displacement Displacement
J3 22.6% 42.8% 20.6%
J4 21.9% 38.7% 20.4%
J5 22.7% 40.2% 23.6%
J6 22.4% 38.3% 23.7%

The analysis using the method presented in this thesis and the ABAQUS analysis

which used 200 elements per member yielded very similar results, with a maximum error

of 1.4%. Once again, the method presented in this thesis gave reasonable results with the

use of only one element per member. The maximum errors between the two ABAQUS

analyses were 22.7% for the horizontal displacements and 42.8% for the vertical

displacements.

4.2.2 Contributions of the Sub-Matrix to the Tangent Stiffness Matrix

18B

Figure 4-9 plots the contributions of the three sub-matrices to the structure tangent

stiffness matrix for the two-story plane frame example. The same procedure was used to

generate Figure 4-9 as was used in the cantilever column example. The relative

horizontal displacements and the rotations were incremented simultaneously for each

story using the relationship given in Equation 4-4. The vertical displacements were set to

zero.

36
 100

90

80

70
Contribution (%) .

60

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100
Horizontal Displacement (in)

Elastic Matrix Geometric Matrix Higher-Order Geometric Matrix

Figure 4-9 Contributions to the Structure Tangent Stiffness Matrix

Figure 4-9 shows that the elastic stiffness matrix constitutes most of the structure’s
X X

response when the displacements are small. The two geometric stiffness matrices

become more important in the analysis as the deformations become larger.

37
38
5 Conclusions
4B

The first objective of this thesis was to present a second-order nonlinear plane

frame analysis method which uses only one element per member. The second objective

was to derive a tangent stiffness matrix for the plane frame element that is the direct

derivative of the resistance vector. Both objectives have been achieved. Formulas for the

local tangent stiffness matrix have been successfully derived by differentiating the

member resistance vector in the displaced position which facilitates an analysis using

only one element per member. These formulas have been checked by finite difference.

The derivation led to the familiar elastic and geometric stiffness matrices used by other

authors with an additional higher order geometric stiffness matrix.

Contributions of each of the sub-matrices to the tangent stiffness matrix were

studied on both the member and structure levels through two numerical examples. At

small deformations, the elastic stiffness matrix contributed the most to the tangent

stiffness matrix. The contributions from the geometric stiffness matrices increased as the

displacements became larger.

A cantilever column example and a two-story plane frame example were analyzed

three different ways for comparison. First, each example was analyzed using the method

presented in this thesis. Second, each example was analyzed with the finite element

modeling software ABAQUS/CAE using only one element per member. A third analysis

39
was performed using ABAQUS/CAE with 200 elements per member. The ABAQUS

analysis which used 200 elements per member was assumed to be the most accurate.

Comparisons were made between it and the other two analyses. The ABAQUS analysis

which used one element per member had maximum errors of 41% for the cantilever

column example and 43% for the two-story plane frame example. The element presented

in this thesis performed much better, with maximum errors of only 1% for the cantilever

column example and 1.4% for the two-story plane frame example.

There are, however, limitations to the application of this element in structural

analysis. The derivation of the tangent stiffness matrix neglects the shear deformation of

the members and assumes that the material remains elastic. Furthermore, this element

only applies to plane frames and has not been extended to three-dimensional frames.

Despite these limitations, using the element presented in this thesis with only one element

per member gives good and computationally efficient results for second-order analysis.

40
 References
5B

Albermani, F. G. A., S. Kitipornchai. “Nonlinear-Analysis of Thin-Walled Structures

Using Least Element Member.” Journal of Structural Engineering-ASCE, 116(1)
(Jan. 1990): 215-234.

Chan, S. L., Z. H. Zhou. “Second-Order Elastic Analysis of Frames Using Single

Imperfect Element per Member.” Journal of Structural Engineering-ASCE,
121(6) (Jun. 1995): 939-945.

Chandra, R., D. N. Trikha, P. Krishna. “Nonlinear analysis of steel space frames.”

Journal of Structural Engineering-ASCE, 116(4) (Apr. 1990): 898-909.

Cheney, W., D. Kincaid. Numerical Mathmatics and Computing, Sixth edition. Belmont,
C.A.: Thomson Brooks/Cole, 2008.

“Choosing a beam element.” ABAQUS Analysis User’s Manual, Version 6.8.

Providence, R.I.: Simulia, 2008.

Gu, J., S. Chan. “Exact tangent stiffness for imperfect beam-column members.” Journal
of Structural Engineering, 126 (Sep. 2000): 1094-1102.

Gu, J., S. Chan. “Tangent stiffness matrix for geometric nonlinear analysis of space
frames.” Journal of Southeast University (English Addition), 21(4) (Dec. 2005):
480-485.

Haktanir, V. “New method for element stiffness matrix of arbitrary planar bars.”
Computers and Structures, 52(4) (Aug. 1994): 679-691.

Izzuddin, B. A. “Simplified buckling analysis of skeletal structures.” Proceedings of the

Institute of Civil Engineers-Structures and Buildings, 159(6) (Aug. 2006): 631-
651.

41
Leu, L. J., J. P. Yang, M. H. Tsai, Y. B. Yang. “Explicit Inelastic Stiffness for Beam
Elements with Uniform and Nonuniform Cross Sections.” Journal of Structural
Engineering, 134(4) (April 2008): 608-618.

Morán, A., E. Oñate, J. Miquel. “A General Procedure for Deriving Symmetric

Expressions for the Secant and Tangent Stiffness Matrices in Finite Element
Analysis.” International Journal for Numerical Methods Engineering, 42 (1998):
219-236.

Oran, M. C. “Tangent Stiffness in Plane Frames.” Journal of Structural Division, 99(6)

(Jun. 1973a): 973-984.

Oran, M. C. “Tangent Stiffness in Space Frames.” Journal of Structural Division, 99(6)

(Jun. 1973b): 987-1001.

So, A. K. W., S. L. Chan. “Buckling and Geometrically Nonlinear Analysis of Frames

using One Element/Member.” Journal of Construction Steel Research, 20 (1991):
271-289.

Shoup, T. E., C. W. McLarnan. “On the use of the Undulating Elastica for the Analusis
of Flexible Link Mechanisms.” Journal of Engineering for Industry, 93(1) (Feb.
1971): 263-267.

Yang, Y. B., H. T. Chiou. “Rigid Body Motion Test for Nonlinear-Analysis with Beam
Elements.” Journal of Engineering Mechanics – ASCE, 113(9) (September
1987): 1404-1419.

Yang, Y. B., S. R. Kuo. Theory and Analysis of Nonlinear Framed Structures.

Englewood Cliffs, N.J.: Prentice-Hall, 1994.

Yang, Y. B., S. R. Kuo, Y. S. Wu. “Incrementally small-deformation theory for

nonlinear analysis of structural frames.” Engineering Structures, 24(6) (2002):
783-798.

Yang, Y. B., S. P. Lin, C. M. Wang. “Rigid element approach for deriving the geometric
stiffness of curved beams for use in buckling analysis.” Journal of Structural
Engineering – ASCE, 133(12) (Dec. 2007): 1762-1771.

42
Yang, Y.B., W. McGuire. “Stiffness Matrix for Geometric Nonlinear Analysis.” Journal
of Structural Engineering – ASCE, 112(4) (April 1986): 853-877.

Zielinski, A. P., F. Frey. “On linearization in nonlinear structural finite element

analysis.” Computers & Science, 79(8) (March 2001): 825-838.

Zhao, D. F., K. K. F. Wong. “New approach for seismic nonlinear analysis of inelastic
framed structures.” Journal of Engineering Mechanics, 132(9) (Sept. 2006):
959-966.

Zhou, Z. H., S. L. Chan. “Elastoplastic and Large Deflection Analysis of Steel Frames
by One Element per Member. I: One Hinge Along Member.” Journal of
Structural Engineering – ASCE, 130(4) (Apr. 2004a): 538-544.

Zhou, Z. H., S. L. Chan. “Elastoplastic and Large Deflection Analysis of Steel Frames
by One Element per Member. II: Three Hinges Along Member.” Journal of
Structural Engineering – ASCE, 130(4) (Apr. 2004b): 545-553.

43
44
Appendix A Member End Moments and Axial Force
6B

A.1
12B Derivation of the Member End Moments

The present derivation is used to determine the end moments of a plane frame

member given in Equation 3-2. Before beginning with the derivation, the governing

equation of second-order Euler theory must be derived. Consider the infinitesimal

section of a plane frame member in its deformed configuration as shown in Figure A-1 .
X X

w
M+dM

V F+dF
M
V+dV
F

Δ Δ+dΔ
dx

Figure A-1 An Infinitesimal Section of a Plane Frame Member

45
The displacements from the original position are Δ and Δ+dΔ at the left and right ends,

respectively. Setting the sum of the horizontal forces equal to zero yields the following

F + dF − F = dF = 0 (A-1)

By setting the summation of the vertical forces equal to zero, the following is determined

V + wdx − V − dV = wdx − dV = 0 (A-2)

A third equation is developed by setting the sum of the moments about the right end

equal to zero.

dx
M + dM − M − Vdx − F (Δ + dΔ − Δ) − wdx
2
dx 2
= dM − Vdx − FdΔ − w =0 (A-3)
2

By dividing Equations A-1 through A-3 by dx the following three equations are obtained

dF
=0 (A-4)
dx
dV
w− =0 (A-5)
dx
dM dΔ dx
−V − F −w =0 (A-6)
dx dx 2

In the limit as dx goes to zero Equation A-6 becomes

dM dΔ
−V − F =0 (A-7)
dx dx

46
Differentiating Equation A-7 and substituting from Equations A-4 and A-5

d 2M d 2Δ
− F =w (A-8)
dx 2 dx 2

Substituting the flexural constitutive equation into Equation A-8

d 4Δ d 2Δ
EI − F =w (A-9)
dx 4 dx 2

With Equation A-9, the governing equation of second-order Euler beam theory, the end

moments of a plane frame member can be derived.

Consider the skeletal member shown in Figure A-2 . This member may be part of a
X X

plane frame. The dotted line represents the deformed configuration of the member.

M2
M1 θ1
θ2 P

x
L

Figure A-2 A Plane Frame Member

The angles θ1 and θ2 are the end rotations at the left and right ends of the member,

respectively. M1 and M2 are the end moments of the left and right ends of the member,

47
respectively. For this derivation, the uniform load along the beam, w is set to zero, and

the axial force in the member is P. The governing differential equation becomes

d 4Δ d 2Δ
EI 4 − P 2 = 0 (A-10)
dx dx

The general solution to Equation A-10 is

λx λx
Δ( x) = a sin + b cos + cx + d (A-11)
L L

where

P
λ=L
EI

By differentiating Equation A-11, the following equations for the rotation and bending

moment are obtained

dΔ aλ λ x bλ λ x
θ ( x) = = cos − sin + c (A-12)
dx L L L L

d 2Δ λ2 ⎛ λx λx ⎞
M ( x) = EI 2
= − EI 2 ⎜
a sin + b cos ⎟ (A-13)
dx L ⎝ L L⎠

Applying the appropriate boundary conditions to Equations A-11 through A-13,

equations for the coefficients a and b are determined.

⎧a ⎫ L ⎡1 − cos λ − λ sin λ cos λ − 1⎤ ⎧θ1 ⎫

⎨ ⎬= ⎢ ⎨ ⎬
λ − sin λ ⎥⎦ ⎩θ 2 ⎭
(A-14)
⎩b ⎭ λ (2 − 2 cos λ − λ sin λ ) ⎣ sin λ − λ cos λ

48
The coefficients c and d can then be calculated using the values of a and b.

a b
c = − sin λ + (1 − cos λ ) ) (A-15)
L L

d = −b (A-16)

To obtain equations for the end moments of the member, two addition boundary

conditions are used.

At x = 0, M = M1:

λ2
− EI b = −M 1 (A-17)
L2

At x = L, M = M2:

λ2
− EI (a sin λ + b cos λ ) = M 2 (A-18)
L2

Equations A-17 and A-18 can be written in matrix form as follows

⎧M1 ⎫ λ2 ⎡ 0 1 ⎤ ⎧a ⎫
⎨ ⎬ = EI ⎢− sin λ ⎨ ⎬
− cos λ ⎥⎦ ⎩b ⎭
(A-19)
⎩M 2 ⎭ L2 ⎣

Substituting a and b into Equation A-19

⎧M1 ⎫ EIλ ⎡sin λ − λ cos λ λ − sin λ ⎤ ⎧θ1 ⎫

⎨ ⎬= ⎢ ⎨ ⎬
sin λ − λ cos λ ⎥⎦ ⎩θ 2 ⎭
(A-20)
⎩M 2 ⎭ L(2 − 2 cos λ − λ sin λ ) ⎣ λ − sin λ

49
The above derivation is valid for an axial force, P of tension. A similar process can be

used to derive the end moments of a plane frame member where P is a compression force.

Equation A-21 is the result of such a derivation.

⎧M1 ⎫ EIλ ⎡λ cosh λ − sinh λ sinh λ − λ ⎤ ⎧θ1 ⎫

⎨ ⎬= ⎢ ⎨ ⎬ (A-21)
⎩M 2 ⎭ L(2 − 2 cosh λ + λ sinh λ ) ⎣ sinh λ − λ λ cosh λ − sinh λ ⎥⎦ ⎩θ 2 ⎭

where

P
λ=L
EI

A.2
13B Derivation of the Axial Force

The derivation of the axial force in a plane frame member given in Equation 3-31

follows. The axial force in a member is equal to the cross sectional area of that member

multiplied by the stress applied to the cross section. This is represented in Equation

A-22.

EA
F= ( L − L0 ) (A-22)
L0

where

L = the length of the member in the deformed position

L0 = the length of the member in the original position

E = the modulus of elasticity of the member

A = the cross sectional area of the member

50
The axial force presented in Equation A-22 leads to an asymmetric higher-order

geometric stiffness matrix, because it does not take into to account the lengthening of the

member due to the end rotations.

In order to determine the lengthening due to the end rotations, consider the

skeletal member shown in Figure A-3 . This member may be part of a plane frame. The
X X

solid line shows the undeformed position of the member, while the dotted line represents

the deformed configuration of the member.

θ1
θ2

x
L

Figure A-3 A Plane Frame Member

The angles θ1 and θ2 are the end rotations at the left and right ends of the member,

respectively. Notice that L in this figure refers to the length of the member in its

undeformed position. To begin, the deformed shape is approximated by a cubic

polynomial.

Δ = ax 3 + bx 2 + cx + d (A-23)

The slope of the deformed member is

θ = 3ax 2 + 2bx + c (A-24)

51
By using the correct boundary conditions the coefficients a, b, c and d are determined.

The following four equations constitute Equation A-25.

θ1 + θ 2
a=
L2
2θ1 + θ 2
b=− (A-25)
L
c = θ1
d =0

The deformed length of the member is

L 2 L
⎛ dΔ ⎞
∫
0
1 + ⎜ ⎟ dx = ∫ 1 + θ 2 dx
⎝ dx ⎠ 0
(A-26)

Using the following first-order Maclaurin series approximation,

y
1+ y ≈ 1+
2

Equation A-26 becomes

L
⎛ θ2⎞
∫0 ⎜⎜⎝1 + 2 ⎟⎟⎠dx
L
⎛ 9a 2 b 4 c2 ⎞
= ∫ ⎜⎜1 + + 6abx 3 + 3acx 2 + 2b 2 x 2 + 2bcx + ⎟⎟dx
0⎝
2 2⎠

9a 2 b 5 3abL3 2b 2 L3 c2L
= L+ + + acL3 + + bcL2 + (A-27)
10 2 3 2

52
Then substituting from Equation A-25, Equation A-27 becomes

⎛ 9(θ 1 + θ 2 )2 3(θ1 + θ 2 )(2θ1 + θ 2 ) ⎞

⎜1 + − + (θ 1 + θ 2 )θ1 ⎟
L⎜⎜
10 2 ⎟
⎟
⎜⎜ + 2(2θ1 + θ 2 ) − (2θ1 + θ 2 )θ1 + θ1
2 2
⎟⎟
⎝ 3 2 ⎠
⎛ θ 2 θθ θ 2⎞
= L⎜⎜1 + 1 − 1 2 + 2 ⎟⎟ (A-28)
⎝ 15 30 15 ⎠

The strain is equal to the change in length over the original length. By subtracting L from

Equation A-28 and then dividing by L, the following strain is found

θ1 2 θ1θ 2 θ22
− + (A-29)
15 30 15

The axial force in the member is determined by multiplying the strain by the modulus of

elasticity, E, to get the stress and then multiplying the stress by the cross sectional area,

as follows

⎛ θ1 2 θ1θ 2 θ 2 2 ⎞
F = EA⎜⎜ − + ⎟ (A-30)
⎝ 15 30 15 ⎟⎠

Equation A-30 is added to Equation A-22 to get the total axial force in the member.

EA ⎛ θ1 2 θ1θ 2 θ 2 2 ⎞
F= (L − L0 ) + EA⎜⎜ − + ⎟ (A-31)
L0 ⎝ 15 30 15 ⎟⎠

Using the axial force given in Equation A-31, the derivation of the higher-order

geometric stiffness matrix leads to the symmetric matrix in Equation 3-32.

53
54
Appendix B ABAQUS Element B21
7B

The B21 element was used in all of the ABAQUS analyses presented in this

thesis. The following comes from the information found in the “Choosing a beam

element” section of the ABAQUS Analysis User’s Manual. According to the naming

convention used by ABAQUS, the B21 element is a two-dimensional beam element that

uses a linear interpolation formulation. It can be used for slender beams as well as for

non-slender or deep beams. The user may specify whether shear deformation of the

beam is considered.

55
56